diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/README.md b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/README.md
new file mode 100644
index 0000000000..5278898fd9
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/README.md
@@ -0,0 +1,197 @@
+# Record: Leaky ReLU Slope + GPTQ Reverse-Cholesky Speedup + PR #1938 (val_bpb = 1.06242)
+
+> **Note:** This README captures only the bare submission record. The full
+> set of insights from our parameter-golf run — every PR iteration we tried,
+> the hyperparameter-tuning experiments behind each design choice, and the
+> ablation results that drove our decisions — is being compiled into a
+> detailed write-up. A more detailed write-up is at: https://www.junchengbillyli.com/llm-notes.html
+
+**val_bpb (3-seed mean) = 1.06242** | σ ≈ 0.00013 | **~15.95 MB** | 8×H100 SXM | 600 s training + 600 s eval
+
+A joint effort by **Tim Shen ([@TimS-ml](https://github.com/TimS-ml))** and **Billy Li ([@lijuncheng16](https://github.com/lijuncheng16))**, with thanks to **Prof. Lin Hao (Fordham University)** for sponsoring the **8×H100 SXM** and **4×RTX 4090** compute used in this submission, **[Xingyuan Ding](https://www.linkedin.com/in/xingyuan-ding-1b1349130)** for additional experiments, **Bill (Yiyuan) Li** for meaningful discussions on tokenizers, **Lijun Yu ([@Lijun-Yu](https://github.com/Lijun-Yu))** for his invaluable insights, and **Hang Zhou ([@greyjoeyzhou](https://github.com/greyjoeyzhou))** for project discussions.
+
+## TL;DR
+
+Extends [PR #1938](https://github.com/openai/parameter-golf/pull/1938) (Billy Li & Tim Shen's *S0/PR1851 + Cap Tokenizer + LQER + Global TTT*, val_bpb=1.0713) with two algorithmically free wins:
+
+1. **Leaky ReLU squared slope 0.5 → 0.3** — `−0.00073` BPB free win; size-neutral, wallclock-neutral. (4-point sweep confirms 0.3 is the minimum — see Key Change 1.)
+2. **GPTQ reverse-Cholesky + triangular solve** instead of the standard `chol → cholesky_inverse → chol(upper)` — mathematically equivalent within fp32 ULP, **2.07–2.24× faster on RTX 4090 cuSOLVER microbench** at the GPTQ workload range. (Key Change 2.)
+
+Both are hardcoded inside `train_gpt.py` (the variant from [PR #1867](https://github.com/openai/parameter-golf/pull/1867)), which also ships **this PR's compliance-tuned defaults on top of PR #1938**: `LQER_TOP_K=1`, `GATED_ATTN_QUANT_GATE=1`, `TTT_BATCH_SIZE=16`, `PHASED_TTT_NUM_PHASES=3`, `GPTQ_RESERVE_SECONDS=16`.
+
+## Result
+
+| Seed | **Post-TTT val_bpb (final)** | Artifact bytes |
+|------|-----------------------------:|---------------:|
+| 1334 | **1.06257**                  | 15,947,664     |
+| 42   | **1.06232**                  | 15,945,920     |
+| 999  | **1.06237**                  | 15,946,532     |
+| **Mean** | **1.06242** (σ ≈ 0.00013) | **15,946,705** |
+
+
+## GPTQ reserve-time accounting
+> **(04-30):** We've noticed that several
+> leaderboard submissions appear to exceed the 10-minute training cap once the
+> full GPTQ pipeline (Hessian collection, quantization, serialize, compress) is
+> accounted for. From our own measurements, `gptq_reserve_seconds=0.5s` is
+> **far insufficient**: GPTQ Hessian collection takes **~3.5-4 s** (depending
+> on calibration batch size), GPTQ quantization itself **~10 s**, and the
+> serialize+compress step adds another **~60-70 s for Brotli** or **~90-100 s
+> for lrzip pergroup**. Among the top leaderboard PRs we surveyed, observed
+> `gptq_reserve_seconds` values range across **0.5 / 4 / 8 s**; this submission
+> uses **16 s** so that the full pipeline completes inside the 600 s training
+> cap with margin. The few-second discrepancy is unlikely to be large enough
+> to materially change the leaderboard score or ranking, but we think it's
+> worth flagging.
+
+## Key Change 1: Leaky ReLU² slope = 0.3
+
+4-point sweep at fixed seed=42 / 1.0× batch / 600 s wallclock:
+
+| slope | TTT BPB | Δ vs 0.30 |
+|------:|--------:|----------:|
+| 0.25  | 1.06151 | +0.00012  |
+| **0.30** | **1.06139** | 0     |
+| 0.35  | 1.06192 | +0.00053  |
+| 0.50 (prior baseline) | 1.06212 | +0.00073 |
+| 0.70  | 1.06267 | +0.00128  |
+
+Shallow V minimum at 0.3, size-neutral, no wallclock cost. Hardcoded in `train_gpt.py` lines 694-695 (Triton kernel) and line 910 (eager fallback).
+
+## Key Change 2: GPTQ reverse-Cholesky Hinv path
+
+Replaces
+
+```python
+Hinv = torch.cholesky_inverse(torch.linalg.cholesky(H))   # 1 chol + 2 tri-solve
+Hinv = torch.linalg.cholesky(Hinv, upper=True)            # 1 chol on dense H^{-1}
+```
+
+with the mathematically equivalent single-pass
+
+```python
+H_flip = torch.flip(H, dims=(0, 1))
+L_flip = torch.linalg.cholesky(H_flip)
+U      = torch.flip(L_flip, dims=(0, 1))
+Hinv   = torch.linalg.solve_triangular(U, eye, upper=True)
+```
+
+(The proof uses `chol(H^{-1}, upper)` uniqueness under the positive-diagonal constraint; full derivation in the authors' Stage 7 ablation note.)
+
+**RTX 4090 cuSOLVER fp32 microbench:**
+
+| n | baseline | reverse_cholesky | speedup |
+|--:|---------:|-----------------:|--------:|
+|  512 | 0.78 ms  | 0.38 ms | **2.07×** |
+| 1024 | 1.80 ms  | 0.82 ms | **2.18×** |
+| 2048 | 3.91 ms  | 1.75 ms | **2.23×** |
+| 4096 | 12.99 ms | 5.81 ms | **2.24×** |
+
+Numerics: max relative error ≤ 5.3e-7 across `n=64..2048`; artifact bytes byte-equivalent within brotli noise. Hardcoded in `train_gpt.py` lines 1870-1874.
+
+## Compliance-tuned defaults (this PR vs PR #1938)
+
+| Hparam | PR #1938 | This | Reason |
+|--------|---------:|-----:|--------|
+| `LQER_TOP_K` | 3 | **1** | top-error matrix (`tok_emb`) only; −0.00044 BPB, saves bytes |
+| `GATED_ATTN_QUANT_GATE` | 0 | **1** | int8 row-quant for `attn_gate_w`; −0.00011 BPB |
+| `TTT_BATCH_SIZE` | 64 | **16** | smaller phased batch |
+| `PHASED_TTT_NUM_PHASES` | 1 | **3** | −0.00118 BPB |
+| `GPTQ_RESERVE_SECONDS` | 4 | **16** | observed Hessian (3.5 s) + quantize (12.2 s) ≈ 16 s; required for `train+GPTQ ≤ 600 s` |
+| `LEAKY_RELU_SQ_SLOPE` (in script) | 0.5 | **0.3** | Key Change 1 |
+| GPTQ Hinv path (in script) | `cholesky_inverse + chol(upper)` | **reverse Cholesky + tri-solve** | Key Change 2 |
+
+All other hparams inherit from `train_gpt.py`'s `Hyperparameters` defaults, which match the PR #1938 envelope.
+
+## Architecture
+
+11L × 512d × 8H / 4KV, MLP 4× (2048 hidden), **LeakyReLU(0.3)²**. Partial RoPE (16/64 dims), layerwise LN scale, tied embeddings (vocab 8192, caseops-augmented), logit softcap=30.0. Depth recurrence (loops layers 3-5, ×2, activated at frac=0.35). Parallel residuals from layer 8. Skip gates. SmearGate with BOS mask. Sparse attention gates. `model_params = 35,945,671`.
+
+## Quantization
+
+Full-Hessian GPTQ + SDClip, on the reverse-Cholesky Hinv path:
+
+- **GPTQ int6** (clip_sigmas=12.85): all attn (`c_q`, `c_k`, `c_v`, `proj`) and MLP (`fc`, `proj`) weights
+- **GPTQ int7 + LQER asymmetric** (rank=4, factor int4, group_size=64): `tok_emb.weight` only (`LQER_TOP_K=1`)
+- **Dedicated int8 row-quant**: `attn_gate_w` (`GATED_ATTN_QUANT_GATE=1`)
+- **fp16 passthrough**: scalar params + small parameter weights
+- **Brotli-11** final compression → artifact ≈ 15.95 MB
+
+## TTT
+
+Phased TTT, 3 phases × 2000 prefix docs, score-first, Adam optimizer, cosine LR (peak 1e-4). LoRA rank=96 over K, MLP, O projections. `TTT_BATCH_SIZE=16`. The script's `total_eval_time` is the canonical eval timer (matches the convention used by past SOTA records).
+
+## Compliance
+
+| Cap | Limit | Observed | Margin |
+|-----|------:|---------:|-------:|
+| Artifact (decimal) | 16,000,000 bytes | 15,947,664 (max of 3 seeds) | 52,336 bytes |
+| `train + GPTQ` | 600 s | 584.1 s + 15.6 s ≈ 599.7 s | ~0.3 s |
+| `total_eval_time` | 600 s | 482.6 s / 485.6 s / 587.7 s | 12–118 s |
+
+## Dataset
+
+This submission uses the **pre-built case-op augmented FineWeb-10B** tokenization from
+[`romeerp/parameter-golf-caseops-v1`](https://huggingface.co/datasets/romeerp/parameter-golf-caseops-v1)
+(pre-built shards), the same dataset that PR #1729 / PR #1736 / PR #1851 use.
+The bijective case-op tokenizer (`fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model`,
+shipped in `tokenizers/`) and the build script (`prepare_caseops_data.py` +
+`lossless_caps.py`) are included for byte-exact rebuild, but **using the
+pre-built shards from `romeerp/parameter-golf-caseops-v1` is the recommended
+path**.
+
+## Reproducing
+
+```bash
+# Option A (recommended): use pre-built shards from HF.
+huggingface-cli download romeerp/parameter-golf-caseops-v1 \
+  --repo-type dataset \
+  --local-dir ./data/datasets/fineweb10B_sp8192_caseops/
+
+# Option B: rebuild locally with the shipped scripts: prepare_caseops_data.py
+
+# Either way, the script expects shards at
+# ./data/datasets/fineweb10B_sp8192_caseops/datasets/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/
+# (the path layout is preserved across both options).
+
+export RUN_ID=repro_seed42
+export SEED=42
+torchrun --nproc_per_node=8 --standalone train_gpt.py
+```
+
+`Hyperparameters` defaults already encode this PR's compliance-tuned envelope (this PR + b-series, on top of PR #1938); no other env exports are needed.
+
+## Builds On
+
+| Layer | Origin |
+|-------|--------|
+| **PR #1938** ([@lijuncheng16](https://github.com/lijuncheng16) & [@TimS-ml](https://github.com/TimS-ml) — *S0/PR1851 + Cap Tokenizer + LQER + Global TTT*, val_bpb=1.0713) | base submission stack |
+| **PR #1867** ([@lijuncheng16](https://github.com/lijuncheng16) & [@TimS-ml](https://github.com/TimS-ml)) | training script |
+| **PR #1851** ([@aquariouseworkman](https://github.com/aquariouseworkman) — SmearGate BOS fix + LQER asymmetric + phased TTT) | architecture / quantization |
+| PR #1797 ([@dexhunter](https://github.com/dexhunter), audit by [@cocohearts](https://github.com/cocohearts)) | SmearGate, LQER asym |
+| PR #1787 ([@nprime06](https://github.com/nprime06)) | SparseAttnGate, FusedCE, MIN_LR |
+| PR #1729 / PR #1736 ([@romeerp](https://github.com/romeerp)) | CaseOps tokenizer + phased TTT |
+| PR #1394 ([@clarkkev](https://github.com/clarkkev)) | GPTQ + SDClip + SP8192 |
+| PR #549 ([@abaybektursun](https://github.com/abaybektursun)) | Score-first TTT framework |
+
+## Acknowledgments
+
+A joint effort by **Tim Shen ([@TimS-ml](https://github.com/TimS-ml))** and **Billy Li ([@lijuncheng16](https://github.com/lijuncheng16))**.
+
+With thanks to:
+
+- **Prof. Lin Hao (Fordham University)** — for sponsoring the 8×H100 SXM and 4×RTX 4090 compute used to produce all sweep, training, and microbench results in this record.
+- **[Xingyuan Ding](https://www.linkedin.com/in/xingyuan-ding-1b1349130)** — for experiments and A100 support.
+- **Bill (Yiyuan) Li** — for meaningful discussions on tokenizers.
+- **Lijun Yu ([@Lijun-Yu](https://github.com/Lijun-Yu))** - for his invaluable insights.
+- **Hang Zhou ([@greyjoeyzhou](https://github.com/greyjoeyzhou))** — for project discussions and for the concurrent auto-research agent infrastructure.
+
+Additional credits (technique stack):
+
+- [@aquariouseworkman](https://github.com/aquariouseworkman) — PR #1851 SmearGate BOS-fix base stack
+- [@cocohearts](https://github.com/cocohearts) — SmearGate BOS audit (PR #1797)
+- [@dexhunter](https://github.com/dexhunter) — SmearGate + LQER asymmetric, phased TTT (PR #1797 / PR #1736)
+- [@romeerp](https://github.com/romeerp) — CaseOps tokenizer (PR #1729 / PR #1736)
+- [@nprime06](https://github.com/nprime06) — SparseAttnGate / FusedCE / MIN_LR (PR #1787)
+- [@abaybektursun](https://github.com/abaybektursun) — Score-first TTT framework (PR #549)
+- [@clarkkev](https://github.com/clarkkev) — GPTQ + SDClip + SP8192 (PR #1394)
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/lossless_caps.py b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/lossless_caps.py
new file mode 100644
index 0000000000..98e472f824
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/lossless_caps.py
@@ -0,0 +1,833 @@
+"""Lossless capitalization pre-encoding helpers.
+
+This module provides a narrow, reversible transform that only touches
+ASCII capital letters `A-Z`. Each uppercase ASCII letter is rewritten as
+`<sentinel><lowercase>`, where `sentinel` is a private-use Unicode
+character that is escaped by doubling if it appears literally in the
+input text.
+
+Example with the default sentinel `\\uE000`:
+
+    "The NASA Launch" -> "\\uE000the \\uE000n\\uE000a\\uE000s\\uE000a \\uE000launch"
+
+The transform is intentionally simple for v1:
+
+- lowercase ASCII letters are unchanged
+- uppercase ASCII letters become sentinel + lowercase letter
+- non-ASCII characters are left untouched
+- literal sentinel characters are escaped as sentinel + sentinel
+
+This makes the transform exactly invertible while allowing a downstream
+tokenizer to reuse lowercase subwords across case variants.
+"""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+from typing import Callable, Iterable
+
+LOSSLESS_CAPS_V1 = "lossless_caps_v1"
+LOSSLESS_CAPS_V2 = "lossless_caps_v2"
+LOSSLESS_CAPS_V3 = "lossless_caps_v3"
+LOSSLESS_CAPS_V4 = "lossless_caps_v4"
+LOSSLESS_CAPS_V5 = "lossless_caps_v5"
+LOSSLESS_CAPS_V6 = "lossless_caps_v6"
+LOSSLESS_CAPS_V7 = "lossless_caps_v7"
+LOSSLESS_CAPS_CASEOPS_V1 = "lossless_caps_caseops_v1"
+IDENTITY = "identity"
+DEFAULT_SENTINEL = "\uE000"
+DEFAULT_V2_TITLE = "\uE001"
+DEFAULT_V2_ALLCAPS = "\uE002"
+DEFAULT_V2_CAPNEXT = "\uE003"
+DEFAULT_V2_ESC = "\uE004"
+DEFAULT_V5_TITLE_MIN_LEN = 7
+DEFAULT_V6_ALLCAPS_MIN_LEN = 3
+DEFAULT_V7_ALLCAPS_MIN_LEN = 4
+
+
+class LosslessCapsError(ValueError):
+    """Raised when a transformed string is malformed."""
+
+
+def _is_ascii_upper(ch: str) -> bool:
+    return "A" <= ch <= "Z"
+
+
+def _is_ascii_lower(ch: str) -> bool:
+    return "a" <= ch <= "z"
+
+
+def _is_ascii_alpha(ch: str) -> bool:
+    return _is_ascii_lower(ch) or _is_ascii_upper(ch)
+
+
+def _validate_distinct_single_chars(*chars: str) -> None:
+    if any(len(ch) != 1 for ch in chars):
+        raise ValueError("all control characters must be exactly one character")
+    if len(set(chars)) != len(chars):
+        raise ValueError("control characters must be distinct")
+
+
+def encode_lossless_caps_v1(text: str, *, sentinel: str = DEFAULT_SENTINEL) -> str:
+    """Encode ASCII capitals reversibly using a one-character sentinel."""
+    if len(sentinel) != 1:
+        raise ValueError("sentinel must be exactly one character")
+    out: list[str] = []
+    for ch in text:
+        if ch == sentinel:
+            out.append(sentinel)
+            out.append(sentinel)
+        elif _is_ascii_upper(ch):
+            out.append(sentinel)
+            out.append(ch.lower())
+        else:
+            out.append(ch)
+    return "".join(out)
+
+
+def decode_lossless_caps_v1(text: str, *, sentinel: str = DEFAULT_SENTINEL) -> str:
+    """Decode the `lossless_caps_v1` transform back to the original text."""
+    if len(sentinel) != 1:
+        raise ValueError("sentinel must be exactly one character")
+    out: list[str] = []
+    i = 0
+    n = len(text)
+    while i < n:
+        ch = text[i]
+        if ch != sentinel:
+            out.append(ch)
+            i += 1
+            continue
+        if i + 1 >= n:
+            raise LosslessCapsError("dangling capitalization sentinel at end of string")
+        nxt = text[i + 1]
+        if nxt == sentinel:
+            out.append(sentinel)
+        elif _is_ascii_lower(nxt):
+            out.append(nxt.upper())
+        else:
+            raise LosslessCapsError(
+                f"invalid sentinel escape sequence {sentinel + nxt!r}; "
+                "expected doubled sentinel or sentinel + lowercase ASCII letter"
+            )
+        i += 2
+    return "".join(out)
+
+
+def encode_lossless_caps_v2(
+    text: str,
+    *,
+    title: str = DEFAULT_V2_TITLE,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    capnext: str = DEFAULT_V2_CAPNEXT,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Encode ASCII word capitalization with cheap word-level markers.
+
+    Rules over maximal ASCII alphabetic runs:
+    - lowercase words stay unchanged
+    - TitleCase words become `title + lowercase(word)`
+    - ALLCAPS words become `allcaps + lowercase(word)`
+    - mixed-case words use:
+      - optional `title` when the first letter is uppercase
+      - `capnext + lowercase(letter)` for subsequent uppercase letters
+    - literal control characters are escaped as `esc + literal`
+    """
+    _validate_distinct_single_chars(title, allcaps, capnext, esc)
+    controls = {title, allcaps, capnext, esc}
+    out: list[str] = []
+    i = 0
+    n = len(text)
+    while i < n:
+        ch = text[i]
+        if ch in controls:
+            out.append(esc)
+            out.append(ch)
+            i += 1
+            continue
+        if not _is_ascii_alpha(ch):
+            out.append(ch)
+            i += 1
+            continue
+
+        j = i + 1
+        while j < n and _is_ascii_alpha(text[j]):
+            j += 1
+        word = text[i:j]
+        lower_word = word.lower()
+
+        if word.islower():
+            out.append(word)
+        elif len(word) >= 2 and word.isupper():
+            out.append(allcaps)
+            out.append(lower_word)
+        elif _is_ascii_upper(word[0]) and word[1:].islower():
+            out.append(title)
+            out.append(lower_word)
+        else:
+            if _is_ascii_upper(word[0]):
+                out.append(title)
+            out.append(lower_word[0])
+            for orig_ch, lower_ch in zip(word[1:], lower_word[1:], strict=True):
+                if _is_ascii_upper(orig_ch):
+                    out.append(capnext)
+                out.append(lower_ch)
+        i = j
+    return "".join(out)
+
+
+def decode_lossless_caps_v2(
+    text: str,
+    *,
+    title: str = DEFAULT_V2_TITLE,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    capnext: str = DEFAULT_V2_CAPNEXT,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Decode the `lossless_caps_v2` transform back to the original text."""
+    _validate_distinct_single_chars(title, allcaps, capnext, esc)
+    out: list[str] = []
+    pending_escape = False
+    pending_word_mode: str | None = None
+    active_allcaps = False
+    pending_capnext = False
+    in_ascii_word = False
+
+    for ch in text:
+        if pending_escape:
+            if pending_word_mode is not None and not _is_ascii_alpha(ch):
+                raise LosslessCapsError("escaped control char cannot satisfy pending word capitalization mode")
+            out.append(ch)
+            pending_escape = False
+            if _is_ascii_alpha(ch):
+                in_ascii_word = True
+            else:
+                in_ascii_word = False
+                active_allcaps = False
+            continue
+
+        if ch == esc:
+            pending_escape = True
+            continue
+        if ch == title:
+            if pending_word_mode is not None or in_ascii_word or pending_capnext:
+                raise LosslessCapsError("invalid title marker placement")
+            pending_word_mode = "title"
+            continue
+        if ch == allcaps:
+            if pending_word_mode is not None or in_ascii_word or pending_capnext:
+                raise LosslessCapsError("invalid allcaps marker placement")
+            pending_word_mode = "allcaps"
+            continue
+        if ch == capnext:
+            if pending_capnext:
+                raise LosslessCapsError("duplicate capnext marker")
+            pending_capnext = True
+            continue
+
+        if _is_ascii_alpha(ch):
+            at_word_start = not in_ascii_word
+            if at_word_start:
+                if pending_word_mode == "allcaps":
+                    out.append(ch.upper())
+                    active_allcaps = True
+                elif pending_word_mode == "title":
+                    out.append(ch.upper())
+                elif pending_capnext:
+                    out.append(ch.upper())
+                else:
+                    out.append(ch)
+                pending_word_mode = None
+                pending_capnext = False
+                in_ascii_word = True
+                continue
+
+            if pending_word_mode is not None:
+                raise LosslessCapsError("word capitalization marker leaked into the middle of a word")
+            if active_allcaps:
+                out.append(ch.upper())
+            elif pending_capnext:
+                out.append(ch.upper())
+            else:
+                out.append(ch)
+            pending_capnext = False
+            continue
+
+        if pending_word_mode is not None or pending_capnext:
+            raise LosslessCapsError("capitalization marker not followed by an ASCII letter")
+        out.append(ch)
+        in_ascii_word = False
+        active_allcaps = False
+
+    if pending_escape:
+        raise LosslessCapsError("dangling escape marker at end of string")
+    if pending_word_mode is not None or pending_capnext:
+        raise LosslessCapsError("dangling capitalization marker at end of string")
+    return "".join(out)
+
+
+def encode_lossless_caps_v3(
+    text: str,
+    *,
+    title: str = DEFAULT_V2_TITLE,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Encode only common word-level capitalization patterns.
+
+    Rules over maximal ASCII alphabetic runs:
+    - lowercase words stay unchanged
+    - TitleCase words become `title + lowercase(word)`
+    - ALLCAPS words become `allcaps + lowercase(word)`
+    - all other mixed-case words are left unchanged
+    - literal control characters are escaped as `esc + literal`
+    """
+    _validate_distinct_single_chars(title, allcaps, esc)
+    controls = {title, allcaps, esc}
+    out: list[str] = []
+    i = 0
+    n = len(text)
+    while i < n:
+        ch = text[i]
+        if ch in controls:
+            out.append(esc)
+            out.append(ch)
+            i += 1
+            continue
+        if not _is_ascii_alpha(ch):
+            out.append(ch)
+            i += 1
+            continue
+
+        j = i + 1
+        while j < n and _is_ascii_alpha(text[j]):
+            j += 1
+        word = text[i:j]
+
+        if word.islower():
+            out.append(word)
+        elif len(word) >= 2 and word.isupper():
+            out.append(allcaps)
+            out.append(word.lower())
+        elif _is_ascii_upper(word[0]) and word[1:].islower():
+            out.append(title)
+            out.append(word.lower())
+        else:
+            out.append(word)
+        i = j
+    return "".join(out)
+
+
+def decode_lossless_caps_v3(
+    text: str,
+    *,
+    title: str = DEFAULT_V2_TITLE,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Decode the `lossless_caps_v3` transform back to the original text."""
+    _validate_distinct_single_chars(title, allcaps, esc)
+    out: list[str] = []
+    pending_escape = False
+    pending_word_mode: str | None = None
+    active_allcaps = False
+    in_ascii_word = False
+
+    for ch in text:
+        if pending_escape:
+            if pending_word_mode is not None and not _is_ascii_alpha(ch):
+                raise LosslessCapsError("escaped control char cannot satisfy pending word capitalization mode")
+            out.append(ch)
+            pending_escape = False
+            if _is_ascii_alpha(ch):
+                in_ascii_word = True
+            else:
+                in_ascii_word = False
+                active_allcaps = False
+            continue
+
+        if ch == esc:
+            pending_escape = True
+            continue
+        if ch == title:
+            if pending_word_mode is not None or in_ascii_word:
+                raise LosslessCapsError("invalid title marker placement")
+            pending_word_mode = "title"
+            continue
+        if ch == allcaps:
+            if pending_word_mode is not None or in_ascii_word:
+                raise LosslessCapsError("invalid allcaps marker placement")
+            pending_word_mode = "allcaps"
+            continue
+
+        if _is_ascii_alpha(ch):
+            at_word_start = not in_ascii_word
+            if at_word_start:
+                if pending_word_mode == "allcaps":
+                    out.append(ch.upper())
+                    active_allcaps = True
+                elif pending_word_mode == "title":
+                    out.append(ch.upper())
+                else:
+                    out.append(ch)
+                pending_word_mode = None
+                in_ascii_word = True
+                continue
+
+            if pending_word_mode is not None:
+                raise LosslessCapsError("word capitalization marker leaked into the middle of a word")
+            out.append(ch.upper() if active_allcaps else ch)
+            continue
+
+        if pending_word_mode is not None:
+            raise LosslessCapsError("capitalization marker not followed by an ASCII letter")
+        out.append(ch)
+        in_ascii_word = False
+        active_allcaps = False
+
+    if pending_escape:
+        raise LosslessCapsError("dangling escape marker at end of string")
+    if pending_word_mode is not None:
+        raise LosslessCapsError("dangling capitalization marker at end of string")
+    return "".join(out)
+
+
+def encode_lossless_caps_v4(
+    text: str,
+    *,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Encode only ALLCAPS ASCII words, leaving all other case untouched."""
+    _validate_distinct_single_chars(allcaps, esc)
+    controls = {allcaps, esc}
+    out: list[str] = []
+    i = 0
+    n = len(text)
+    while i < n:
+        ch = text[i]
+        if ch in controls:
+            out.append(esc)
+            out.append(ch)
+            i += 1
+            continue
+        if not _is_ascii_alpha(ch):
+            out.append(ch)
+            i += 1
+            continue
+        j = i + 1
+        while j < n and _is_ascii_alpha(text[j]):
+            j += 1
+        word = text[i:j]
+        if len(word) >= 2 and word.isupper():
+            out.append(allcaps)
+            out.append(word.lower())
+        else:
+            out.append(word)
+        i = j
+    return "".join(out)
+
+
+def decode_lossless_caps_v4(
+    text: str,
+    *,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Decode the `lossless_caps_v4` transform back to the original text."""
+    _validate_distinct_single_chars(allcaps, esc)
+    out: list[str] = []
+    pending_escape = False
+    pending_allcaps = False
+    in_ascii_word = False
+    active_allcaps = False
+
+    for ch in text:
+        if pending_escape:
+            if pending_allcaps and not _is_ascii_alpha(ch):
+                raise LosslessCapsError("escaped control char cannot satisfy pending allcaps mode")
+            out.append(ch)
+            pending_escape = False
+            if _is_ascii_alpha(ch):
+                in_ascii_word = True
+            else:
+                in_ascii_word = False
+                active_allcaps = False
+            continue
+
+        if ch == esc:
+            pending_escape = True
+            continue
+        if ch == allcaps:
+            if pending_allcaps or in_ascii_word:
+                raise LosslessCapsError("invalid allcaps marker placement")
+            pending_allcaps = True
+            continue
+
+        if _is_ascii_alpha(ch):
+            if not in_ascii_word:
+                active_allcaps = pending_allcaps
+                pending_allcaps = False
+                in_ascii_word = True
+            out.append(ch.upper() if active_allcaps else ch)
+            continue
+
+        if pending_allcaps:
+            raise LosslessCapsError("allcaps marker not followed by an ASCII letter")
+        out.append(ch)
+        in_ascii_word = False
+        active_allcaps = False
+
+    if pending_escape:
+        raise LosslessCapsError("dangling escape marker at end of string")
+    if pending_allcaps:
+        raise LosslessCapsError("dangling allcaps marker at end of string")
+    return "".join(out)
+
+
+def encode_lossless_caps_v5(
+    text: str,
+    *,
+    title: str = DEFAULT_V2_TITLE,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+    title_min_len: int = DEFAULT_V5_TITLE_MIN_LEN,
+) -> str:
+    """Encode ALLCAPS words and only sufficiently long TitleCase words."""
+    _validate_distinct_single_chars(title, allcaps, esc)
+    controls = {title, allcaps, esc}
+    out: list[str] = []
+    i = 0
+    n = len(text)
+    while i < n:
+        ch = text[i]
+        if ch in controls:
+            out.append(esc)
+            out.append(ch)
+            i += 1
+            continue
+        if not _is_ascii_alpha(ch):
+            out.append(ch)
+            i += 1
+            continue
+        j = i + 1
+        while j < n and _is_ascii_alpha(text[j]):
+            j += 1
+        word = text[i:j]
+        if len(word) >= 2 and word.isupper():
+            out.append(allcaps)
+            out.append(word.lower())
+        elif len(word) >= title_min_len and _is_ascii_upper(word[0]) and word[1:].islower():
+            out.append(title)
+            out.append(word.lower())
+        else:
+            out.append(word)
+        i = j
+    return "".join(out)
+
+
+def decode_lossless_caps_v5(
+    text: str,
+    *,
+    title: str = DEFAULT_V2_TITLE,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Decode the `lossless_caps_v5` transform back to the original text."""
+    return decode_lossless_caps_v3(text, title=title, allcaps=allcaps, esc=esc)
+
+
+def encode_lossless_caps_v6(
+    text: str,
+    *,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+    allcaps_min_len: int = DEFAULT_V6_ALLCAPS_MIN_LEN,
+) -> str:
+    """Encode only ALLCAPS words with length >= allcaps_min_len."""
+    _validate_distinct_single_chars(allcaps, esc)
+    controls = {allcaps, esc}
+    out: list[str] = []
+    i = 0
+    n = len(text)
+    while i < n:
+        ch = text[i]
+        if ch in controls:
+            out.append(esc)
+            out.append(ch)
+            i += 1
+            continue
+        if not _is_ascii_alpha(ch):
+            out.append(ch)
+            i += 1
+            continue
+        j = i + 1
+        while j < n and _is_ascii_alpha(text[j]):
+            j += 1
+        word = text[i:j]
+        if len(word) >= allcaps_min_len and word.isupper():
+            out.append(allcaps)
+            out.append(word.lower())
+        else:
+            out.append(word)
+        i = j
+    return "".join(out)
+
+
+def decode_lossless_caps_v6(
+    text: str,
+    *,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Decode the `lossless_caps_v6` transform back to the original text."""
+    return decode_lossless_caps_v4(text, allcaps=allcaps, esc=esc)
+
+
+def encode_lossless_caps_v7(
+    text: str,
+    *,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+    allcaps_min_len: int = DEFAULT_V7_ALLCAPS_MIN_LEN,
+) -> str:
+    """Encode only ALLCAPS words with length >= 4."""
+    return encode_lossless_caps_v6(
+        text,
+        allcaps=allcaps,
+        esc=esc,
+        allcaps_min_len=allcaps_min_len,
+    )
+
+
+def decode_lossless_caps_v7(
+    text: str,
+    *,
+    allcaps: str = DEFAULT_V2_ALLCAPS,
+    esc: str = DEFAULT_V2_ESC,
+) -> str:
+    """Decode the `lossless_caps_v7` transform back to the original text."""
+    return decode_lossless_caps_v6(text, allcaps=allcaps, esc=esc)
+
+
+def get_text_transform(name: str | None) -> Callable[[str], str]:
+    """Return the forward text transform for the given config name."""
+    normalized = IDENTITY if name in {None, "", IDENTITY} else str(name)
+    if normalized == IDENTITY:
+        return lambda text: text
+    if normalized == LOSSLESS_CAPS_V1:
+        return encode_lossless_caps_v1
+    if normalized == LOSSLESS_CAPS_V2:
+        return encode_lossless_caps_v2
+    if normalized == LOSSLESS_CAPS_V3:
+        return encode_lossless_caps_v3
+    if normalized == LOSSLESS_CAPS_V4:
+        return encode_lossless_caps_v4
+    if normalized == LOSSLESS_CAPS_V5:
+        return encode_lossless_caps_v5
+    if normalized == LOSSLESS_CAPS_V6:
+        return encode_lossless_caps_v6
+    if normalized == LOSSLESS_CAPS_V7:
+        return encode_lossless_caps_v7
+    if normalized == LOSSLESS_CAPS_CASEOPS_V1:
+        return encode_lossless_caps_v2
+    raise ValueError(f"unsupported text_transform={name!r}")
+
+
+def get_text_inverse_transform(name: str | None) -> Callable[[str], str]:
+    """Return the inverse transform for the given config name."""
+    normalized = IDENTITY if name in {None, "", IDENTITY} else str(name)
+    if normalized == IDENTITY:
+        return lambda text: text
+    if normalized == LOSSLESS_CAPS_V1:
+        return decode_lossless_caps_v1
+    if normalized == LOSSLESS_CAPS_V2:
+        return decode_lossless_caps_v2
+    if normalized == LOSSLESS_CAPS_V3:
+        return decode_lossless_caps_v3
+    if normalized == LOSSLESS_CAPS_V4:
+        return decode_lossless_caps_v4
+    if normalized == LOSSLESS_CAPS_V5:
+        return decode_lossless_caps_v5
+    if normalized == LOSSLESS_CAPS_V6:
+        return decode_lossless_caps_v6
+    if normalized == LOSSLESS_CAPS_V7:
+        return decode_lossless_caps_v7
+    if normalized == LOSSLESS_CAPS_CASEOPS_V1:
+        return decode_lossless_caps_v2
+    raise ValueError(f"unsupported text_transform={name!r}")
+
+
+def normalize_text_transform_name(name: str | None) -> str:
+    """Normalize empty/None transform names to the identity transform."""
+    return IDENTITY if name in {None, "", IDENTITY} else str(name)
+
+
+def get_text_transform_control_symbols(name: str | None) -> list[str]:
+    """Return reserved control symbols used by a transform, if any."""
+    normalized = normalize_text_transform_name(name)
+    if normalized == IDENTITY:
+        return []
+    if normalized == LOSSLESS_CAPS_V1:
+        return [DEFAULT_SENTINEL]
+    if normalized == LOSSLESS_CAPS_V2:
+        return [DEFAULT_V2_TITLE, DEFAULT_V2_ALLCAPS, DEFAULT_V2_CAPNEXT, DEFAULT_V2_ESC]
+    if normalized == LOSSLESS_CAPS_CASEOPS_V1:
+        return [DEFAULT_V2_TITLE, DEFAULT_V2_ALLCAPS, DEFAULT_V2_CAPNEXT, DEFAULT_V2_ESC]
+    if normalized in {LOSSLESS_CAPS_V3, LOSSLESS_CAPS_V5}:
+        return [DEFAULT_V2_TITLE, DEFAULT_V2_ALLCAPS, DEFAULT_V2_ESC]
+    if normalized in {LOSSLESS_CAPS_V4, LOSSLESS_CAPS_V6, LOSSLESS_CAPS_V7}:
+        return [DEFAULT_V2_ALLCAPS, DEFAULT_V2_ESC]
+    raise ValueError(f"unsupported text_transform={name!r}")
+
+
+def infer_text_transform_from_manifest(tokenizer_path: str | Path) -> str:
+    """Best-effort lookup of a tokenizer's text transform from a local manifest."""
+    tokenizer_path = Path(tokenizer_path).expanduser().resolve()
+    manifest_candidates = [
+        tokenizer_path.parent.parent / "manifest.json",
+        tokenizer_path.parent / "manifest.json",
+    ]
+    for manifest_path in manifest_candidates:
+        if not manifest_path.is_file():
+            continue
+        try:
+            payload = json.loads(manifest_path.read_text(encoding="utf-8"))
+        except (OSError, json.JSONDecodeError):
+            continue
+        tokenizers = payload.get("tokenizers")
+        if not isinstance(tokenizers, list):
+            continue
+        for tokenizer_meta in tokenizers:
+            if not isinstance(tokenizer_meta, dict):
+                continue
+            model_path = tokenizer_meta.get("model_path") or tokenizer_meta.get("path")
+            if not model_path:
+                continue
+            candidate = (manifest_path.parent / str(model_path)).resolve()
+            if candidate == tokenizer_path:
+                return normalize_text_transform_name(tokenizer_meta.get("text_transform"))
+    return IDENTITY
+
+
+def surface_piece_original_byte_counts(
+    surfaces: Iterable[str],
+    *,
+    text_transform_name: str | None = None,
+    sentinel: str = DEFAULT_SENTINEL,
+) -> list[int]:
+    """Return exact original UTF-8 byte counts contributed by each surface piece.
+
+    `surfaces` must be the exact decoded text fragments emitted by SentencePiece
+    in order, e.g. `piece.surface` from `encode_as_immutable_proto`.
+    """
+    normalized = normalize_text_transform_name(text_transform_name)
+    if normalized == IDENTITY:
+        return [len(surface.encode("utf-8")) for surface in surfaces]
+    if normalized == LOSSLESS_CAPS_V1:
+        if len(sentinel) != 1:
+            raise ValueError("sentinel must be exactly one character")
+        sentinel_bytes = len(sentinel.encode("utf-8"))
+        pending_sentinel = False
+        counts: list[int] = []
+        for surface in surfaces:
+            piece_bytes = 0
+            for ch in surface:
+                if pending_sentinel:
+                    if ch == sentinel:
+                        piece_bytes += sentinel_bytes
+                    elif _is_ascii_lower(ch):
+                        piece_bytes += 1
+                    else:
+                        raise LosslessCapsError(
+                            f"invalid continuation {ch!r} after capitalization sentinel"
+                        )
+                    pending_sentinel = False
+                    continue
+                if ch == sentinel:
+                    pending_sentinel = True
+                else:
+                    piece_bytes += len(ch.encode("utf-8"))
+            counts.append(piece_bytes)
+        if pending_sentinel:
+            raise LosslessCapsError("dangling capitalization sentinel across piece boundary")
+        return counts
+    if normalized not in {LOSSLESS_CAPS_V2, LOSSLESS_CAPS_V3, LOSSLESS_CAPS_V4, LOSSLESS_CAPS_V5, LOSSLESS_CAPS_V6, LOSSLESS_CAPS_V7, LOSSLESS_CAPS_CASEOPS_V1}:
+        raise ValueError(f"unsupported text_transform={text_transform_name!r}")
+
+    title = DEFAULT_V2_TITLE
+    allcaps = DEFAULT_V2_ALLCAPS
+    capnext = DEFAULT_V2_CAPNEXT
+    esc = DEFAULT_V2_ESC
+    if normalized in {LOSSLESS_CAPS_V2, LOSSLESS_CAPS_CASEOPS_V1}:
+        _validate_distinct_single_chars(title, allcaps, capnext, esc)
+    elif normalized in {LOSSLESS_CAPS_V4, LOSSLESS_CAPS_V6, LOSSLESS_CAPS_V7}:
+        _validate_distinct_single_chars(allcaps, esc)
+    else:
+        _validate_distinct_single_chars(title, allcaps, esc)
+    pending_escape = False
+    pending_word_mode: str | None = None
+    active_allcaps = False
+    pending_capnext = False
+    in_ascii_word = False
+    counts: list[int] = []
+    for surface in surfaces:
+        piece_bytes = 0
+        for ch in surface:
+            if pending_escape:
+                if pending_word_mode is not None and not _is_ascii_alpha(ch):
+                    raise LosslessCapsError("escaped control char cannot satisfy pending word capitalization mode")
+                piece_bytes += len(ch.encode("utf-8"))
+                pending_escape = False
+                if _is_ascii_alpha(ch):
+                    in_ascii_word = True
+                else:
+                    in_ascii_word = False
+                    active_allcaps = False
+                continue
+            if ch == esc:
+                pending_escape = True
+                continue
+            if normalized in {LOSSLESS_CAPS_V2, LOSSLESS_CAPS_V3, LOSSLESS_CAPS_V5, LOSSLESS_CAPS_CASEOPS_V1} and ch == title:
+                if pending_word_mode is not None or in_ascii_word or pending_capnext:
+                    raise LosslessCapsError("invalid title marker placement")
+                pending_word_mode = "title"
+                continue
+            if ch == allcaps:
+                if pending_word_mode is not None or in_ascii_word or pending_capnext:
+                    raise LosslessCapsError("invalid allcaps marker placement")
+                pending_word_mode = "allcaps"
+                continue
+            if normalized in {LOSSLESS_CAPS_V2, LOSSLESS_CAPS_CASEOPS_V1} and ch == capnext:
+                if pending_capnext:
+                    raise LosslessCapsError("duplicate capnext marker")
+                pending_capnext = True
+                continue
+
+            if _is_ascii_alpha(ch):
+                at_word_start = not in_ascii_word
+                if at_word_start:
+                    piece_bytes += 1
+                    active_allcaps = pending_word_mode == "allcaps"
+                    pending_word_mode = None
+                    pending_capnext = False
+                    in_ascii_word = True
+                    continue
+                if pending_word_mode is not None:
+                    raise LosslessCapsError("word capitalization marker leaked into the middle of a word")
+                piece_bytes += 1
+                pending_capnext = False
+                continue
+
+            if pending_word_mode is not None or pending_capnext:
+                raise LosslessCapsError("capitalization marker not followed by an ASCII letter")
+            piece_bytes += len(ch.encode("utf-8"))
+            in_ascii_word = False
+            active_allcaps = False
+        counts.append(piece_bytes)
+    if pending_escape:
+        raise LosslessCapsError("dangling escape marker across piece boundary")
+    if pending_word_mode is not None or pending_capnext:
+        raise LosslessCapsError("dangling capitalization marker across piece boundary")
+    return counts
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/prepare_caseops_data.py b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/prepare_caseops_data.py
new file mode 100644
index 0000000000..5c3f13e69c
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/prepare_caseops_data.py
@@ -0,0 +1,177 @@
+"""Prepare CaseOps-tokenized FineWeb shards + per-token byte sidecar.
+
+CaseOps (``lossless_caps_caseops_v1``) is a bijective, character-level text
+transform that introduces four operator tokens in place of explicit
+capitalization: TITLE, ALLCAPS, CAPNEXT, ESC. The transform is fully
+reversible — no information is lost relative to the untransformed UTF-8
+text, so BPB stays computable on TRUE byte counts.
+
+Forward pipeline:
+  1. Read the canonical FineWeb-10B doc stream (``docs_selected.jsonl``
+     produced by ``data/download_hf_docs_and_tokenize.py`` in the root repo).
+  2. Apply ``encode_lossless_caps_v2`` (the caseops_v1 alias) to each doc.
+  3. Tokenize with the shipped SP model
+     ``tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model``
+     (reserves TITLE/ALLCAPS/CAPNEXT/ESC + sentinel as user_defined_symbols).
+  4. Write uint16 train/val shards (``fineweb_{train,val}_XXXXXX.bin``).
+  5. For the VAL stream only, emit per-token byte sidecar shards
+     (``fineweb_val_bytes_XXXXXX.bin``, uint16 parallel arrays) that record
+     each token's ORIGINAL pre-transform UTF-8 byte count. BPB is computed
+     from these canonical bytes so the score is on the untransformed text
+     (not the transformed representation).
+
+Output layout — matches what ``train_gpt.py`` expects under
+``DATA_DIR=./data`` with ``CASEOPS_ENABLED=1``:
+
+    data/datasets/fineweb10B_sp8192_caseops/datasets/
+      tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model
+      datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/
+        fineweb_train_000000.bin
+        fineweb_train_000001.bin
+        ...
+        fineweb_val_000000.bin
+        fineweb_val_bytes_000000.bin
+
+Usage:
+
+    python3 prepare_caseops_data.py \\
+        --docs ./fineweb10B_raw/docs_selected.jsonl \\
+        --out  ./data/datasets/fineweb10B_sp8192_caseops/datasets \\
+        --sp   ./tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model
+
+Requirements: sentencepiece, numpy. CPU-only. Runs once; reused across seeds.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import pathlib
+import struct
+import sys
+
+import numpy as np
+import sentencepiece as spm
+
+# Local import — lossless_caps.py ships next to this script.
+sys.path.insert(0, str(pathlib.Path(__file__).resolve().parent))
+from lossless_caps import (  # noqa: E402
+    LOSSLESS_CAPS_CASEOPS_V1,
+    encode_lossless_caps_v2,
+    surface_piece_original_byte_counts,
+)
+
+
+SHARD_MAGIC = 20240520
+SHARD_VERSION = 1
+SHARD_TOKENS = 10_000_000  # tokens per shard — matches the main pipeline
+BOS_ID = 1  # SP model's <s> control token; train_gpt.py:_find_docs requires BOS per doc
+
+
+def _write_shard(out_path: pathlib.Path, arr: np.ndarray) -> None:
+    """Write a uint16 shard in the standard header-prefixed format."""
+    assert arr.dtype == np.uint16
+    header = np.zeros(256, dtype=np.int32)
+    header[0] = SHARD_MAGIC
+    header[1] = SHARD_VERSION
+    header[2] = int(arr.size)
+    with out_path.open("wb") as fh:
+        fh.write(header.tobytes())
+        fh.write(arr.tobytes())
+
+
+def _iter_docs(docs_path: pathlib.Path):
+    """Yield doc strings from a jsonl file (one json object per line)."""
+    with docs_path.open("r", encoding="utf-8") as fh:
+        for line in fh:
+            line = line.strip()
+            if not line:
+                continue
+            obj = json.loads(line)
+            # Support both {"text": ...} and raw strings.
+            yield obj["text"] if isinstance(obj, dict) else obj
+
+
+def _token_original_byte_counts(
+    sp: spm.SentencePieceProcessor,
+    original_text: str,
+    transformed_text: str,
+) -> np.ndarray:
+    """Per-token canonical (pre-transform) UTF-8 byte counts.
+
+    Delegates to ``surface_piece_original_byte_counts`` in ``lossless_caps.py``
+    — the canonical exporter used by the PR #1729 / HF-hosted CaseOps dataset.
+    Operator pieces (U+E001..U+E004) contribute 0 original bytes; letter pieces
+    contribute their pre-transform UTF-8 byte count.
+    """
+    proto = sp.encode_as_immutable_proto(transformed_text)
+    byte_counts = surface_piece_original_byte_counts(
+        (piece.surface for piece in proto.pieces),
+        text_transform_name=LOSSLESS_CAPS_CASEOPS_V1,
+    )
+    return np.asarray(list(byte_counts), dtype=np.uint16)
+
+
+def main() -> None:
+    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
+    ap.add_argument("--docs", required=True, type=pathlib.Path, help="Path to docs_selected.jsonl")
+    ap.add_argument("--out",  required=True, type=pathlib.Path, help="Output datasets dir")
+    ap.add_argument("--sp",   required=True, type=pathlib.Path, help="Path to CaseOps SP model")
+    ap.add_argument("--val-docs", type=int, default=10_000, help="Validation docs count")
+    args = ap.parse_args()
+
+    sp = spm.SentencePieceProcessor(model_file=str(args.sp))
+    print(f"loaded sp: vocab={sp.vocab_size()}", flush=True)
+
+    train_out = args.out / "datasets" / "fineweb10B_sp8192_lossless_caps_caseops_v1_reserved"
+    train_out.mkdir(parents=True, exist_ok=True)
+
+    val_buf_tokens: list[int] = []
+    val_buf_bytes: list[int] = []
+    train_buf: list[int] = []
+    val_written = 0
+    train_written = 0
+    n_docs = 0
+
+    for text in _iter_docs(args.docs):
+        transformed = encode_lossless_caps_v2(text)
+        token_ids = [BOS_ID] + sp.encode(transformed, out_type=int)
+        if n_docs < args.val_docs:
+            # Validation doc — also compute byte sidecar
+            byte_counts = _token_original_byte_counts(sp, text, transformed)
+            val_buf_tokens.extend(token_ids)
+            val_buf_bytes.append(0)  # BOS contributes 0 original bytes
+            val_buf_bytes.extend(int(b) for b in byte_counts)
+            if len(val_buf_tokens) >= SHARD_TOKENS:
+                _write_shard(train_out / f"fineweb_val_{val_written:06d}.bin",
+                             np.array(val_buf_tokens[:SHARD_TOKENS], dtype=np.uint16))
+                _write_shard(train_out / f"fineweb_val_bytes_{val_written:06d}.bin",
+                             np.array(val_buf_bytes[:SHARD_TOKENS], dtype=np.uint16))
+                val_buf_tokens = val_buf_tokens[SHARD_TOKENS:]
+                val_buf_bytes = val_buf_bytes[SHARD_TOKENS:]
+                val_written += 1
+        else:
+            train_buf.extend(token_ids)
+            if len(train_buf) >= SHARD_TOKENS:
+                _write_shard(train_out / f"fineweb_train_{train_written:06d}.bin",
+                             np.array(train_buf[:SHARD_TOKENS], dtype=np.uint16))
+                train_buf = train_buf[SHARD_TOKENS:]
+                train_written += 1
+        n_docs += 1
+        if n_docs % 10_000 == 0:
+            print(f"  processed {n_docs} docs  train_shards={train_written}  val_shards={val_written}", flush=True)
+
+    # Flush tail buffers into final (possibly short) shards.
+    if val_buf_tokens:
+        _write_shard(train_out / f"fineweb_val_{val_written:06d}.bin",
+                     np.array(val_buf_tokens, dtype=np.uint16))
+        _write_shard(train_out / f"fineweb_val_bytes_{val_written:06d}.bin",
+                     np.array(val_buf_bytes, dtype=np.uint16))
+    if train_buf:
+        _write_shard(train_out / f"fineweb_train_{train_written:06d}.bin",
+                     np.array(train_buf, dtype=np.uint16))
+
+    print(f"done. docs={n_docs} train_shards={train_written + (1 if train_buf else 0)} val_shards={val_written + (1 if val_buf_tokens else 0)}")
+
+
+if __name__ == "__main__":
+    main()
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/submission.json b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/submission.json
new file mode 100644
index 0000000000..162864c629
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/submission.json
@@ -0,0 +1,37 @@
+{
+  "author": "Tim Shen (@TimS-ml) & Billy Li (@lijuncheng16)",
+  "github_id": "TimS-ml",
+  "name": "Record: Leaky ReLU Slope + GPTQ Reverse-Cholesky Speedup + PR #1938 (val_bpb = 1.06242)",
+  "date": "2026-04-29",
+  "track": "10min_16mb",
+  "val_bpb": 1.06242,
+  "val_bpb_std": 0.00013,
+  "seeds": [1334, 42, 999],
+  "seed_results": {
+    "1334": {"val_bpb": 1.06257073, "artifact_bytes": 15947664},
+    "42":   {"val_bpb": 1.06232455, "artifact_bytes": 15945920},
+    "999":  {"val_bpb": 1.06236517, "artifact_bytes": 15946532}
+  },
+  "hardware": "8xH100 80GB SXM",
+  "pytorch_version": "2.9.1+cu128",
+  "technique_summary": "SP8192/CaseOps + 11L GQA + LeakyReLU(0.3)\u00b2 + Depth Recurrence (L3-5 \u00d72) + Parallel Residuals (L8+) + SmearGate + SparseAttnGate + GPTQ int6 Reverse-Cholesky + LQER asym (top-1) + Int8 Attn Gate + Phased TTT (3ph) + Brotli",
+  "compliance": {
+    "train_under_600s": true,
+    "artifact_under_16mb": true,
+    "eval_under_600s": true,
+    "three_seeds": true,
+    "score_first_ttt": true,
+    "smeargate_bos_fix": true
+  },
+  "attribution": {
+    "pr1938": "@lijuncheng16; @TimS-ml (PR #1938)",
+    "pr1867": "@lijuncheng16; @TimS-ml (PR #1867)",
+    "pr1851": "@aquariouseworkman (PR #1851)",
+    "smeargate_bos_audit": "@cocohearts (PR #1797)",
+    "lqer_asym": "@dexhunter (PR #1797)",
+    "caseops_tokenizer": "@romeerp (PR #1729, #1736); shards from romeerp/parameter-golf-caseops-v1",
+    "sparse_attn_gate": "@nprime06 (PR #1787)",
+    "phased_ttt": "@dexhunter (PR #1736), @abaybektursun (PR #549)",
+    "gptq_sdclip": "@clarkkev (PR #1394)"
+  }
+}
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model
new file mode 100644
index 0000000000..fffc8bb306
Binary files /dev/null and b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model differ
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.vocab b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.vocab
new file mode 100644
index 0000000000..9940e493c6
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.vocab
@@ -0,0 +1,8192 @@
+<pad>	0
+<s>	0
+</s>	0
+<unk>	0
+	0
+	0
+	0
+	0
+<0x00>	0
+<0x01>	0
+<0x02>	0
+<0x03>	0
+<0x04>	0
+<0x05>	0
+<0x06>	0
+<0x07>	0
+<0x08>	0
+<0x09>	0
+<0x0A>	0
+<0x0B>	0
+<0x0C>	0
+<0x0D>	0
+<0x0E>	0
+<0x0F>	0
+<0x10>	0
+<0x11>	0
+<0x12>	0
+<0x13>	0
+<0x14>	0
+<0x15>	0
+<0x16>	0
+<0x17>	0
+<0x18>	0
+<0x19>	0
+<0x1A>	0
+<0x1B>	0
+<0x1C>	0
+<0x1D>	0
+<0x1E>	0
+<0x1F>	0
+<0x20>	0
+<0x21>	0
+<0x22>	0
+<0x23>	0
+<0x24>	0
+<0x25>	0
+<0x26>	0
+<0x27>	0
+<0x28>	0
+<0x29>	0
+<0x2A>	0
+<0x2B>	0
+<0x2C>	0
+<0x2D>	0
+<0x2E>	0
+<0x2F>	0
+<0x30>	0
+<0x31>	0
+<0x32>	0
+<0x33>	0
+<0x34>	0
+<0x35>	0
+<0x36>	0
+<0x37>	0
+<0x38>	0
+<0x39>	0
+<0x3A>	0
+<0x3B>	0
+<0x3C>	0
+<0x3D>	0
+<0x3E>	0
+<0x3F>	0
+<0x40>	0
+<0x41>	0
+<0x42>	0
+<0x43>	0
+<0x44>	0
+<0x45>	0
+<0x46>	0
+<0x47>	0
+<0x48>	0
+<0x49>	0
+<0x4A>	0
+<0x4B>	0
+<0x4C>	0
+<0x4D>	0
+<0x4E>	0
+<0x4F>	0
+<0x50>	0
+<0x51>	0
+<0x52>	0
+<0x53>	0
+<0x54>	0
+<0x55>	0
+<0x56>	0
+<0x57>	0
+<0x58>	0
+<0x59>	0
+<0x5A>	0
+<0x5B>	0
+<0x5C>	0
+<0x5D>	0
+<0x5E>	0
+<0x5F>	0
+<0x60>	0
+<0x61>	0
+<0x62>	0
+<0x63>	0
+<0x64>	0
+<0x65>	0
+<0x66>	0
+<0x67>	0
+<0x68>	0
+<0x69>	0
+<0x6A>	0
+<0x6B>	0
+<0x6C>	0
+<0x6D>	0
+<0x6E>	0
+<0x6F>	0
+<0x70>	0
+<0x71>	0
+<0x72>	0
+<0x73>	0
+<0x74>	0
+<0x75>	0
+<0x76>	0
+<0x77>	0
+<0x78>	0
+<0x79>	0
+<0x7A>	0
+<0x7B>	0
+<0x7C>	0
+<0x7D>	0
+<0x7E>	0
+<0x7F>	0
+<0x80>	0
+<0x81>	0
+<0x82>	0
+<0x83>	0
+<0x84>	0
+<0x85>	0
+<0x86>	0
+<0x87>	0
+<0x88>	0
+<0x89>	0
+<0x8A>	0
+<0x8B>	0
+<0x8C>	0
+<0x8D>	0
+<0x8E>	0
+<0x8F>	0
+<0x90>	0
+<0x91>	0
+<0x92>	0
+<0x93>	0
+<0x94>	0
+<0x95>	0
+<0x96>	0
+<0x97>	0
+<0x98>	0
+<0x99>	0
+<0x9A>	0
+<0x9B>	0
+<0x9C>	0
+<0x9D>	0
+<0x9E>	0
+<0x9F>	0
+<0xA0>	0
+<0xA1>	0
+<0xA2>	0
+<0xA3>	0
+<0xA4>	0
+<0xA5>	0
+<0xA6>	0
+<0xA7>	0
+<0xA8>	0
+<0xA9>	0
+<0xAA>	0
+<0xAB>	0
+<0xAC>	0
+<0xAD>	0
+<0xAE>	0
+<0xAF>	0
+<0xB0>	0
+<0xB1>	0
+<0xB2>	0
+<0xB3>	0
+<0xB4>	0
+<0xB5>	0
+<0xB6>	0
+<0xB7>	0
+<0xB8>	0
+<0xB9>	0
+<0xBA>	0
+<0xBB>	0
+<0xBC>	0
+<0xBD>	0
+<0xBE>	0
+<0xBF>	0
+<0xC0>	0
+<0xC1>	0
+<0xC2>	0
+<0xC3>	0
+<0xC4>	0
+<0xC5>	0
+<0xC6>	0
+<0xC7>	0
+<0xC8>	0
+<0xC9>	0
+<0xCA>	0
+<0xCB>	0
+<0xCC>	0
+<0xCD>	0
+<0xCE>	0
+<0xCF>	0
+<0xD0>	0
+<0xD1>	0
+<0xD2>	0
+<0xD3>	0
+<0xD4>	0
+<0xD5>	0
+<0xD6>	0
+<0xD7>	0
+<0xD8>	0
+<0xD9>	0
+<0xDA>	0
+<0xDB>	0
+<0xDC>	0
+<0xDD>	0
+<0xDE>	0
+<0xDF>	0
+<0xE0>	0
+<0xE1>	0
+<0xE2>	0
+<0xE3>	0
+<0xE4>	0
+<0xE5>	0
+<0xE6>	0
+<0xE7>	0
+<0xE8>	0
+<0xE9>	0
+<0xEA>	0
+<0xEB>	0
+<0xEC>	0
+<0xED>	0
+<0xEE>	0
+<0xEF>	0
+<0xF0>	0
+<0xF1>	0
+<0xF2>	0
+<0xF3>	0
+<0xF4>	0
+<0xF5>	0
+<0xF6>	0
+<0xF7>	0
+<0xF8>	0
+<0xF9>	0
+<0xFA>	0
+<0xFB>	0
+<0xFC>	0
+<0xFD>	0
+<0xFE>	0
+<0xFF>	0
+▁t	-0
+▁a	-1
+in	-2
+he	-3
+re	-4
+on	-5
+er	-6
+▁the	-7
+▁s	-8
+▁w	-9
+or	-10
+at	-11
+ou	-12
+nd	-13
+it	-14
+▁c	-15
+es	-16
+▁f	-17
+is	-18
+en	-19
+ing	-20
+▁b	-21
+▁p	-22
+▁o	-23
+an	-24
+al	-25
+ed	-26
+▁to	-27
+ar	-28
+▁m	-29
+▁and	-30
+▁in	-31
+▁of	-32
+le	-33
+▁d	-34
+as	-35
+ic	-36
+▁h	-37
+om	-38
+ion	-39
+▁th	-40
+il	-41
+st	-42
+▁l	-43
+ro	-44
+ent	-45
+ve	-46
+▁y	-47
+▁e	-48
+▁re	-49
+▁n	-50
+▁g	-51
+ac	-52
+et	-53
+▁you	-54
+ly	-55
+the	-56
+id	-57
+ay	-58
+▁for	-59
+▁is	-60
+▁on	-61
+▁be	-62
+am	-63
+ow	-64
+se	-65
+ot	-66
+ad	-67
+ol	-68
+ig	-69
+ct	-70
+im	-71
+ch	-72
+▁u	-73
+ver	-74
+ith	-75
+ut	-76
+▁st	-77
+el	-78
+ation	-79
+▁with	-80
+ir	-81
+▁that	-82
+th	-83
+ur	-84
+ce	-85
+▁he	-86
+▁it	-87
+ill	-88
+ter	-89
+if	-90
+▁al	-91
+▁an	-92
+ul	-93
+ke	-94
+our	-95
+ag	-96
+ers	-97
+▁pro	-98
+▁wh	-99
+▁as	-100
+▁are	-101
+▁we	-102
+▁ha	-103
+oo	-104
+out	-105
+un	-106
+pp	-107
+us	-108
+ab	-109
+ate	-110
+ess	-111
+▁at	-112
+▁con	-113
+▁com	-114
+▁or	-115
+ra	-116
+em	-117
+ore	-118
+ri	-119
+est	-120
+igh	-121
+rom	-122
+▁-	-123
+▁ne	-124
+op	-125
+▁se	-126
+▁your	-127
+qu	-128
+ld	-129
+and	-130
+ist	-131
+▁(	-132
+res	-133
+ment	-134
+▁ex	-135
+ant	-136
+pe	-137
+ity	-138
+art	-139
+▁v	-140
+ive	-141
+▁r	-142
+all	-143
+▁have	-144
+ort	-145
+ust	-146
+▁was	-147
+um	-148
+▁this	-149
+▁from	-150
+▁de	-151
+oc	-152
+▁sh	-153
+ies	-154
+os	-155
+▁su	-156
+ain	-157
+▁will	-158
+▁can	-159
+▁ch	-160
+ight	-161
+▁by	-162
+nt	-163
+ome	-164
+ard	-165
+▁not	-166
+te	-167
+ud	-168
+▁le	-169
+red	-170
+we	-171
+▁wor	-172
+ost	-173
+ie	-174
+ge	-175
+▁pl	-176
+▁ab	-177
+ther	-178
+iv	-179
+king	-180
+ide	-181
+ast	-182
+for	-183
+per	-184
+gh	-185
+..	-186
+rou	-187
+fe	-188
+ial	-189
+▁all	-190
+pl	-191
+ack	-192
+ine	-193
+▁j	-194
+ould	-195
+od	-196
+ice	-197
+ell	-198
+▁has	-199
+ind	-200
+act	-201
+ne	-202
+ure	-203
+cl	-204
+one	-205
+ear	-206
+▁do	-207
+so	-208
+▁k	-209
+are	-210
+▁us	-211
+▁ad	-212
+ake	-213
+age	-214
+ks	-215
+▁me	-216
+ip	-217
+▁out	-218
+▁but	-219
+ry	-220
+ap	-221
+ally	-222
+▁up	-223
+▁whe	-224
+ions	-225
+com	-226
+to	-227
+able	-228
+▁our	-229
+ail	-230
+▁en	-231
+▁more	-232
+▁comp	-233
+very	-234
+ite	-235
+og	-236
+▁my	-237
+▁“	-238
+ime	-239
+▁so	-240
+ich	-241
+her	-242
+▁cl	-243
+▁their	-244
+ood	-245
+ong	-246
+ated	-247
+ber	-248
+▁sa	-249
+pt	-250
+ame	-251
+iz	-252
+▁they	-253
+▁one	-254
+▁ac	-255
+du	-256
+ike	-257
+▁te	-258
+ous	-259
+▁about	-260
+ak	-261
+ans	-262
+ase	-263
+ace	-264
+ass	-265
+▁cont	-266
+ia	-267
+ru	-268
+con	-269
+▁im	-270
+ire	-271
+ign	-272
+▁fe	-273
+▁who	-274
+ance	-275
+ree	-276
+▁off	-277
+ach	-278
+▁man	-279
+bl	-280
+reat	-281
+now	-282
+▁go	-283
+▁new	-284
+au	-285
+▁"	-286
+ays	-287
+▁his	-288
+av	-289
+ick	-290
+ib	-291
+▁year	-292
+▁app	-293
+ff	-294
+▁res	-295
+form	-296
+▁qu	-297
+erv	-298
+ove	-299
+ary	-300
+port	-301
+ction	-302
+ult	-303
+this	-304
+ations	-305
+▁per	-306
+ook	-307
+ile	-308
+▁also	-309
+▁get	-310
+day	-311
+▁time	-312
+▁which	-313
+ents	-314
+ep	-315
+▁like	-316
+ount	-317
+vel	-318
+▁some	-319
+ue	-320
+▁any	-321
+ven	-322
+▁other	-323
+▁tr	-324
+▁sp	-325
+cess	-326
+ph	-327
+▁been	-328
+che	-329
+▁part	-330
+ings	-331
+ical	-332
+▁over	-333
+sh	-334
+ark	-335
+pro	-336
+▁un	-337
+▁dis	-338
+ress	-339
+ence	-340
+be	-341
+wh	-342
+▁them	-343
+low	-344
+ition	-345
+ors	-346
+int	-347
+▁pe	-348
+ild	-349
+▁her	-350
+▁ar	-351
+▁ag	-352
+▁when	-353
+ord	-354
+▁ro	-355
+ov	-356
+ound	-357
+ough	-358
+ple	-359
+▁just	-360
+su	-361
+▁said	-362
+ish	-363
+own	-364
+irst	-365
+ang	-366
+ren	-367
+▁need	-368
+ck	-369
+ose	-370
+ob	-371
+▁spe	-372
+ll	-373
+▁pre	-374
+▁what	-375
+▁there	-376
+▁pr	-377
+gr	-378
+▁if	-379
+old	-380
+ens	-381
+wn	-382
+you	-383
+▁work	-384
+de	-385
+▁than	-386
+lp	-387
+vers	-388
+▁would	-389
+▁know	-390
+ning	-391
+oy	-392
+ater	-393
+fter	-394
+▁had	-395
+▁am	-396
+ade	-397
+▁des	-398
+▁were	-399
+▁sc	-400
+rough	-401
+▁rec	-402
+clud	-403
+itt	-404
+▁how	-405
+▁comm	-406
+ory	-407
+▁into	-408
+ople	-409
+...	-410
+man	-411
+▁–	-412
+▁help	-413
+nder	-414
+ful	-415
+▁first	-416
+▁em	-417
+▁its	-418
+je	-419
+ool	-420
+▁add	-421
+ft	-422
+▁make	-423
+▁tw	-424
+but	-425
+▁prov	-426
+▁bl	-427
+ink	-428
+anc	-429
+▁people	-430
+end	-431
+▁includ	-432
+▁col	-433
+▁she	-434
+ict	-435
+rent	-436
+get	-437
+iew	-438
+▁produ	-439
+ian	-440
+▁serv	-441
+ex	-442
+round	-443
+ious	-444
+ates	-445
+tern	-446
+▁every	-447
+▁only	-448
+▁bec	-449
+aw	-450
+▁bu	-451
+▁may	-452
+▁no	-453
+▁back	-454
+▁through	-455
+ating	-456
+up	-457
+ual	-458
+ices	-459
+sp	-460
+new	-461
+▁ph	-462
+▁want	-463
+▁most	-464
+arch	-465
+▁see	-466
+ments	-467
+▁well	-468
+bs	-469
+pport	-470
+ities	-471
+wor	-472
+▁use	-473
+▁two	-474
+iss	-475
+co	-476
+▁bet	-477
+les	-478
+cts	-479
+▁after	-480
+ise	-481
+ause	-482
+ife	-483
+hing	-484
+ons	-485
+fore	-486
+iness	-487
+▁play	-488
+ty	-489
+ily	-490
+row	-491
+▁|	-492
+▁act	-493
+▁&	-494
+▁co	-495
+air	-496
+lect	-497
+ting	-498
+▁fl	-499
+enc	-500
+whe	-501
+br	-502
+▁exper	-503
+xt	-504
+▁stud	-505
+cial	-506
+urn	-507
+cent	-508
+sel	-509
+▁these	-510
+ond	-511
+tr	-512
+ts	-513
+▁years	-514
+als	-515
+vent	-516
+▁cons	-517
+▁under	-518
+▁$	-519
+aking	-520
+alth	-521
+▁bus	-522
+▁ke	-523
+▁very	-524
+fr	-525
+ubl	-526
+▁loc	-527
+▁good	-528
+▁sm	-529
+oth	-530
+▁now	-531
+cre	-532
+▁many	-533
+ational	-534
+ased	-535
+▁wee	-536
+▁reg	-537
+ific	-538
+▁gu	-539
+its	-540
+▁tra	-541
+▁way	-542
+ures	-543
+formation	-544
+▁dif	-545
+erson	-546
+ange	-547
+any	-548
+▁great	-549
+fect	-550
+ms	-551
+ience	-552
+oint	-553
+▁start	-554
+ert	-555
+▁fin	-556
+▁day	-557
+other	-558
+▁rel	-559
+▁sy	-560
+▁could	-561
+▁rem	-562
+uring	-563
+read	-564
+ath	-565
+▁ind	-566
+▁don	-567
+ility	-568
+ible	-569
+ident	-570
+▁even	-571
+▁mu	-572
+▁here	-573
+▁own	-574
+ics	-575
+pr	-576
+).	-577
+▁where	-578
+hed	-579
+▁inv	-580
+▁acc	-581
+▁home	-582
+velop	-583
+oss	-584
+line	-585
+▁show	-586
+amp	-587
+comm	-588
+▁find	-589
+▁should	-590
+▁ent	-591
+ble	-592
+▁high	-593
+▁then	-594
+chool	-595
+ular	-596
+▁down	-597
+▁br	-598
+▁right	-599
+ix	-600
+▁read	-601
+▁best	-602
+ited	-603
+▁car	-604
+▁av	-605
+olog	-606
+||	-607
+me	-608
+▁fam	-609
+▁ass	-610
+▁set	-611
+▁too	-612
+fl	-613
+gan	-614
+ied	-615
+▁min	-616
+arn	-617
+▁mon	-618
+view	-619
+▁take	-620
+ock	-621
+ward	-622
+▁design	-623
+▁che	-624
+io	-625
+pect	-626
+ull	-627
+sc	-628
+gram	-629
+stem	-630
+▁business	-631
+▁much	-632
+▁information	-633
+way	-634
+eng	-635
+az	-636
+ax	-637
+▁last	-638
+ins	-639
+▁inst	-640
+ys	-641
+▁long	-642
+ener	-643
+mon	-644
+rest	-645
+omet	-646
+ative	-647
+ason	-648
+▁made	-649
+▁because	-650
+app	-651
+▁did	-652
+.”	-653
+min	-654
+rit	-655
+ject	-656
+med	-657
+ps	-658
+land	-659
+▁call	-660
+atch	-661
+let	-662
+▁before	-663
+ollow	-664
+not	-665
+▁team	-666
+ave	-667
+ash	-668
+how	-669
+oun	-670
+▁sur	-671
+there	-672
+uc	-673
+▁cre	-674
+ner	-675
+▁world	-676
+▁him	-677
+amer	-678
+▁mem	-679
+▁again	-680
+ient	-681
+▁requ	-682
+▁hand	-683
+▁person	-684
+▁art	-685
+▁op	-686
+▁diffe	-687
+my	-688
+▁develop	-689
+mer	-690
+▁gr	-691
+riend	-692
+▁does	-693
+▁life	-694
+▁vis	-695
+),	-696
+arket	-697
+ouse	-698
+▁being	-699
+▁inte	-700
+aut	-701
+▁end	-702
+▁dec	-703
+ros	-704
+ustom	-705
+bo	-706
+ale	-707
+▁such	-708
+ality	-709
+ower	-710
+unity	-711
+▁pres	-712
+▁look	-713
+ving	-714
+▁those	-715
+ince	-716
+chn	-717
+▁med	-718
+no	-719
+▁sk	-720
+ger	-721
+▁while	-722
+▁free	-723
+▁love	-724
+▁provid	-725
+der	-726
+ailable	-727
+irect	-728
+▁support	-729
+ages	-730
+ters	-731
+akes	-732
+ton	-733
+ank	-734
+▁around	-735
+▁sim	-736
+▁used	-737
+▁each	-738
+ced	-739
+▁pol	-740
+ier	-741
+ces	-742
+▁post	-743
+ement	-744
+they	-745
+ives	-746
+▁att	-747
+what	-748
+▁cur	-749
+ah	-750
+▁bo	-751
+col	-752
+▁lead	-753
+▁mod	-754
+ather	-755
+ital	-756
+ional	-757
+by	-758
+ually	-759
+▁char	-760
+ize	-761
+self	-762
+▁ser	-763
+▁care	-764
+she	-765
+rand	-766
+ivers	-767
+▁eas	-768
+▁ed	-769
+▁child	-770
+und	-771
+ublic	-772
+▁loo	-773
+▁det	-774
+▁follow	-775
+▁really	-776
+▁ret	-777
+▁program	-778
+▁system	-779
+imes	-780
+▁run	-781
+▁num	-782
+▁cour	-783
+ets	-784
+do	-785
+cur	-786
+ub	-787
+▁trans	-788
+ines	-789
+▁found	-790
+▁list	-791
+▁inter	-792
+▁feel	-793
+▁think	-794
+▁available	-795
+led	-796
+ittle	-797
+▁going	-798
+ness	-799
+▁count	-800
+iet	-801
+ways	-802
+▁somet	-803
+fin	-804
+▁del	-805
+ann	-806
+iel	-807
+▁fun	-808
+▁month	-809
+ene	-810
+hip	-811
+▁ra	-812
+▁str	-813
+▁top	-814
+▁custom	-815
+▁still	-816
+▁ins	-817
+ames	-818
+▁prof	-819
+cond	-820
+,”	-821
+ten	-822
+▁place	-823
+mar	-824
+▁cle	-825
+with	-826
+▁though	-827
+▁few	-828
+gg	-829
+▁sign	-830
+▁fr	-831
+jo	-832
+▁both	-833
+▁three	-834
+arm	-835
+▁sub	-836
+cept	-837
+que	-838
+▁i	-839
+▁gl	-840
+▁different	-841
+▁happ	-842
+bers	-843
+orn	-844
+▁health	-845
+lease	-846
+ved	-847
+lic	-848
+▁week	-849
+olut	-850
+ired	-851
+▁open	-852
+▁adv	-853
+▁pur	-854
+▁keep	-855
+ross	-856
+▁exp	-857
+▁lot	-858
+▁friend	-859
+when	-860
+part	-861
+ants	-862
+▁market	-863
+▁say	-864
+ata	-865
+ween	-866
+▁family	-867
+▁school	-868
+of	-869
+▁little	-870
+pres	-871
+alk	-872
+▁game	-873
+ars	-874
+▁same	-875
+▁company	-876
+less	-877
+."	-878
+▁using	-879
+ering	-880
+atur	-881
+▁process	-882
+iving	-883
+▁report	-884
+▁techn	-885
+▁feat	-886
+▁lar	-887
+lo	-888
+americ	-889
+▁fil	-890
+▁import	-891
+thing	-892
+▁ev	-893
+▁point	-894
+go	-895
+▁including	-896
+▁comple	-897
+▁experience	-898
+ology	-899
+▁things	-900
+▁met	-901
+ves	-902
+▁during	-903
+▁water	-904
+▁next	-905
+▁between	-906
+stand	-907
+search	-908
+▁aut	-909
+ues	-910
+▁real	-911
+ished	-912
+ized	-913
+ever	-914
+ody	-915
+ining	-916
+▁disc	-917
+▁state	-918
+▁service	-919
+tra	-920
+att	-921
+▁come	-922
+▁mus	-923
+ware	-924
+roup	-925
+▁win	-926
+▁allow	-927
+pre	-928
+▁small	-929
+▁tri	-930
+▁dri	-931
+gy	-932
+▁contin	-933
+▁ear	-934
+cer	-935
+ists	-936
+▁form	-937
+▁full	-938
+▁sure	-939
+▁always	-940
+▁rest	-941
+ric	-942
+nor	-943
+work	-944
+▁cr	-945
+ield	-946
+▁big	-947
+▁var	-948
+▁organ	-949
+▁rece	-950
+▁put	-951
+▁opt	-952
+joy	-953
+▁online	-954
+▁build	-955
+oot	-956
+▁try	-957
+▁cent	-958
+par	-959
+els	-960
+▁appro	-961
+car	-962
+▁interest	-963
+▁plan	-964
+▁phot	-965
+ik	-966
+▁better	-967
+▁days	-968
+▁bel	-969
+▁pri	-970
+▁pat	-971
+uss	-972
+▁since	-973
+that	-974
+sw	-975
+ards	-976
+▁got	-977
+▁poss	-978
+▁class	-979
+ffect	-980
+cri	-981
+▁something	-982
+▁cap	-983
+ours	-984
+▁main	-985
+▁quest	-986
+▁partic	-987
+istr	-988
+pa	-989
+▁special	-990
+▁ide	-991
+reen	-992
+▁pass	-993
+fer	-994
+▁ve	-995
+▁today	-996
+▁result	-997
+▁sl	-998
+gu	-999
+▁pay	-1000
+off	-1001
+▁mov	-1002
+chr	-1003
+▁ty	-1004
+▁hard	-1005
+ipp	-1006
+▁gener	-1007
+▁direct	-1008
+▁perform	-1009
+▁stand	-1010
+▁bro	-1011
+illion	-1012
+▁head	-1013
+iversity	-1014
+▁provide	-1015
+can	-1016
+ott	-1017
+▁resp	-1018
+▁let	-1019
+▁second	-1020
+ok	-1021
+oney	-1022
+ploy	-1023
+ont	-1024
+ients	-1025
+▁give	-1026
+▁val	-1027
+▁wom	-1028
+▁another	-1029
+▁manag	-1030
+▁cor	-1031
+▁current	-1032
+vis	-1033
+▁meet	-1034
+,"	-1035
+▁without	-1036
+here	-1037
+ope	-1038
+▁able	-1039
+▁book	-1040
+▁services	-1041
+sy	-1042
+rence	-1043
+▁wr	-1044
+▁ann	-1045
+may	-1046
+ley	-1047
+▁vide	-1048
+▁number	-1049
+▁int	-1050
+raph	-1051
+▁area	-1052
+▁bre	-1053
+▁soc	-1054
+▁old	-1055
+ww	-1056
+vir	-1057
+ium	-1058
+▁cost	-1059
+▁must	-1060
+ically	-1061
+ems	-1062
+ately	-1063
+▁—	-1064
+ior	-1065
+por	-1066
+▁access	-1067
+▁ter	-1068
+▁invest	-1069
+▁important	-1070
+son	-1071
+▁unt	-1072
+dr	-1073
+bu	-1074
+▁working	-1075
+ense	-1076
+▁students	-1077
+▁webs	-1078
+▁public	-1079
+most	-1080
+gin	-1081
+ling	-1082
+▁never	-1083
+▁sal	-1084
+▁light	-1085
+ertain	-1086
+ral	-1087
+▁fact	-1088
+aj	-1089
+▁suc	-1090
+just	-1091
+▁order	-1092
+▁proble	-1093
+▁grow	-1094
+cont	-1095
+hes	-1096
+▁visit	-1097
+▁incre	-1098
+▁proper	-1099
+▁might	-1100
+vern	-1101
+▁local	-1102
+▁imp	-1103
+more	-1104
+▁site	-1105
+ruct	-1106
+outh	-1107
+ample	-1108
+add	-1109
+▁sing	-1110
+▁enjoy	-1111
+ember	-1112
+▁four	-1113
+room	-1114
+▁event	-1115
+gether	-1116
+▁level	-1117
+▁exc	-1118
+count	-1119
+fort	-1120
+▁fac	-1121
+book	-1122
+▁looking	-1123
+comp	-1124
+ds	-1125
+ute	-1126
+▁food	-1127
+ission	-1128
+play	-1129
+ording	-1130
+pend	-1131
+▁test	-1132
+rol	-1133
+tend	-1134
+▁night	-1135
+▁season	-1136
+hor	-1137
+▁expl	-1138
+▁data	-1139
+▁until	-1140
+▁spec	-1141
+val	-1142
+ential	-1143
+▁room	-1144
+ories	-1145
+▁offer	-1146
+▁dr	-1147
+▁children	-1148
+▁making	-1149
+▁el	-1150
+ency	-1151
+▁less	-1152
+▁within	-1153
+▁eng	-1154
+▁email	-1155
+▁products	-1156
+come	-1157
+cle	-1158
+ently	-1159
+▁view	-1160
+▁'	-1161
+lection	-1162
+▁profess	-1163
+▁pop	-1164
+serv	-1165
+▁product	-1166
+▁sit	-1167
+des	-1168
+ids	-1169
+use	-1170
+ieve	-1171
+▁job	-1172
+ats	-1173
+▁oper	-1174
+▁prom	-1175
+▁power	-1176
+▁pos	-1177
+ases	-1178
+year	-1179
+ains	-1180
+ched	-1181
+ature	-1182
+▁addition	-1183
+▁contact	-1184
+▁supp	-1185
+▁intern	-1186
+▁est	-1187
+ste	-1188
+tt	-1189
+your	-1190
+▁community	-1191
+▁law	-1192
+ney	-1193
+tw	-1194
+ared	-1195
+ribut	-1196
+af	-1197
+ique	-1198
+ills	-1199
+▁quality	-1200
+aff	-1201
+▁turn	-1202
+▁against	-1203
+▁prote	-1204
+▁impro	-1205
+▁past	-1206
+ches	-1207
+icle	-1208
+inc	-1209
+ivid	-1210
+university	-1211
+pos	-1212
+reg	-1213
+▁together	-1214
+▁ob	-1215
+pm	-1216
+utes	-1217
+▁away	-1218
+gl	-1219
+▁mar	-1220
+▁project	-1221
+ator	-1222
+ified	-1223
+after	-1224
+ined	-1225
+▁conf	-1226
+rap	-1227
+ode	-1228
+▁website	-1229
+sm	-1230
+iday	-1231
+▁indust	-1232
+▁ask	-1233
+▁easy	-1234
+ability	-1235
+▁complete	-1236
+ider	-1237
+▁needs	-1238
+irl	-1239
+▁eff	-1240
+ina	-1241
+please	-1242
+face	-1243
+▁creat	-1244
+▁learn	-1245
+back	-1246
+▁pract	-1247
+these	-1248
+▁success	-1249
+▁tru	-1250
+ips	-1251
+time	-1252
+▁bas	-1253
+ju	-1254
+▁beaut	-1255
+▁please	-1256
+▁detail	-1257
+free	-1258
+▁equ	-1259
+▁pain	-1260
+▁leg	-1261
+ready	-1262
+augh	-1263
+▁members	-1264
+▁change	-1265
+cal	-1266
+atter	-1267
+!!	-1268
+▁course	-1269
+▁chang	-1270
+ng	-1271
+leg	-1272
+▁include	-1273
+▁large	-1274
+christ	-1275
+arent	-1276
+iety	-1277
+▁money	-1278
+yle	-1279
+▁along	-1280
+ries	-1281
+▁says	-1282
+▁group	-1283
+ift	-1284
+oci	-1285
+▁check	-1286
+ream	-1287
+ised	-1288
+▁fav	-1289
+▁dist	-1290
+bro	-1291
+aster	-1292
+opp	-1293
+▁applic	-1294
+▁name	-1295
+▁house	-1296
+▁near	-1297
+over	-1298
+▁offic	-1299
+aterial	-1300
+ird	-1301
+bar	-1302
+alf	-1303
+load	-1304
+eu	-1305
+▁expect	-1306
+▁music	-1307
+cut	-1308
+ism	-1309
+ring	-1310
+▁video	-1311
+▁activ	-1312
+uk	-1313
+une	-1314
+▁times	-1315
+▁present	-1316
+ane	-1317
+iting	-1318
+cour	-1319
+▁buy	-1320
+igin	-1321
+▁why	-1322
+ump	-1323
+side	-1324
+▁employ	-1325
+iver	-1326
+net	-1327
+ising	-1328
+▁perfect	-1329
+ably	-1330
+ocus	-1331
+▁opport	-1332
+ered	-1333
+▁thing	-1334
+ote	-1335
+spe	-1336
+hn	-1337
+ery	-1338
+▁below	-1339
+▁.	-1340
+veral	-1341
+urs	-1342
+cor	-1343
+ny	-1344
+oor	-1345
+state	-1346
+▁air	-1347
+▁ref	-1348
+▁pa	-1349
+ify	-1350
+ball	-1351
+ury	-1352
+enef	-1353
+itions	-1354
+▁case	-1355
+▁research	-1356
+▁account	-1357
+ots	-1358
+vernment	-1359
+▁live	-1360
+ored	-1361
+ived	-1362
+▁line	-1363
+▁beh	-1364
+▁hum	-1365
+aring	-1366
+▁create	-1367
+la	-1368
+▁following	-1369
+▁writ	-1370
+▁short	-1371
+▁hist	-1372
+▁hot	-1373
+▁makes	-1374
+▁sw	-1375
+▁kind	-1376
+▁review	-1377
+▁soft	-1378
+▁ce	-1379
+▁left	-1380
+▁page	-1381
+▁treat	-1382
+itive	-1383
+ision	-1384
+▁/	-1385
+viron	-1386
+dis	-1387
+▁having	-1388
+▁side	-1389
+▁dem	-1390
+▁conne	-1391
+uck	-1392
+fact	-1393
+▁getting	-1394
+vious	-1395
+▁story	-1396
+▁works	-1397
+▁share	-1398
+ization	-1399
+while	-1400
+▁across	-1401
+▁minutes	-1402
+▁enough	-1403
+▁move	-1404
+▁ever	-1405
+don	-1406
+iam	-1407
+osp	-1408
+▁personal	-1409
+ffe	-1410
+ma	-1411
+▁far	-1412
+ca	-1413
+▁already	-1414
+▁trad	-1415
+▁cond	-1416
+tain	-1417
+▁understand	-1418
+ention	-1419
+▁effect	-1420
+▁const	-1421
+▁record	-1422
+john	-1423
+▁won	-1424
+▁hours	-1425
+▁possible	-1426
+▁done	-1427
+mber	-1428
+▁million	-1429
+▁news	-1430
+▁par	-1431
+aim	-1432
+ush	-1433
+▁comes	-1434
+light	-1435
+▁based	-1436
+▁several	-1437
+▁nor	-1438
+▁cut	-1439
+▁educ	-1440
+▁women	-1441
+ze	-1442
+ienc	-1443
+pol	-1444
+unch	-1445
+▁camp	-1446
+▁dev	-1447
+▁social	-1448
+house	-1449
+ividual	-1450
+aces	-1451
+▁become	-1452
+ured	-1453
+▁mil	-1454
+ister	-1455
+oice	-1456
+right	-1457
+ron	-1458
+uture	-1459
+american	-1460
+ajor	-1461
+▁bit	-1462
+▁country	-1463
+iven	-1464
+▁boo	-1465
+city	-1466
+▁benef	-1467
+▁focus	-1468
+▁es	-1469
+▁often	-1470
+ford	-1471
+▁cou	-1472
+▁men	-1473
+raw	-1474
+▁tell	-1475
+▁:	-1476
+nes	-1477
+▁development	-1478
+mal	-1479
+▁net	-1480
+ances	-1481
+intern	-1482
+▁pot	-1483
+▁jo	-1484
+▁quick	-1485
+▁...	-1486
+iful	-1487
+▁vers	-1488
+▁hold	-1489
+▁material	-1490
+rec	-1491
+▁city	-1492
+▁body	-1493
+ork	-1494
+▁friends	-1495
+jan	-1496
+▁ext	-1497
+school	-1498
+head	-1499
+ides	-1500
+▁didn	-1501
+pri	-1502
+▁bring	-1503
+▁fund	-1504
+▁purch	-1505
+▁sum	-1506
+ograph	-1507
+vironment	-1508
+ows	-1509
+▁blog	-1510
+raft	-1511
+inst	-1512
+▁issu	-1513
+yes	-1514
+▁elect	-1515
+▁mult	-1516
+ference	-1517
+ai	-1518
+nov	-1519
+ource	-1520
+osed	-1521
+post	-1522
+▁return	-1523
+ns	-1524
+▁doing	-1525
+met	-1526
+olution	-1527
+▁pie	-1528
+from	-1529
+ham	-1530
+idd	-1531
+▁tre	-1532
+well	-1533
+mc	-1534
+▁others	-1535
+▁months	-1536
+ners	-1537
+▁ste	-1538
+▁mom	-1539
+▁wa	-1540
+ones	-1541
+▁individual	-1542
+▁pack	-1543
+har	-1544
+▁non	-1545
+▁travel	-1546
+▁doesn	-1547
+▁called	-1548
+ser	-1549
+where	-1550
+lege	-1551
+▁started	-1552
+▁space	-1553
+po	-1554
+gar	-1555
+▁deal	-1556
+▁compet	-1557
+aug	-1558
+▁took	-1559
+however	-1560
+▁store	-1561
+▁mind	-1562
+▁sol	-1563
+some	-1564
+▁came	-1565
+.,	-1566
+ination	-1567
+conom	-1568
+▁clean	-1569
+reet	-1570
+str	-1571
+▁future	-1572
+▁la	-1573
+▁yet	-1574
+▁lo	-1575
+ateg	-1576
+▁rele	-1577
+wood	-1578
+▁industry	-1579
+▁price	-1580
+▁cover	-1581
+▁pict	-1582
+▁games	-1583
+acy	-1584
+▁means	-1585
+curity	-1586
+▁low	-1587
+will	-1588
+roll	-1589
+ged	-1590
+arning	-1591
+unic	-1592
+bed	-1593
+world	-1594
+▁oc	-1595
+sk	-1596
+lish	-1597
+west	-1598
+na	-1599
+▁major	-1600
+▁percent	-1601
+▁key	-1602
+aust	-1603
+▁control	-1604
+top	-1605
+▁*	-1606
+ply	-1607
+▁lim	-1608
+▁heart	-1609
+▁least	-1610
+▁went	-1611
+▁tem	-1612
+jects	-1613
+duct	-1614
+▁offers	-1615
+▁five	-1616
+▁range	-1617
+cr	-1618
+▁chall	-1619
+▁once	-1620
+ye	-1621
+see	-1622
+ption	-1623
+anies	-1624
+▁tool	-1625
+god	-1626
+ilities	-1627
+▁features	-1628
+▁fore	-1629
+high	-1630
+ster	-1631
+bre	-1632
+▁early	-1633
+▁x	-1634
+▁origin	-1635
+▁series	-1636
+▁specific	-1637
+▁land	-1638
+▁wind	-1639
+united	-1640
+▁--	-1641
+▁tal	-1642
+▁comb	-1643
+▁hope	-1644
+▁require	-1645
+anks	-1646
+commend	-1647
+hol	-1648
+▁step	-1649
+illed	-1650
+ither	-1651
+dep	-1652
+▁discuss	-1653
+▁upd	-1654
+▁invol	-1655
+▁mean	-1656
+org	-1657
+ilt	-1658
+ask	-1659
+pet	-1660
+▁girl	-1661
+▁single	-1662
+nam	-1663
+▁ago	-1664
+mat	-1665
+▁thought	-1666
+aken	-1667
+▁priv	-1668
+▁close	-1669
+▁told	-1670
+resh	-1671
+itor	-1672
+▁strong	-1673
+▁beautiful	-1674
+▁appear	-1675
+stud	-1676
+char	-1677
+york	-1678
+▁happen	-1679
+//	-1680
+▁questions	-1681
+last	-1682
+▁favor	-1683
+oose	-1684
+cy	-1685
+▁stay	-1686
+down	-1687
+board	-1688
+▁government	-1689
+north	-1690
+mit	-1691
+▁customers	-1692
+like	-1693
+tract	-1694
+urther	-1695
+▁added	-1696
+▁brand	-1697
+euro	-1698
+gs	-1699
+://	-1700
+▁study	-1701
+aps	-1702
+rel	-1703
+present	-1704
+wed	-1705
+▁charac	-1706
+▁begin	-1707
+oct	-1708
+rain	-1709
+ians	-1710
+▁redu	-1711
+ted	-1712
+▁hig	-1713
+ibr	-1714
+opy	-1715
+press	-1716
+▁mess	-1717
+▁port	-1718
+▁everything	-1719
+fully	-1720
+▁actually	-1721
+irm	-1722
+ources	-1723
+▁tax	-1724
+▁install	-1725
+atural	-1726
+▁clear	-1727
+▁ener	-1728
+lab	-1729
+bur	-1730
+wards	-1731
+hers	-1732
+ern	-1733
+icy	-1734
+▁search	-1735
+ret	-1736
+ama	-1737
+▁taking	-1738
+south	-1739
+put	-1740
+▁field	-1741
+ensive	-1742
+hy	-1743
+rs	-1744
+▁technology	-1745
+▁announ	-1746
+▁address	-1747
+▁environment	-1748
+▁def	-1749
+atest	-1750
+▁certain	-1751
+▁unique	-1752
+acc	-1753
+▁watch	-1754
+▁improve	-1755
+ancial	-1756
+lebr	-1757
+cast	-1758
+apr	-1759
+▁face	-1760
+isk	-1761
+▁staff	-1762
+gle	-1763
+▁media	-1764
+ship	-1765
+▁host	-1766
+▁super	-1767
+▁results	-1768
+▁content	-1769
+▁bar	-1770
+ural	-1771
+▁training	-1772
+▁half	-1773
+center	-1774
+▁anal	-1775
+▁typ	-1776
+▁mor	-1777
+ports	-1778
+rop	-1779
+▁young	-1780
+▁someone	-1781
+▁companies	-1782
+ocu	-1783
+▁tot	-1784
+friday	-1785
+county	-1786
+▁later	-1787
+unities	-1788
+▁sound	-1789
+▁recommend	-1790
+national	-1791
+ctor	-1792
+ief	-1793
+▁vol	-1794
+▁building	-1795
+ram	-1796
+▁everyone	-1797
+sure	-1798
+tty	-1799
+eral	-1800
+uff	-1801
+▁contr	-1802
+▁miss	-1803
+enn	-1804
+ided	-1805
+▁press	-1806
+▁known	-1807
+bus	-1808
+▁card	-1809
+▁oil	-1810
+▁professional	-1811
+▁particip	-1812
+▁office	-1813
+▁whole	-1814
+▁problem	-1815
+public	-1816
+cially	-1817
+urch	-1818
+▁z	-1819
+▁break	-1820
+ests	-1821
+▁mot	-1822
+phone	-1823
+▁film	-1824
+▁provides	-1825
+▁wond	-1826
+sue	-1827
+▁official	-1828
+ica	-1829
+▁ident	-1830
+▁bed	-1831
+coming	-1832
+▁six	-1833
+satur	-1834
+▁particular	-1835
+▁ess	-1836
+itted	-1837
+care	-1838
+ables	-1839
+▁respons	-1840
+▁simple	-1841
+name	-1842
+▁seen	-1843
+ging	-1844
+▁color	-1845
+▁log	-1846
+▁size	-1847
+ape	-1848
+pat	-1849
+▁front	-1850
+health	-1851
+ibility	-1852
+hod	-1853
+prof	-1854
+ison	-1855
+▁age	-1856
+▁issues	-1857
+place	-1858
+▁mag	-1859
+▁web	-1860
+▁areas	-1861
+elf	-1862
+▁sell	-1863
+europe	-1864
+park	-1865
+▁publ	-1866
+▁tour	-1867
+▁fall	-1868
+▁dig	-1869
+oved	-1870
+imate	-1871
+▁cred	-1872
+march	-1873
+▁given	-1874
+set	-1875
+pecially	-1876
+▁performance	-1877
+lin	-1878
+sund	-1879
+▁reason	-1880
+▁anything	-1881
+▁events	-1882
+▁htt	-1883
+ee	-1884
+▁energy	-1885
+atic	-1886
+▁due	-1887
+▁designed	-1888
+bor	-1889
+▁management	-1890
+ilar	-1891
+tc	-1892
+▁black	-1893
+▁author	-1894
+▁believe	-1895
+▁above	-1896
+▁prob	-1897
+june	-1898
+▁cal	-1899
+▁details	-1900
+ety	-1901
+▁relations	-1902
+▁assist	-1903
+mo	-1904
+used	-1905
+▁latest	-1906
+▁phone	-1907
+▁foot	-1908
+then	-1909
+ledge	-1910
+▁print	-1911
+iod	-1912
+asing	-1913
+saturday	-1914
+obal	-1915
+▁aud	-1916
+was	-1917
+gest	-1918
+▁effort	-1919
+mor	-1920
+▁econom	-1921
+obile	-1922
+direct	-1923
+▁pretty	-1924
+▁skin	-1925
+aur	-1926
+▁sex	-1927
+▁history	-1928
+ug	-1929
+ino	-1930
+ille	-1931
+sept	-1932
+▁claim	-1933
+▁weeks	-1934
+ites	-1935
+ouch	-1936
+inter	-1937
+▁property	-1938
+ulation	-1939
+▁connect	-1940
+based	-1941
+▁according	-1942
+july	-1943
+▁ca	-1944
+▁link	-1945
+▁select	-1946
+▁experienc	-1947
+ators	-1948
+key	-1949
+ii	-1950
+tv	-1951
+▁soon	-1952
+▁ant	-1953
+▁idea	-1954
+▁points	-1955
+▁includes	-1956
+inking	-1957
+▁living	-1958
+itter	-1959
+▁pick	-1960
+every	-1961
+▁popular	-1962
+itch	-1963
+weet	-1964
+▁comput	-1965
+ges	-1966
+ogn	-1967
+▁protect	-1968
+▁phys	-1969
+aily	-1970
+▁defin	-1971
+▁among	-1972
+▁continue	-1973
+▁inc	-1974
+▁question	-1975
+his	-1976
+▁cult	-1977
+sal	-1978
+▁compl	-1979
+news	-1980
+irt	-1981
+itional	-1982
+▁fre	-1983
+astic	-1984
+dav	-1985
+▁fire	-1986
+ruction	-1987
+mas	-1988
+....	-1989
+manag	-1990
+ufact	-1991
+▁white	-1992
+nding	-1993
+feb	-1994
+vert	-1995
+▁receive	-1996
+ands	-1997
+april	-1998
+▁talk	-1999
+▁opportunity	-2000
+▁parent	-2001
+first	-2002
+▁taken	-2003
+▁dou	-2004
+kend	-2005
+uation	-2006
+▁track	-2007
+▁type	-2008
+▁further	-2009
+▁value	-2010
+▁celebr	-2011
+ental	-2012
+oh	-2013
+li	-2014
+▁date	-2015
+▁amount	-2016
+ming	-2017
+▁kids	-2018
+▁items	-2019
+▁currently	-2020
+▁docu	-2021
+idence	-2022
+▁comfort	-2023
+▁di	-2024
+▁hon	-2025
+▁required	-2026
+ires	-2027
+selves	-2028
+▁descri	-2029
+▁answ	-2030
+▁coming	-2031
+▁fit	-2032
+monday	-2033
+▁various	-2034
+ding	-2035
+ese	-2036
+▁safe	-2037
+▁human	-2038
+▁entire	-2039
+▁text	-2040
+▁trying	-2041
+▁whether	-2042
+▁bad	-2043
+▁sent	-2044
+▁forward	-2045
+▁,	-2046
+aching	-2047
+▁yourself	-2048
+sunday	-2049
+den	-2050
+elt	-2051
+iate	-2052
+▁concer	-2053
+▁summer	-2054
+▁consider	-2055
+mic	-2056
+▁stop	-2057
+many	-2058
+fo	-2059
+▁sugg	-2060
+▁wide	-2061
+wes	-2062
+▁happy	-2063
+ume	-2064
+word	-2065
+▁ready	-2066
+itting	-2067
+▁member	-2068
+▁almost	-2069
+aged	-2070
+ises	-2071
+▁diff	-2072
+icated	-2073
+ler	-2074
+pper	-2075
+uary	-2076
+▁dam	-2077
+▁box	-2078
+ained	-2079
+▁needed	-2080
+▁attend	-2081
+black	-2082
+soc	-2083
+▁example	-2084
+▁style	-2085
+sing	-2086
+▁final	-2087
+▁ach	-2088
+ried	-2089
+▁either	-2090
+▁software	-2091
+ival	-2092
+▁via	-2093
+mus	-2094
+azing	-2095
+▁meas	-2096
+▁famil	-2097
+ux	-2098
+▁red	-2099
+president	-2100
+ole	-2101
+▁crit	-2102
+▁wanted	-2103
+▁hom	-2104
+home	-2105
+▁provided	-2106
+▁especially	-2107
+urance	-2108
+ana	-2109
+ray	-2110
+icult	-2111
+▁third	-2112
+▁recently	-2113
+sim	-2114
+▁ways	-2115
+▁seem	-2116
+▁isn	-2117
+ek	-2118
+▁potential	-2119
+lim	-2120
+▁relationship	-2121
+▁held	-2122
+verage	-2123
+mr	-2124
+▁quite	-2125
+▁running	-2126
+▁ensure	-2127
+ples	-2128
+▁couple	-2129
+life	-2130
+war	-2131
+click	-2132
+▁sle	-2133
+▁road	-2134
+▁insp	-2135
+ively	-2136
+▁previous	-2137
+▁favorite	-2138
+itten	-2139
+▁regular	-2140
+▁nice	-2141
+gress	-2142
+▁hol	-2143
+mb	-2144
+▁paper	-2145
+▁song	-2146
+▁common	-2147
+clus	-2148
+▁issue	-2149
+▁options	-2150
+▁veh	-2151
+ground	-2152
+▁financial	-2153
+▁throughout	-2154
+▁cup	-2155
+▁users	-2156
+▁natural	-2157
+design	-2158
+wind	-2159
+law	-2160
+▁bott	-2161
+▁period	-2162
+cas	-2163
+▁flow	-2164
+aby	-2165
+▁clients	-2166
+▁engine	-2167
+▁init	-2168
+coun	-2169
+know	-2170
+▁choose	-2171
+cons	-2172
+ees	-2173
+▁morning	-2174
+aced	-2175
+▁created	-2176
+▁jour	-2177
+.)	-2178
+▁sales	-2179
+▁version	-2180
+ada	-2181
+ung	-2182
+▁else	-2183
+▁takes	-2184
+▁extra	-2185
+▁players	-2186
+▁inde	-2187
+▁behind	-2188
+▁cool	-2189
+▁download	-2190
+bi	-2191
+aging	-2192
+▁er	-2193
+▁reading	-2194
+ishing	-2195
+▁lab	-2196
+sen	-2197
+san	-2198
+▁recent	-2199
+▁writing	-2200
+band	-2201
+▁click	-2202
+▁consum	-2203
+▁fresh	-2204
+uge	-2205
+▁probably	-2206
+mich	-2207
+▁stri	-2208
+ecut	-2209
+cil	-2210
+▁role	-2211
+▁war	-2212
+▁tit	-2213
+loc	-2214
+▁asked	-2215
+ober	-2216
+sa	-2217
+▁problems	-2218
+▁received	-2219
+ague	-2220
+▁represent	-2221
+also	-2222
+ipping	-2223
+inks	-2224
+class	-2225
+▁walk	-2226
+▁wonder	-2227
+▁leave	-2228
+anced	-2229
+angu	-2230
+lar	-2231
+▁sun	-2232
+icles	-2233
+▁deter	-2234
+road	-2235
+▁security	-2236
+▁ele	-2237
+overed	-2238
+ken	-2239
+▁party	-2240
+▁education	-2241
+▁article	-2242
+ades	-2243
+rew	-2244
+star	-2245
+▁shows	-2246
+istrict	-2247
+list	-2248
+gener	-2249
+▁collect	-2250
+rodu	-2251
+tues	-2252
+cap	-2253
+aving	-2254
+august	-2255
+wel	-2256
+▁simply	-2257
+▁increase	-2258
+▁impact	-2259
+▁books	-2260
+▁hy	-2261
+▁action	-2262
+▁anyone	-2263
+eth	-2264
+mark	-2265
+ev	-2266
+artment	-2267
+hib	-2268
+▁etc	-2269
+states	-2270
+thurs	-2271
+▁release	-2272
+▁original	-2273
+ony	-2274
+▁ful	-2275
+october	-2276
+gree	-2277
+ror	-2278
+istration	-2279
+▁self	-2280
+▁model	-2281
+olutions	-2282
+▁knowledge	-2283
+atform	-2284
+▁customer	-2285
+▁hit	-2286
+▁fast	-2287
+ffic	-2288
+▁coun	-2289
+▁medical	-2290
+chen	-2291
+mod	-2292
+▁mix	-2293
+january	-2294
+▁table	-2295
+▁outside	-2296
+▁exam	-2297
+▁changes	-2298
+▁however	-2299
+▁skills	-2300
+associ	-2301
+▁likely	-2302
+september	-2303
+▁difficult	-2304
+▁method	-2305
+▁accept	-2306
+oid	-2307
+ooking	-2308
+▁inside	-2309
+box	-2310
+▁pan	-2311
+bel	-2312
+--	-2313
+tod	-2314
+oogle	-2315
+▁abs	-2316
+▁goal	-2317
+cing	-2318
+▁ple	-2319
+▁board	-2320
+hen	-2321
+ification	-2322
+wednes	-2323
+idge	-2324
+▁ground	-2325
+▁programs	-2326
+▁police	-2327
+▁career	-2328
+▁town	-2329
+▁match	-2330
+tuesday	-2331
+▁signific	-2332
+facebook	-2333
+bit	-2334
+▁activities	-2335
+adem	-2336
+▁learning	-2337
+▁systems	-2338
+▁weekend	-2339
+▁manufact	-2340
+▁ball	-2341
+▁meeting	-2342
+america	-2343
+▁looks	-2344
+aker	-2345
+check	-2346
+▁private	-2347
+▁contract	-2348
+▁position	-2349
+iple	-2350
+▁continu	-2351
+▁cir	-2352
+▁recogn	-2353
+thursday	-2354
+▁matter	-2355
+▁screen	-2356
+▁words	-2357
+▁necess	-2358
+▁nothing	-2359
+▁application	-2360
+▁lives	-2361
+▁contribut	-2362
+▁ri	-2363
+best	-2364
+ules	-2365
+oly	-2366
+adu	-2367
+nal	-2368
+ville	-2369
+▁fail	-2370
+▁hair	-2371
+cover	-2372
+rem	-2373
+▁anim	-2374
+▁photos	-2375
+irth	-2376
+austral	-2377
+▁mat	-2378
+▁included	-2379
+▁lay	-2380
+wednesday	-2381
+▁projects	-2382
+▁sat	-2383
+water	-2384
+friend	-2385
+hold	-2386
+▁worth	-2387
+ening	-2388
+field	-2389
+▁round	-2390
+▁longer	-2391
+service	-2392
+ops	-2393
+ude	-2394
+▁stor	-2395
+business	-2396
+ellow	-2397
+▁practice	-2398
+▁exist	-2399
+▁additional	-2400
+market	-2401
+iqu	-2402
+ought	-2403
+▁located	-2404
+▁similar	-2405
+pal	-2406
+arter	-2407
+phot	-2408
+date	-2409
+nove	-2410
+▁former	-2411
+iter	-2412
+tim	-2413
+street	-2414
+▁treatment	-2415
+found	-2416
+▁amazing	-2417
+department	-2418
+ights	-2419
+▁total	-2420
+▁fig	-2421
+thanks	-2422
+▁suggest	-2423
+iol	-2424
+▁playing	-2425
+aign	-2426
+miss	-2427
+ls	-2428
+▁risk	-2429
+arget	-2430
+win	-2431
+▁associ	-2432
+▁kit	-2433
+▁communic	-2434
+action	-2435
+▁send	-2436
+▁drive	-2437
+▁repl	-2438
+▁worked	-2439
+dom	-2440
+real	-2441
+bon	-2442
+wal	-2443
+soft	-2444
+▁variety	-2445
+▁seems	-2446
+▁choice	-2447
+del	-2448
+▁rather	-2449
+▁infl	-2450
+ipment	-2451
+izing	-2452
+▁death	-2453
+mag	-2454
+▁saf	-2455
+example	-2456
+unt	-2457
+▁touch	-2458
+▁providing	-2459
+google	-2460
+airs	-2461
+▁purchase	-2462
+▁true	-2463
+▁autom	-2464
+▁growth	-2465
+about	-2466
+▁comment	-2467
+▁plans	-2468
+▁student	-2469
+▁loss	-2470
+▁draw	-2471
+ba	-2472
+▁polit	-2473
+▁general	-2474
+good	-2475
+ga	-2476
+trans	-2477
+▁chance	-2478
+rect	-2479
+▁production	-2480
+▁significant	-2481
+▁international	-2482
+great	-2483
+bal	-2484
+hot	-2485
+▁late	-2486
+ipe	-2487
+lier	-2488
+ercial	-2489
+group	-2490
+mission	-2491
+▁cook	-2492
+empt	-2493
+▁teac	-2494
+▁character	-2495
+▁function	-2496
+bt	-2497
+▁arri	-2498
+calif	-2499
+power	-2500
+▁leading	-2501
+▁court	-2502
+▁file	-2503
+▁allows	-2504
+▁bud	-2505
+international	-2506
+ef	-2507
+iff	-2508
+ivery	-2509
+opping	-2510
+why	-2511
+▁image	-2512
+hood	-2513
+▁network	-2514
+▁cell	-2515
+ession	-2516
+▁wel	-2517
+ha	-2518
+▁sn	-2519
+▁challeng	-2520
+▁leaders	-2521
+▁national	-2522
+mp	-2523
+green	-2524
+▁ideas	-2525
+▁wall	-2526
+▁built	-2527
+▁cand	-2528
+▁remain	-2529
+▁boy	-2530
+iddle	-2531
+▁encour	-2532
+▁collection	-2533
+▁sil	-2534
+eds	-2535
+hel	-2536
+▁movie	-2537
+november	-2538
+▁investig	-2539
+sl	-2540
+labor	-2541
+imum	-2542
+▁mach	-2543
+▁moment	-2544
+gen	-2545
+ondon	-2546
+▁platform	-2547
+ached	-2548
+rich	-2549
+▁decided	-2550
+bert	-2551
+▁quickly	-2552
+▁rad	-2553
+web	-2554
+imin	-2555
+white	-2556
+▁exact	-2557
+▁tick	-2558
+▁piece	-2559
+call	-2560
+▁easily	-2561
+cret	-2562
+▁usually	-2563
+▁bag	-2564
+under	-2565
+▁prevent	-2566
+▁stat	-2567
+▁estab	-2568
+▁excell	-2569
+▁wood	-2570
+▁display	-2571
+▁bal	-2572
+▁prior	-2573
+sch	-2574
+london	-2575
+▁cause	-2576
+cup	-2577
+▁regard	-2578
+▁credit	-2579
+even	-2580
+▁decision	-2581
+sur	-2582
+ederal	-2583
+east	-2584
+▁computer	-2585
+▁location	-2586
+▁clos	-2587
+▁higher	-2588
+mm	-2589
+▁lost	-2590
+dece	-2591
+idents	-2592
+icks	-2593
+▁depend	-2594
+aste	-2595
+itut	-2596
+▁cannot	-2597
+▁ill	-2598
+super	-2599
+▁http	-2600
+▁sche	-2601
+▁apply	-2602
+▁feature	-2603
+▁introdu	-2604
+amed	-2605
+▁daily	-2606
+using	-2607
+▁safety	-2608
+▁commit	-2609
+▁expected	-2610
+bec	-2611
+oud	-2612
+techn	-2613
+jack	-2614
+▁rep	-2615
+zy	-2616
+▁heav	-2617
+step	-2618
+today	-2619
+december	-2620
+▁hor	-2621
+▁mass	-2622
+ospital	-2623
+▁involved	-2624
+pped	-2625
+▁respect	-2626
+sam	-2627
+▁myself	-2628
+▁weight	-2629
+fam	-2630
+▁user	-2631
+david	-2632
+reng	-2633
+▁base	-2634
+ky	-2635
+▁released	-2636
+aves	-2637
+▁photo	-2638
+▁parents	-2639
+▁goes	-2640
+▁serious	-2641
+▁integr	-2642
+mac	-2643
+▁rate	-2644
+▁note	-2645
+oes	-2646
+find	-2647
+▁enter	-2648
+arden	-2649
+ruary	-2650
+▁floor	-2651
+english	-2652
+▁green	-2653
+▁cy	-2654
+▁vict	-2655
+ibrary	-2656
+order	-2657
+roups	-2658
+illing	-2659
+▁mother	-2660
+▁reported	-2661
+iding	-2662
+▁saw	-2663
+attle	-2664
+▁rights	-2665
+▁stra	-2666
+▁woman	-2667
+february	-2668
+mem	-2669
+ester	-2670
+▁deep	-2671
+works	-2672
+erg	-2673
+wide	-2674
+▁smo	-2675
+though	-2676
+child	-2677
+oring	-2678
+▁dress	-2679
+ead	-2680
+▁hands	-2681
+▁word	-2682
+▁expert	-2683
+bum	-2684
+▁dom	-2685
+▁squ	-2686
+big	-2687
+▁benefits	-2688
+▁feed	-2689
+▁gen	-2690
+▁marketing	-2691
+ban	-2692
+iction	-2693
+▁digital	-2694
+night	-2695
+▁center	-2696
+▁af	-2697
+thern	-2698
+program	-2699
+iant	-2700
+ructure	-2701
+▁hear	-2702
+▁campaign	-2703
+www	-2704
+twitter	-2705
+mary	-2706
+▁tools	-2707
+californ	-2708
+ength	-2709
+college	-2710
+once	-2711
+since	-2712
+▁related	-2713
+▁langu	-2714
+▁surpr	-2715
+men	-2716
+▁save	-2717
+body	-2718
+umb	-2719
+bscri	-2720
+▁effic	-2721
+▁costs	-2722
+▁helps	-2723
+▁door	-2724
+▁gives	-2725
+▁ep	-2726
+▁roll	-2727
+▁request	-2728
+▁announced	-2729
+hall	-2730
+point	-2731
+▁source	-2732
+▁partners	-2733
+afr	-2734
+▁immed	-2735
+▁huge	-2736
+atives	-2737
+services	-2738
+▁wed	-2739
+▁ut	-2740
+make	-2741
+▁ability	-2742
+▁complet	-2743
+▁stock	-2744
+▁global	-2745
+▁mobile	-2746
+▁code	-2747
+▁gets	-2748
+change	-2749
+iring	-2750
+lex	-2751
+arant	-2752
+love	-2753
+asons	-2754
+enge	-2755
+izes	-2756
+▁modern	-2757
+dri	-2758
+▁earn	-2759
+▁agre	-2760
+force	-2761
+▁materials	-2762
+erc	-2763
+▁arch	-2764
+iation	-2765
+da	-2766
+vol	-2767
+reme	-2768
+▁giving	-2769
+▁trust	-2770
+akers	-2771
+▁standard	-2772
+pay	-2773
+brit	-2774
+according	-2775
+▁sale	-2776
+cure	-2777
+▁option	-2778
+▁attack	-2779
+▁restaur	-2780
+california	-2781
+tex	-2782
+▁award	-2783
+di	-2784
+▁write	-2785
+▁published	-2786
+▁admin	-2787
+roy	-2788
+plement	-2789
+▁played	-2790
+▁patients	-2791
+▁ones	-2792
+▁dest	-2793
+ocol	-2794
+▁gar	-2795
+christmas	-2796
+▁warm	-2797
+▁parts	-2798
+dem	-2799
+fil	-2800
+ida	-2801
+▁ded	-2802
+▁eye	-2803
+▁maintain	-2804
+cu	-2805
+pose	-2806
+have	-2807
+▁fine	-2808
+ael	-2809
+▁written	-2810
+ivity	-2811
+▁organization	-2812
+▁sus	-2813
+▁interested	-2814
+▁fab	-2815
+vey	-2816
+india	-2817
+▁copy	-2818
+▁sold	-2819
+▁planning	-2820
+ault	-2821
+develop	-2822
+▁dry	-2823
+▁instead	-2824
+▁images	-2825
+▁target	-2826
+hist	-2827
+▁legal	-2828
+irit	-2829
+open	-2830
+tri	-2831
+rown	-2832
+▁innov	-2833
+erve	-2834
+erous	-2835
+▁highly	-2836
+plan	-2837
+thank	-2838
+each	-2839
+ned	-2840
+▁sense	-2841
+paul	-2842
+▁join	-2843
+zz	-2844
+▁remember	-2845
+phil	-2846
+▁attention	-2847
+bay	-2848
+▁inj	-2849
+▁rout	-2850
+▁mid	-2851
+▁heat	-2852
+ashion	-2853
+▁interesting	-2854
+▁sens	-2855
+▁picture	-2856
+▁levels	-2857
+▁bi	-2858
+▁ur	-2859
+who	-2860
+orge	-2861
+▁appe	-2862
+hand	-2863
+▁starting	-2864
+▁star	-2865
+▁policy	-2866
+washing	-2867
+▁average	-2868
+▁interview	-2869
+acing	-2870
+▁answer	-2871
+▁lic	-2872
+ography	-2873
+▁themselves	-2874
+▁stuff	-2875
+uel	-2876
+acking	-2877
+▁consult	-2878
+▁president	-2879
+▁conditions	-2880
+▁item	-2881
+▁depart	-2882
+▁document	-2883
+▁dream	-2884
+▁equipment	-2885
+▁cases	-2886
+▁positive	-2887
+hi	-2888
+▁gift	-2889
+▁healthy	-2890
+▁families	-2891
+apan	-2892
+▁dead	-2893
+▁lower	-2894
+▁contain	-2895
+▁deliver	-2896
+▁player	-2897
+▁qual	-2898
+angel	-2899
+▁sou	-2900
+ipped	-2901
+fire	-2902
+long	-2903
+mart	-2904
+▁businesses	-2905
+build	-2906
+▁firm	-2907
+▁gold	-2908
+▁dise	-2909
+▁streng	-2910
+▁lif	-2911
+two	-2912
+aches	-2913
+▁completely	-2914
+ben	-2915
+▁bur	-2916
+▁effective	-2917
+iny	-2918
+nds	-2919
+apple	-2920
+▁pull	-2921
+▁solution	-2922
+ension	-2923
+▁approach	-2924
+▁wasn	-2925
+▁upon	-2926
+▁achieve	-2927
+atory	-2928
+▁fully	-2929
+▁prices	-2930
+▁multiple	-2931
+▁condition	-2932
+ado	-2933
+▁groups	-2934
+rought	-2935
+inary	-2936
+system	-2937
+▁baby	-2938
+gre	-2939
+▁slow	-2940
+▁shoot	-2941
+review	-2942
+▁prim	-2943
+wa	-2944
+▁began	-2945
+istic	-2946
+▁trip	-2947
+▁stories	-2948
+▁directly	-2949
+▁www	-2950
+▁express	-2951
+▁resources	-2952
+▁wrong	-2953
+lt	-2954
+▁sett	-2955
+▁cat	-2956
+office	-2957
+▁message	-2958
+▁fight	-2959
+uses	-2960
+▁reports	-2961
+▁neigh	-2962
+▁reach	-2963
+▁hour	-2964
+ffee	-2965
+take	-2966
+▁pers	-2967
+▁sites	-2968
+▁insurance	-2969
+note	-2970
+▁wal	-2971
+ctions	-2972
+▁band	-2973
+isl	-2974
+▁opportunities	-2975
+olf	-2976
+anch	-2977
+educ	-2978
+▁neg	-2979
+▁subject	-2980
+dy	-2981
+▁camer	-2982
+▁comments	-2983
+vest	-2984
+▁prem	-2985
+dan	-2986
+▁broad	-2987
+produ	-2988
+▁aim	-2989
+▁fans	-2990
+enty	-2991
+▁defe	-2992
+▁advice	-2993
+ties	-2994
+gold	-2995
+▁types	-2996
+ysis	-2997
+▁except	-2998
+▁attract	-2999
+washington	-3000
+erved	-3001
+▁successful	-3002
+▁wait	-3003
+▁plant	-3004
+tom	-3005
+▁dru	-3006
+aughter	-3007
+▁attempt	-3008
+japan	-3009
+▁adult	-3010
+rock	-3011
+▁offe	-3012
+▁stru	-3013
+▁viol	-3014
+▁suit	-3015
+▁affect	-3016
+▁cert	-3017
+▁sen	-3018
+ruit	-3019
+▁thinking	-3020
+▁implement	-3021
+itation	-3022
+▁opp	-3023
+team	-3024
+▁reve	-3025
+▁solutions	-3026
+▁employees	-3027
+fast	-3028
+itely	-3029
+vision	-3030
+▁fix	-3031
+▁fair	-3032
+comes	-3033
+▁birth	-3034
+▁hop	-3035
+▁obs	-3036
+cert	-3037
+▁ir	-3038
+▁teams	-3039
+▁traditional	-3040
+▁propos	-3041
+▁pictures	-3042
+▁excellent	-3043
+▁creating	-3044
+china	-3045
+▁led	-3046
+rob	-3047
+▁mount	-3048
+onse	-3049
+▁tur	-3050
+inner	-3051
+▁capt	-3052
+wom	-3053
+▁feeling	-3054
+people	-3055
+▁separ	-3056
+▁necessary	-3057
+sum	-3058
+▁saying	-3059
+▁commercial	-3060
+▁goals	-3061
+▁execut	-3062
+bet	-3063
+▁gave	-3064
+show	-3065
+▁dark	-3066
+▁countries	-3067
+▁promot	-3068
+iable	-3069
+▁carry	-3070
+mail	-3071
+▁itself	-3072
+person	-3073
+william	-3074
+director	-3075
+iced	-3076
+▁vot	-3077
+▁extreme	-3078
+▁sleep	-3079
+▁wish	-3080
+tee	-3081
+▁tele	-3082
+▁div	-3083
+▁essential	-3084
+ffects	-3085
+company	-3086
+▁section	-3087
+ores	-3088
+vide	-3089
+det	-3090
+▁serve	-3091
+▁places	-3092
+▁glass	-3093
+flor	-3094
+▁sometimes	-3095
+arts	-3096
+itute	-3097
+sun	-3098
+association	-3099
+▁ble	-3100
+▁club	-3101
+▁har	-3102
+▁imm	-3103
+pan	-3104
+aught	-3105
+vin	-3106
+”.	-3107
+ancy	-3108
+▁blood	-3109
+unte	-3110
+▁wonderful	-3111
+cs	-3112
+mil	-3113
+▁budget	-3114
+uth	-3115
+cher	-3116
+▁growing	-3117
+▁schools	-3118
+ibly	-3119
+▁applications	-3120
+istry	-3121
+▁cancer	-3122
+▁felt	-3123
+▁vehicle	-3124
+▁cross	-3125
+▁sweet	-3126
+▁aware	-3127
+yond	-3128
+court	-3129
+▁spent	-3130
+▁complex	-3131
+▁limited	-3132
+▁delivery	-3133
+▁organiz	-3134
+club	-3135
+bra	-3136
+▁eat	-3137
+▁distribut	-3138
+▁album	-3139
+door	-3140
+boo	-3141
+▁impress	-3142
+▁dro	-3143
+texas	-3144
+fun	-3145
+although	-3146
+▁loved	-3147
+▁progress	-3148
+hest	-3149
+ta	-3150
+ads	-3151
+haps	-3152
+ils	-3153
+▁kn	-3154
+pite	-3155
+michael	-3156
+run	-3157
+aries	-3158
+▁region	-3159
+▁director	-3160
+contact	-3161
+▁correct	-3162
+▁aff	-3163
+▁son	-3164
+▁terms	-3165
+corpor	-3166
+▁smart	-3167
+council	-3168
+▁grand	-3169
+camp	-3170
+live	-3171
+▁speed	-3172
+▁requirements	-3173
+▁block	-3174
+canada	-3175
+▁physical	-3176
+bing	-3177
+▁apart	-3178
+▁temper	-3179
+▁respond	-3180
+oll	-3181
+▁mention	-3182
+▁spend	-3183
+▁weather	-3184
+▁profession	-3185
+▁blue	-3186
+▁developed	-3187
+rang	-3188
+▁shop	-3189
+▁towards	-3190
+▁occur	-3191
+▁individuals	-3192
+▁nut	-3193
+rote	-3194
+opt	-3195
+cell	-3196
+mitted	-3197
+▁conduct	-3198
+▁straight	-3199
+▁heard	-3200
+▁situation	-3201
+▁cold	-3202
+▁finish	-3203
+chan	-3204
+ustr	-3205
+▁sort	-3206
+sol	-3207
+research	-3208
+fa	-3209
+pack	-3210
+sign	-3211
+▁exerc	-3212
+▁wants	-3213
+▁jud	-3214
+itary	-3215
+aint	-3216
+windows	-3217
+asy	-3218
+▁eyes	-3219
+▁nature	-3220
+▁quarter	-3221
+online	-3222
+▁moving	-3223
+indian	-3224
+▁finally	-3225
+▁difference	-3226
+lu	-3227
+pc	-3228
+▁statement	-3229
+▁mark	-3230
+▁race	-3231
+▁device	-3232
+ceed	-3233
+▁dog	-3234
+▁offering	-3235
+made	-3236
+james	-3237
+▁shot	-3238
+aks	-3239
+wise	-3240
+custom	-3241
+▁pattern	-3242
+ady	-3243
+▁ing	-3244
+▁ver	-3245
+sub	-3246
+▁construction	-3247
+ox	-3248
+▁enh	-3249
+▁avoid	-3250
+▁became	-3251
+▁brought	-3252
+▁college	-3253
+▁sty	-3254
+▁feet	-3255
+ka	-3256
+▁convers	-3257
+▁apprec	-3258
+rench	-3259
+main	-3260
+hum	-3261
+hop	-3262
+oon	-3263
+food	-3264
+tor	-3265
+▁stage	-3266
+▁charge	-3267
+▁ru	-3268
+▁pet	-3269
+▁nation	-3270
+ube	-3271
+▁spot	-3272
+▁title	-3273
+internet	-3274
+ological	-3275
+iment	-3276
+▁wedding	-3277
+▁advant	-3278
+keep	-3279
+▁ord	-3280
+▁annual	-3281
+oke	-3282
+test	-3283
+▁kitchen	-3284
+aling	-3285
+lead	-3286
+):	-3287
+rad	-3288
+iber	-3289
+color	-3290
+card	-3291
+only	-3292
+▁wear	-3293
+▁seven	-3294
+european	-3295
+game	-3296
+▁girls	-3297
+▁tried	-3298
+▁turned	-3299
+aled	-3300
+olute	-3301
+▁homes	-3302
+▁hotel	-3303
+▁je	-3304
+oms	-3305
+▁owners	-3306
+".	-3307
+▁imag	-3308
+▁uses	-3309
+rences	-3310
+jun	-3311
+dule	-3312
+▁stress	-3313
+▁language	-3314
+▁investment	-3315
+▁guarant	-3316
+urt	-3317
+▁cas	-3318
+ker	-3319
+fall	-3320
+▁hous	-3321
+▁guide	-3322
+▁helping	-3323
+oves	-3324
+▁ten	-3325
+sey	-3326
+▁ge	-3327
+because	-3328
+▁retail	-3329
+▁mic	-3330
+mount	-3331
+▁rent	-3332
+▁internet	-3333
+▁exactly	-3334
+rant	-3335
+▁wrote	-3336
+▁consist	-3337
+blue	-3338
+▁medic	-3339
+plus	-3340
+▁bath	-3341
+▁nearly	-3342
+usa	-3343
+bc	-3344
+▁pow	-3345
+wall	-3346
+▁partner	-3347
+mine	-3348
+special	-3349
+information	-3350
+cks	-3351
+▁entertain	-3352
+▁term	-3353
+▁categ	-3354
+▁brow	-3355
+▁influ	-3356
+community	-3357
+cul	-3358
+▁maybe	-3359
+general	-3360
+▁offered	-3361
+iance	-3362
+▁active	-3363
+▁client	-3364
+ancing	-3365
+▁strateg	-3366
+▁beginning	-3367
+micro	-3368
+▁double	-3369
+ares	-3370
+▁max	-3371
+▁concept	-3372
+music	-3373
+▁wife	-3374
+▁challenge	-3375
+elect	-3376
+ocr	-3377
+ios	-3378
+icking	-3379
+idered	-3380
+▁posted	-3381
+:||	-3382
+▁bow	-3383
+▁overall	-3384
+ses	-3385
+town	-3386
+dig	-3387
+mad	-3388
+▁path	-3389
+hill	-3390
+conne	-3391
+ustain	-3392
+▁pages	-3393
+▁definitely	-3394
+mont	-3395
+iately	-3396
+league	-3397
+▁pieces	-3398
+▁activity	-3399
+log	-3400
+los	-3401
+ocks	-3402
+ounds	-3403
+▁lots	-3404
+▁ahead	-3405
+▁billion	-3406
+estival	-3407
+▁bottom	-3408
+never	-3409
+opped	-3410
+church	-3411
+▁thous	-3412
+week	-3413
+▁altern	-3414
+mot	-3415
+▁appoint	-3416
+▁decl	-3417
+▁expand	-3418
+▁independ	-3419
+illa	-3420
+wr	-3421
+iles	-3422
+▁inform	-3423
+▁reduce	-3424
+which	-3425
+active	-3426
+mel	-3427
+▁acqu	-3428
+▁finished	-3429
+craft	-3430
+▁benefit	-3431
+roud	-3432
+▁gas	-3433
+icious	-3434
+agn	-3435
+ican	-3436
+▁paid	-3437
+▁devices	-3438
+▁surround	-3439
+▁fat	-3440
+cam	-3441
+▁thank	-3442
+ago	-3443
+▁hus	-3444
+▁farm	-3445
+police	-3446
+lee	-3447
+▁sports	-3448
+sou	-3449
+antly	-3450
+▁collabor	-3451
+cong	-3452
+▁miles	-3453
+▁extend	-3454
+▁response	-3455
+▁excited	-3456
+smith	-3457
+▁decor	-3458
+look	-3459
+▁emot	-3460
+▁length	-3461
+management	-3462
+follow	-3463
+▁exhib	-3464
+▁advert	-3465
+▁earlier	-3466
+▁comfortable	-3467
+ml	-3468
+bell	-3469
+▁easier	-3470
+inte	-3471
+page	-3472
+watch	-3473
+▁adjust	-3474
+▁corpor	-3475
+▁po	-3476
+atever	-3477
+island	-3478
+anta	-3479
+writ	-3480
+▁largest	-3481
+▁conference	-3482
+amin	-3483
+▁eight	-3484
+▁dating	-3485
+▁mer	-3486
+▁assess	-3487
+florida	-3488
+▁transport	-3489
+boy	-3490
+vant	-3491
+▁states	-3492
+▁opening	-3493
+▁belie	-3494
+▁demand	-3495
+rd	-3496
+van	-3497
+▁served	-3498
+▁storage	-3499
+▁flav	-3500
+isions	-3501
+▁damage	-3502
+▁estate	-3503
+start	-3504
+▁analy	-3505
+gal	-3506
+exper	-3507
+▁candid	-3508
+lake	-3509
+▁bank	-3510
+▁jobs	-3511
+▁tips	-3512
+quest	-3513
+▁thanks	-3514
+oper	-3515
+oving	-3516
+▁coffee	-3517
+jer	-3518
+▁lik	-3519
+during	-3520
+▁culture	-3521
+▁named	-3522
+▁pen	-3523
+▁sem	-3524
+▁pred	-3525
+▁numbers	-3526
+▁completed	-3527
+print	-3528
+▁background	-3529
+arc	-3530
+▁rates	-3531
+russ	-3532
+fit	-3533
+▁histor	-3534
+australia	-3535
+▁rock	-3536
+▁understanding	-3537
+ivil	-3538
+▁cash	-3539
+▁immediately	-3540
+▁driving	-3541
+bank	-3542
+working	-3543
+▁responsible	-3544
+▁plus	-3545
+▁considered	-3546
+▁creative	-3547
+lying	-3548
+▁mis	-3549
+organ	-3550
+undred	-3551
+▁father	-3552
+▁gradu	-3553
+▁truly	-3554
+▁afford	-3555
+▁setting	-3556
+hens	-3557
+media	-3558
+cd	-3559
+compl	-3560
+author	-3561
+ails	-3562
+za	-3563
+▁haven	-3564
+▁cards	-3565
+▁fashion	-3566
+vict	-3567
+bul	-3568
+▁beyond	-3569
+onstr	-3570
+▁wine	-3571
+ota	-3572
+oil	-3573
+jes	-3574
+sell	-3575
+azine	-3576
+▁experienced	-3577
+nded	-3578
+▁knew	-3579
+▁analysis	-3580
+cel	-3581
+family	-3582
+▁occas	-3583
+▁normal	-3584
+download	-3585
+icing	-3586
+▁button	-3587
+education	-3588
+ously	-3589
+▁accur	-3590
+▁owner	-3591
+▁reviews	-3592
+orial	-3593
+▁welcome	-3594
+▁purpose	-3595
+▁prefer	-3596
+▁husband	-3597
+aded	-3598
+ategy	-3599
+thom	-3600
+▁cream	-3601
+ampions	-3602
+central	-3603
+rooms	-3604
+▁pool	-3605
+▁regul	-3606
+va	-3607
+ald	-3608
+ocolate	-3609
+▁frame	-3610
+▁colors	-3611
+▁runs	-3612
+▁talking	-3613
+▁disease	-3614
+▁political	-3615
+chic	-3616
+▁degree	-3617
+publ	-3618
+▁guys	-3619
+lement	-3620
+▁threat	-3621
+iar	-3622
+▁machine	-3623
+republic	-3624
+▁shopping	-3625
+▁ingred	-3626
+▁holiday	-3627
+▁born	-3628
+text	-3629
+cat	-3630
+anned	-3631
+arily	-3632
+▁train	-3633
+icken	-3634
+sil	-3635
+▁satis	-3636
+▁initial	-3637
+priate	-3638
+▁launch	-3639
+▁onto	-3640
+posted	-3641
+brown	-3642
+▁park	-3643
+▁session	-3644
+lou	-3645
+▁paint	-3646
+▁sugar	-3647
+▁changed	-3648
+▁guy	-3649
+▁files	-3650
+jour	-3651
+▁looked	-3652
+case	-3653
+▁slight	-3654
+▁evening	-3655
+pare	-3656
+▁indic	-3657
+▁fem	-3658
+istered	-3659
+▁highest	-3660
+▁helped	-3661
+grand	-3662
+▁shipping	-3663
+▁prec	-3664
+aud	-3665
+▁volunte	-3666
+▁surv	-3667
+equ	-3668
+▁lines	-3669
+ences	-3670
+▁object	-3671
+▁fill	-3672
+children	-3673
+▁increased	-3674
+help	-3675
+▁proud	-3676
+▁himself	-3677
+pd	-3678
+▁opin	-3679
+weight	-3680
+▁proced	-3681
+lor	-3682
+▁clin	-3683
+▁artist	-3684
+bow	-3685
+▁egg	-3686
+oph	-3687
+▁spirit	-3688
+▁particularly	-3689
+fund	-3690
+▁drop	-3691
+through	-3692
+▁camera	-3693
+▁winter	-3694
+▁federal	-3695
+district	-3696
+▁fant	-3697
+▁mission	-3698
+ws	-3699
+▁daughter	-3700
+itect	-3701
+▁strugg	-3702
+▁traffic	-3703
+▁absolute	-3704
+lord	-3705
+pass	-3706
+product	-3707
+▁journey	-3708
+▁produce	-3709
+ris	-3710
+▁craft	-3711
+▁rules	-3712
+science	-3713
+▁pressure	-3714
+hu	-3715
+offic	-3716
+rey	-3717
+istics	-3718
+▁models	-3719
+▁economic	-3720
+copy	-3721
+▁allowed	-3722
+▁selection	-3723
+iling	-3724
+▁patient	-3725
+full	-3726
+atl	-3727
+lies	-3728
+▁spread	-3729
+▁highlight	-3730
+const	-3731
+▁incred	-3732
+▁vac	-3733
+profess	-3734
+div	-3735
+thes	-3736
+▁efforts	-3737
+union	-3738
+▁stream	-3739
+▁secure	-3740
+going	-3741
+▁spring	-3742
+▁existing	-3743
+han	-3744
+rapy	-3745
+▁bul	-3746
+▁increasing	-3747
+▁furn	-3748
+▁videos	-3749
+▁subst	-3750
+clusive	-3751
+▁letter	-3752
+french	-3753
+▁peace	-3754
+british	-3755
+pects	-3756
+mes	-3757
+▁bill	-3758
+tit	-3759
+obama	-3760
+▁behav	-3761
+▁residents	-3762
+▁manager	-3763
+master	-3764
+▁frequ	-3765
+▁mail	-3766
+▁relax	-3767
+▁fan	-3768
+hd	-3769
+project	-3770
+▁measure	-3771
+social	-3772
+xim	-3773
+contin	-3774
+▁heavy	-3775
+▁trade	-3776
+▁couldn	-3777
+▁middle	-3778
+▁powerful	-3779
+▁challenges	-3780
+▁beauty	-3781
+▁certainly	-3782
+icon	-3783
+▁basis	-3784
+fortun	-3785
+ologies	-3786
+development	-3787
+porary	-3788
+▁subscri	-3789
+resp	-3790
+▁obtain	-3791
+▁shared	-3792
+▁dedicated	-3793
+▁smooth	-3794
+▁liter	-3795
+unk	-3796
+▁update	-3797
+▁moved	-3798
+▁seeing	-3799
+▁multi	-3800
+▁gain	-3801
+▁wild	-3802
+lished	-3803
+▁cru	-3804
+dec	-3805
+mill	-3806
+annel	-3807
+▁payment	-3808
+▁taste	-3809
+uce	-3810
+acity	-3811
+▁links	-3812
+support	-3813
+▁schedule	-3814
+iger	-3815
+▁ban	-3816
+▁rese	-3817
+includ	-3818
+▁alone	-3819
+▁instruct	-3820
+somet	-3821
+whether	-3822
+▁cars	-3823
+▁gone	-3824
+encies	-3825
+▁science	-3826
+▁demonstr	-3827
+▁organizations	-3828
+imal	-3829
+▁whose	-3830
+▁transfer	-3831
+blog	-3832
+pping	-3833
+▁refer	-3834
+▁hundred	-3835
+roid	-3836
+▁aren	-3837
+▁brother	-3838
+term	-3839
+oe	-3840
+▁supply	-3841
+▁effects	-3842
+ying	-3843
+▁ton	-3844
+front	-3845
+pared	-3846
+▁protection	-3847
+▁basic	-3848
+▁steps	-3849
+rian	-3850
+▁characters	-3851
+uled	-3852
+▁competition	-3853
+▁church	-3854
+ico	-3855
+apter	-3856
+▁guests	-3857
+river	-3858
+standing	-3859
+▁communities	-3860
+▁passion	-3861
+gn	-3862
+tre	-3863
+fol	-3864
+ona	-3865
+▁graph	-3866
+▁extremely	-3867
+oices	-3868
+▁butter	-3869
+ronic	-3870
+▁cast	-3871
+▁compon	-3872
+▁finding	-3873
+western	-3874
+apers	-3875
+iture	-3876
+▁forget	-3877
+▁reasons	-3878
+times	-3879
+foundation	-3880
+sequ	-3881
+▁strength	-3882
+▁unit	-3883
+next	-3884
+▁posts	-3885
+aming	-3886
+▁garden	-3887
+▁sched	-3888
+bour	-3889
+ened	-3890
+flow	-3891
+▁notice	-3892
+▁scient	-3893
+▁fa	-3894
+▁hospital	-3895
+ht	-3896
+ni	-3897
+learn	-3898
+anging	-3899
+ergency	-3900
+commission	-3901
+▁guid	-3902
+▁ideal	-3903
+▁happened	-3904
+▁rob	-3905
+site	-3906
+ston	-3907
+asure	-3908
+before	-3909
+▁shown	-3910
+▁voice	-3911
+▁manufacture	-3912
+▁coach	-3913
+▁leader	-3914
+▁determin	-3915
+bill	-3916
+▁breat	-3917
+▁remind	-3918
+ari	-3919
+▁academ	-3920
+▁eval	-3921
+security	-3922
+▁au	-3923
+foot	-3924
+▁waiting	-3925
+both	-3926
+osing	-3927
+▁batter	-3928
+alex	-3929
+▁luck	-3930
+▁classes	-3931
+rief	-3932
+▁solid	-3933
+stone	-3934
+▁beg	-3935
+▁pair	-3936
+▁biggest	-3937
+▁meaning	-3938
+▁although	-3939
+▁experiences	-3940
+▁stick	-3941
+tot	-3942
+engine	-3943
+▁lack	-3944
+▁trou	-3945
+▁adding	-3946
+▁passed	-3947
+gers	-3948
+▁views	-3949
+▁covered	-3950
+▁fear	-3951
+▁regarding	-3952
+num	-3953
+▁beat	-3954
+another	-3955
+berg	-3956
+▁snow	-3957
+tur	-3958
+▁cit	-3959
+▁gun	-3960
+▁score	-3961
+▁senior	-3962
+▁surface	-3963
+mi	-3964
+lied	-3965
+ifying	-3966
+azon	-3967
+itar	-3968
+apt	-3969
+aven	-3970
+rial	-3971
+▁contains	-3972
+cc	-3973
+▁sand	-3974
+▁scene	-3975
+ams	-3976
+fre	-3977
+sand	-3978
+▁electric	-3979
+orm	-3980
+irection	-3981
+▁compared	-3982
+▁standards	-3983
+ration	-3984
+ya	-3985
+zer	-3986
+ordin	-3987
+lyn	-3988
+▁continues	-3989
+▁associated	-3990
+olar	-3991
+▁conven	-3992
+▁advantage	-3993
+bas	-3994
+ags	-3995
+video	-3996
+▁disapp	-3997
+data	-3998
+tour	-3999
+▁window	-4000
+nic	-4001
+▁ju	-4002
+rick	-4003
+orney	-4004
+▁den	-4005
+enter	-4006
+▁starts	-4007
+▁classic	-4008
+▁package	-4009
+grade	-4010
+▁minute	-4011
+▁calls	-4012
+▁names	-4013
+iques	-4014
+▁throw	-4015
+▁continued	-4016
+▁died	-4017
+▁recipe	-4018
+iverse	-4019
+▁master	-4020
+▁military	-4021
+dou	-4022
+commod	-4023
+▁und	-4024
+▁experts	-4025
+▁professionals	-4026
+▁ice	-4027
+mical	-4028
+related	-4029
+▁dinner	-4030
+▁keeping	-4031
+▁artists	-4032
+england	-4033
+ipes	-4034
+▁guess	-4035
+▁figure	-4036
+ara	-4037
+▁dang	-4038
+▁closed	-4039
+▁talent	-4040
+▁ran	-4041
+▁epis	-4042
+▁district	-4043
+▁pm	-4044
+noon	-4045
+store	-4046
+▁manage	-4047
+institute	-4048
+▁wat	-4049
+▁cab	-4050
+beach	-4051
+their	-4052
+▁followed	-4053
+▁technical	-4054
+wis	-4055
+▁circ	-4056
+▁larger	-4057
+▁reflect	-4058
+gery	-4059
+▁nove	-4060
+illage	-4061
+valley	-4062
+▁connection	-4063
+mass	-4064
+atures	-4065
+▁listed	-4066
+▁sustain	-4067
+▁football	-4068
+▁cra	-4069
+▁knows	-4070
+▁trail	-4071
+cm	-4072
+▁critical	-4073
+▁awes	-4074
+ula	-4075
+▁friendly	-4076
+▁jump	-4077
+▁strategy	-4078
+▁laun	-4079
+▁fabric	-4080
+allery	-4081
+▁fish	-4082
+▁capital	-4083
+mex	-4084
+▁shape	-4085
+▁officials	-4086
+aly	-4087
+jim	-4088
+▁income	-4089
+▁audience	-4090
+▁resident	-4091
+carol	-4092
+▁virt	-4093
+▁winning	-4094
+fair	-4095
+▁medium	-4096
+oma	-4097
+▁gra	-4098
+▁opened	-4099
+▁studies	-4100
+▁efficient	-4101
+bes	-4102
+elling	-4103
+ette	-4104
+▁teaching	-4105
+▁sym	-4106
+▁facilities	-4107
+▁fol	-4108
+quality	-4109
+▁developing	-4110
+▁ship	-4111
+▁advent	-4112
+▁tim	-4113
+▁bright	-4114
+hard	-4115
+▁motor	-4116
+▁doctor	-4117
+▁learned	-4118
+▁restaurant	-4119
+mu	-4120
+okes	-4121
+▁salt	-4122
+connect	-4123
+▁arrest	-4124
+▁watching	-4125
+wing	-4126
+season	-4127
+▁kept	-4128
+▁ingredients	-4129
+women	-4130
+▁handle	-4131
+▁awesome	-4132
+ums	-4133
+▁placed	-4134
+▁thousands	-4135
+level	-4136
+▁poor	-4137
+▁initi	-4138
+ements	-4139
+▁entry	-4140
+▁bought	-4141
+▁station	-4142
+▁workers	-4143
+▁rich	-4144
+acement	-4145
+▁exciting	-4146
+▁leaves	-4147
+▁leaving	-4148
+▁grant	-4149
+▁produced	-4150
+▁updated	-4151
+▁institut	-4152
+conference	-4153
+spring	-4154
+igation	-4155
+edy	-4156
+ki	-4157
+vi	-4158
+scott	-4159
+ishes	-4160
+▁older	-4161
+▁established	-4162
+▁greater	-4163
+▁appropriate	-4164
+ixt	-4165
+antic	-4166
+▁princ	-4167
+▁neighbor	-4168
+ech	-4169
+▁explore	-4170
+franc	-4171
+▁plenty	-4172
+series	-4173
+yer	-4174
+cles	-4175
+venue	-4176
+▁force	-4177
+▁spons	-4178
+itors	-4179
+▁catch	-4180
+▁presented	-4181
+jul	-4182
+chicago	-4183
+▁famous	-4184
+▁tickets	-4185
+▁monitor	-4186
+max	-4187
+inese	-4188
+inger	-4189
+brand	-4190
+ograp	-4191
+▁speak	-4192
+▁sharing	-4193
+gi	-4194
+happ	-4195
+type	-4196
+▁claims	-4197
+▁facility	-4198
+bling	-4199
+looking	-4200
+▁cry	-4201
+jesus	-4202
+▁rooms	-4203
+▁department	-4204
+aa	-4205
+access	-4206
+atically	-4207
+▁testing	-4208
+size	-4209
+award	-4210
+▁requires	-4211
+▁whatever	-4212
+▁announce	-4213
+▁sport	-4214
+▁remove	-4215
+▁mo	-4216
+lands	-4217
+wil	-4218
+party	-4219
+▁veget	-4220
+academ	-4221
+walk	-4222
+amazon	-4223
+invest	-4224
+▁driver	-4225
+▁victim	-4226
+▁vote	-4227
+▁theme	-4228
+▁advance	-4229
+jud	-4230
+pac	-4231
+zen	-4232
+▁remains	-4233
+▁useful	-4234
+▁becoming	-4235
+ups	-4236
+tech	-4237
+ooper	-4238
+▁shall	-4239
+▁buying	-4240
+email	-4241
+▁prepar	-4242
+▁designs	-4243
+▁properties	-4244
+ctors	-4245
+tenance	-4246
+▁discover	-4247
+▁flat	-4248
+▁structure	-4249
+▁swe	-4250
+▁yester	-4251
+rome	-4252
+atively	-4253
+code	-4254
+▁map	-4255
+cost	-4256
+▁flex	-4257
+”,	-4258
+prote	-4259
+commit	-4260
+▁selected	-4261
+cin	-4262
+▁stores	-4263
+report	-4264
+▁repair	-4265
+▁vision	-4266
+berry	-4267
+osure	-4268
+▁tree	-4269
+actions	-4270
+▁brings	-4271
+▁beach	-4272
+▁diagn	-4273
+▁tables	-4274
+▁happens	-4275
+▁valu	-4276
+muse	-4277
+lastname	-4278
+join	-4279
+being	-4280
+itude	-4281
+price	-4282
+▁desk	-4283
+▁signed	-4284
+▁street	-4285
+christian	-4286
+girl	-4287
+agram	-4288
+chris	-4289
+pective	-4290
+origin	-4291
+▁serving	-4292
+??	-4293
+oom	-4294
+ifts	-4295
+▁balance	-4296
+▁bra	-4297
+▁alleg	-4298
+▁plants	-4299
+fortunately	-4300
+▁menu	-4301
+burg	-4302
+books	-4303
+▁crow	-4304
+▁plastic	-4305
+gas	-4306
+bles	-4307
+▁articles	-4308
+vere	-4309
+▁yesterday	-4310
+days	-4311
+uries	-4312
+ximately	-4313
+▁journal	-4314
+atab	-4315
+▁rain	-4316
+▁chocolate	-4317
+iency	-4318
+▁perhaps	-4319
+outs	-4320
+▁memory	-4321
+▁primary	-4322
+ira	-4323
+ped	-4324
+▁anti	-4325
+▁wouldn	-4326
+ec	-4327
+▁fee	-4328
+mond	-4329
+▁scen	-4330
+chinese	-4331
+▁operations	-4332
+▁ul	-4333
+▁secret	-4334
+▁direction	-4335
+cohol	-4336
+government	-4337
+hotel	-4338
+▁pray	-4339
+▁prepared	-4340
+▁drug	-4341
+▁funds	-4342
+▁library	-4343
+▁spending	-4344
+haw	-4345
+claim	-4346
+society	-4347
+▁temperature	-4348
+got	-4349
+▁actual	-4350
+▁corner	-4351
+▁economy	-4352
+ading	-4353
+george	-4354
+limited	-4355
+▁status	-4356
+known	-4357
+▁util	-4358
+▁movement	-4359
+enses	-4360
+▁faith	-4361
+▁meant	-4362
+hab	-4363
+sex	-4364
+tic	-4365
+rael	-4366
+▁mist	-4367
+▁discount	-4368
+wer	-4369
+▁mechan	-4370
+ceo	-4371
+▁cris	-4372
+lock	-4373
+▁resist	-4374
+frank	-4375
+▁ride	-4376
+copyright	-4377
+▁numerous	-4378
+▁raised	-4379
+",	-4380
+chief	-4381
+▁busy	-4382
+greg	-4383
+africa	-4384
+▁forms	-4385
+medical	-4386
+▁orders	-4387
+▁changing	-4388
+▁syn	-4389
+sn	-4390
+▁locations	-4391
+gra	-4392
+dered	-4393
+vention	-4394
+adv	-4395
+▁focused	-4396
+▁exercise	-4397
+▁communication	-4398
+ket	-4399
+aper	-4400
+▁dep	-4401
+▁core	-4402
+link	-4403
+journal	-4404
+▁injury	-4405
+agon	-4406
+▁eating	-4407
+**	-4408
+odes	-4409
+creat	-4410
+abilities	-4411
+tem	-4412
+mid	-4413
+▁youth	-4414
+microsoft	-4415
+ci	-4416
+israel	-4417
+▁notes	-4418
+▁independent	-4419
+sea	-4420
+▁listen	-4421
+▁survey	-4422
+▁showing	-4423
+!!!	-4424
+riage	-4425
+▁favour	-4426
+rict	-4427
+▁obvious	-4428
+share	-4429
+▁rare	-4430
+lets	-4431
+▁advanced	-4432
+far	-4433
+▁brown	-4434
+congress	-4435
+essions	-4436
+▁afternoon	-4437
+▁psy	-4438
+lines	-4439
+trump	-4440
+▁putting	-4441
+▁evidence	-4442
+▁combination	-4443
+technology	-4444
+▁worry	-4445
+ini	-4446
+▁wire	-4447
+▁records	-4448
+cos	-4449
+▁ended	-4450
+umin	-4451
+▁suff	-4452
+source	-4453
+▁animals	-4454
+▁methods	-4455
+▁reality	-4456
+▁practices	-4457
+pu	-4458
+dev	-4459
+▁drink	-4460
+▁slightly	-4461
+pad	-4462
+anth	-4463
+iences	-4464
+want	-4465
+second	-4466
+ss	-4467
+▁micro	-4468
+current	-4469
+▁enjoyed	-4470
+▁caused	-4471
+▁missing	-4472
+die	-4473
+hun	-4474
+▁limit	-4475
+sports	-4476
+kit	-4477
+▁determine	-4478
+kin	-4479
+▁tend	-4480
+▁metal	-4481
+gb	-4482
+maybe	-4483
+▁agency	-4484
+▁sources	-4485
+centre	-4486
+▁sector	-4487
+▁showed	-4488
+ded	-4489
+▁cleaning	-4490
+▁operating	-4491
+fi	-4492
+▁inf	-4493
+▁environmental	-4494
+▁estim	-4495
+mike	-4496
+▁lux	-4497
+indust	-4498
+▁joined	-4499
+tru	-4500
+olic	-4501
+▁diet	-4502
+▁shel	-4503
+▁relig	-4504
+bob	-4505
+meet	-4506
+▁mill	-4507
+bud	-4508
+▁flu	-4509
+▁dance	-4510
+▁compre	-4511
+▁teachers	-4512
+▁shut	-4513
+peter	-4514
+▁thoughts	-4515
+hus	-4516
+imp	-4517
+estic	-4518
+ivered	-4519
+▁sch	-4520
+ander	-4521
+robert	-4522
+▁updates	-4523
+need	-4524
+ript	-4525
+▁funding	-4526
+▁organis	-4527
+sor	-4528
+eter	-4529
+earth	-4530
+▁capacity	-4531
+▁dete	-4532
+▁lose	-4533
+employ	-4534
+igan	-4535
+isher	-4536
+perform	-4537
+▁colour	-4538
+▁profile	-4539
+▁assistance	-4540
+▁nar	-4541
+▁songs	-4542
+rupt	-4543
+▁visual	-4544
+bridge	-4545
+ampionship	-4546
+▁appears	-4547
+▁accident	-4548
+▁visitors	-4549
+did	-4550
+ownt	-4551
+library	-4552
+▁previously	-4553
+available	-4554
+anna	-4555
+hope	-4556
+scre	-4557
+▁repe	-4558
+▁steel	-4559
+lie	-4560
+▁thr	-4561
+▁chair	-4562
+▁filled	-4563
+azz	-4564
+hon	-4565
+▁du	-4566
+those	-4567
+▁brain	-4568
+▁sea	-4569
+▁south	-4570
+android	-4571
+▁managed	-4572
+jeff	-4573
+ception	-4574
+ales	-4575
+▁cart	-4576
+▁swim	-4577
+▁decide	-4578
+▁corporate	-4579
+▁coll	-4580
+democr	-4581
+▁charges	-4582
+ili	-4583
+ferences	-4584
+▁markets	-4585
+inch	-4586
+▁deb	-4587
+istan	-4588
+▁https	-4589
+▁honest	-4590
+▁chat	-4591
+▁spl	-4592
+▁neck	-4593
+▁cheese	-4594
+▁allowing	-4595
+squ	-4596
+▁raise	-4597
+▁stret	-4598
+building	-4599
+wild	-4600
+▁rev	-4601
+▁extensive	-4602
+acks	-4603
+▁teacher	-4604
+▁accommod	-4605
+edom	-4606
+global	-4607
+▁helpful	-4608
+▁ath	-4609
+louis	-4610
+▁legisl	-4611
+▁central	-4612
+▁walking	-4613
+▁committed	-4614
+ett	-4615
+away	-4616
+▁firstname	-4617
+hr	-4618
+▁lunch	-4619
+hensive	-4620
+▁absolutely	-4621
+▁fold	-4622
+energy	-4623
+ervation	-4624
+▁documents	-4625
+ellig	-4626
+manager	-4627
+▁asking	-4628
+▁coverage	-4629
+assist	-4630
+▁vehicles	-4631
+▁pal	-4632
+francis	-4633
+iration	-4634
+aked	-4635
+inn	-4636
+pop	-4637
+uty	-4638
+appro	-4639
+welcome	-4640
+▁outdoor	-4641
+▁smaller	-4642
+pract	-4643
+▁tast	-4644
+▁selling	-4645
+▁register	-4646
+style	-4647
+digital	-4648
+gon	-4649
+▁milk	-4650
+▁inspired	-4651
+▁leadership	-4652
+parent	-4653
+▁promote	-4654
+▁decisions	-4655
+human	-4656
+▁architect	-4657
+pot	-4658
+rose	-4659
+▁north	-4660
+▁population	-4661
+?”	-4662
+pow	-4663
+▁earth	-4664
+ola	-4665
+sky	-4666
+tun	-4667
+network	-4668
+▁reached	-4669
+committee	-4670
+▁elements	-4671
+▁relationships	-4672
+▁counter	-4673
+▁establish	-4674
+▁expensive	-4675
+▁partnership	-4676
+france	-4677
+angeles	-4678
+pie	-4679
+▁sauce	-4680
+medic	-4681
+▁parties	-4682
+▁delicious	-4683
+py	-4684
+bac	-4685
+dog	-4686
+▁becomes	-4687
+▁dy	-4688
+legal	-4689
+▁agree	-4690
+chair	-4691
+▁lovely	-4692
+▁windows	-4693
+▁concerns	-4694
+▁expertise	-4695
+admin	-4696
+▁topic	-4697
+▁winner	-4698
+▁society	-4699
+▁van	-4700
+inste	-4701
+kings	-4702
+comple	-4703
+▁radio	-4704
+four	-4705
+▁spect	-4706
+▁upcoming	-4707
+nut	-4708
+icip	-4709
+travel	-4710
+▁willing	-4711
+aser	-4712
+photo	-4713
+iers	-4714
+▁registered	-4715
+non	-4716
+▁relevant	-4717
+saf	-4718
+worth	-4719
+▁truth	-4720
+▁rid	-4721
+▁launched	-4722
+▁connected	-4723
+▁fourth	-4724
+▁century	-4725
+omb	-4726
+▁task	-4727
+atin	-4728
+suit	-4729
+▁sounds	-4730
+▁bac	-4731
+▁format	-4732
+▁techniques	-4733
+buy	-4734
+▁movies	-4735
+▁eth	-4736
+▁fees	-4737
+ocket	-4738
+cook	-4739
+▁doll	-4740
+▁plays	-4741
+pur	-4742
+osis	-4743
+▁kid	-4744
+▁instru	-4745
+▁despite	-4746
+▁bon	-4747
+▁situ	-4748
+museum	-4749
+rics	-4750
+▁enhance	-4751
+!)	-4752
+asp	-4753
+omin	-4754
+prise	-4755
+▁approximately	-4756
+plo	-4757
+three	-4758
+▁loan	-4759
+ym	-4760
+rie	-4761
+born	-4762
+▁joint	-4763
+▁defense	-4764
+▁anywhere	-4765
+ius	-4766
+▁citiz	-4767
+▁values	-4768
+lot	-4769
+prov	-4770
+▁stre	-4771
+thomas	-4772
+▁explain	-4773
+▁dur	-4774
+▁civil	-4775
+▁panel	-4776
+▁innovative	-4777
+▁commitment	-4778
+▁discussion	-4779
+bled	-4780
+▁delivered	-4781
+▁operation	-4782
+kel	-4783
+▁essay	-4784
+hom	-4785
+virgin	-4786
+eless	-4787
+▁readers	-4788
+▁educational	-4789
+hou	-4790
+ubs	-4791
+kes	-4792
+▁nav	-4793
+▁separate	-4794
+een	-4795
+▁maintenance	-4796
+▁recommended	-4797
+prem	-4798
+secut	-4799
+▁surgery	-4800
+def	-4801
+▁meat	-4802
+dc	-4803
+▁sav	-4804
+ita	-4805
+oles	-4806
+▁feels	-4807
+▁waste	-4808
+▁tun	-4809
+month	-4810
+▁generally	-4811
+▁assign	-4812
+ero	-4813
+▁rap	-4814
+▁toward	-4815
+▁courses	-4816
+▁controll	-4817
+cov	-4818
+▁mental	-4819
+mur	-4820
+▁cheap	-4821
+▁fellow	-4822
+sem	-4823
+▁parking	-4824
+job	-4825
+▁identify	-4826
+fed	-4827
+cean	-4828
+idering	-4829
+▁encourage	-4830
+▁fantastic	-4831
+▁presence	-4832
+▁optim	-4833
+▁killed	-4834
+cription	-4835
+▁calling	-4836
+▁emergency	-4837
+ensions	-4838
+▁cooking	-4839
+cape	-4840
+lind	-4841
+▁tea	-4842
+▁photograph	-4843
+does	-4844
+▁argu	-4845
+elines	-4846
+icies	-4847
+▁roof	-4848
+▁editor	-4849
+jon	-4850
+icians	-4851
+▁sympt	-4852
+▁agreement	-4853
+nel	-4854
+rings	-4855
+▁edge	-4856
+rid	-4857
+▁adults	-4858
+▁landsc	-4859
+▁feedback	-4860
+▁vs	-4861
+ipal	-4862
+making	-4863
+▁charged	-4864
+hell	-4865
+▁joy	-4866
+▁nine	-4867
+▁chicken	-4868
+▁guest	-4869
+archers	-4870
+▁arrang	-4871
+sar	-4872
+games	-4873
+▁seat	-4874
+▁server	-4875
+▁hearing	-4876
+▁sitting	-4877
+▁processes	-4878
+▁sets	-4879
+▁fuel	-4880
+▁rise	-4881
+▁appearance	-4882
+▁consumers	-4883
+▁mur	-4884
+▁mostly	-4885
+▁academic	-4886
+▁combined	-4887
+eed	-4888
+hal	-4889
+stock	-4890
+▁od	-4891
+▁accompl	-4892
+cred	-4893
+▁fruit	-4894
+asia	-4895
+▁signs	-4896
+▁generation	-4897
+▁trial	-4898
+▁featured	-4899
+▁executive	-4900
+▁installation	-4901
+irmed	-4902
+letter	-4903
+▁intended	-4904
+▁prepare	-4905
+▁tight	-4906
+chester	-4907
+▁transl	-4908
+▁teen	-4909
+▁furniture	-4910
+▁therefore	-4911
+ko	-4912
+shop	-4913
+nament	-4914
+▁category	-4915
+▁rat	-4916
+local	-4917
+communic	-4918
+wi	-4919
+gent	-4920
+heart	-4921
+▁bowl	-4922
+▁push	-4923
+little	-4924
+▁cities	-4925
+▁provider	-4926
+xx	-4927
+cross	-4928
+▁infr	-4929
+▁detailed	-4930
+ifications	-4931
+jac	-4932
+▁kick	-4933
+▁factors	-4934
+▁dogs	-4935
+▁alternative	-4936
+▁brief	-4937
+▁breast	-4938
+▁properly	-4939
+burn	-4940
+▁battle	-4941
+▁responsibility	-4942
+bost	-4943
+iled	-4944
+▁seconds	-4945
+▁confident	-4946
+▁motiv	-4947
+johnson	-4948
+▁studio	-4949
+▁confidence	-4950
+clean	-4951
+▁brands	-4952
+▁info	-4953
+martin	-4954
+agement	-4955
+▁chemical	-4956
+span	-4957
+ulated	-4958
+▁participate	-4959
+amb	-4960
+drive	-4961
+atabase	-4962
+▁sister	-4963
+▁removed	-4964
+unning	-4965
+▁exchange	-4966
+▁majority	-4967
+▁entertainment	-4968
+rat	-4969
+matt	-4970
+happy	-4971
+▁mort	-4972
+▁holding	-4973
+▁bringing	-4974
+▁measures	-4975
+▁transform	-4976
+kn	-4977
+kins	-4978
+young	-4979
+▁approved	-4980
+▁kill	-4981
+▁meal	-4982
+othing	-4983
+record	-4984
+▁caught	-4985
+▁purchased	-4986
+ja	-4987
+secret	-4988
+▁tough	-4989
+▁novel	-4990
+▁seeking	-4991
+▁scheduled	-4992
+▁ven	-4993
+▁lock	-4994
+▁described	-4995
+guide	-4996
+▁female	-4997
+▁occasion	-4998
+▁websites	-4999
+joe	-5000
+festival	-5001
+▁participants	-5002
+jose	-5003
+ulty	-5004
+visit	-5005
+▁clim	-5006
+▁importance	-5007
+▁achie	-5008
+▁writer	-5009
+students	-5010
+▁totally	-5011
+▁distance	-5012
+dro	-5013
+priv	-5014
+▁stars	-5015
+americans	-5016
+▁returned	-5017
+jones	-5018
+applic	-5019
+▁psych	-5020
+interest	-5021
+execut	-5022
+▁minor	-5023
+ager	-5024
+▁rom	-5025
+▁opinion	-5026
+kar	-5027
+icate	-5028
+canad	-5029
+▁satisf	-5030
+▁wearing	-5031
+ushed	-5032
+econom	-5033
+▁guarantee	-5034
+anda	-5035
+▁ment	-5036
+▁sides	-5037
+▁thick	-5038
+▁familiar	-5039
+▁performed	-5040
+table	-5041
+▁contem	-5042
+▁supporting	-5043
+dam	-5044
+iner	-5045
+prene	-5046
+mother	-5047
+▁centre	-5048
+turn	-5049
+▁hundreds	-5050
+wat	-5051
+▁shap	-5052
+▁maximum	-5053
+▁appeared	-5054
+hors	-5055
+▁vir	-5056
+▁ring	-5057
+▁coast	-5058
+▁units	-5059
+sometimes	-5060
+australian	-5061
+ols	-5062
+paper	-5063
+▁seek	-5064
+boston	-5065
+control	-5066
+▁topics	-5067
+hospital	-5068
+▁restrict	-5069
+▁investigation	-5070
+asks	-5071
+expl	-5072
+space	-5073
+▁monthly	-5074
+▁ourselves	-5075
+trust	-5076
+▁fell	-5077
+ushing	-5078
+▁teach	-5079
+▁switch	-5080
+▁climate	-5081
+mony	-5082
+ulture	-5083
+▁installed	-5084
+otton	-5085
+host	-5086
+kore	-5087
+▁doors	-5088
+reprene	-5089
+▁depending	-5090
+inct	-5091
+▁comprehensive	-5092
+break	-5093
+▁trees	-5094
+▁magazine	-5095
+pdf	-5096
+oga	-5097
+irds	-5098
+▁mentioned	-5099
+▁worldwide	-5100
+try	-5101
+alog	-5102
+▁quiet	-5103
+▁sizes	-5104
+▁sexual	-5105
+▁clearly	-5106
+hs	-5107
+coast	-5108
+▁leads	-5109
+▁foreign	-5110
+▁impl	-5111
+▁begins	-5112
+▁supported	-5113
+guard	-5114
+ulous	-5115
+▁gather	-5116
+▁episode	-5117
+▁conversation	-5118
+map	-5119
+▁lap	-5120
+itness	-5121
+▁mouth	-5122
+▁minimum	-5123
+▁advertising	-5124
+minister	-5125
+sat	-5126
+▁hun	-5127
+▁competitive	-5128
+cir	-5129
+ixture	-5130
+▁square	-5131
+cry	-5132
+ighter	-5133
+eper	-5134
+steve	-5135
+▁celebrate	-5136
+▁tests	-5137
+▁valid	-5138
+country	-5139
+▁visiting	-5140
+bb	-5141
+eline	-5142
+idden	-5143
+▁vary	-5144
+▁married	-5145
+small	-5146
+summer	-5147
+ho	-5148
+nav	-5149
+core	-5150
+▁chief	-5151
+▁veter	-5152
+fox	-5153
+sus	-5154
+▁pros	-5155
+▁apparent	-5156
+▁birthday	-5157
+german	-5158
+ohio	-5159
+otic	-5160
+enjoy	-5161
+▁tail	-5162
+academy	-5163
+japanese	-5164
+▁flowers	-5165
+▁otherwise	-5166
+▁affordable	-5167
+verse	-5168
+▁invent	-5169
+▁struct	-5170
+▁breakfast	-5171
+fill	-5172
+tube	-5173
+▁flo	-5174
+▁cells	-5175
+▁ticket	-5176
+▁bond	-5177
+estyle	-5178
+victor	-5179
+▁species	-5180
+▁accounts	-5181
+wear	-5182
+▁ang	-5183
+rights	-5184
+▁apps	-5185
+abs	-5186
+mosp	-5187
+▁bike	-5188
+rical	-5189
+▁reco	-5190
+pected	-5191
+▁audio	-5192
+▁picked	-5193
+▁delight	-5194
+▁featuring	-5195
+arp	-5196
+iser	-5197
+tered	-5198
+▁scored	-5199
+rate	-5200
+▁boost	-5201
+mult	-5202
+▁deals	-5203
+▁draft	-5204
+▁teeth	-5205
+▁shooting	-5206
+▁truck	-5207
+african	-5208
+);	-5209
+▁vent	-5210
+▁bread	-5211
+station	-5212
+▁animal	-5213
+▁policies	-5214
+daily	-5215
+endar	-5216
+▁honor	-5217
+▁faster	-5218
+fc	-5219
+ora	-5220
+ailability	-5221
+kat	-5222
+▁pepper	-5223
+dar	-5224
+beaut	-5225
+number	-5226
+▁alcohol	-5227
+▁marriage	-5228
+▁possibly	-5229
+▁introduced	-5230
+▁technologies	-5231
+prom	-5232
+ryan	-5233
+▁regularly	-5234
+mobile	-5235
+▁noted	-5236
+▁predict	-5237
+kind	-5238
+range	-5239
+▁battery	-5240
+▁confirm	-5241
+▁illustr	-5242
+▁officer	-5243
+gradu	-5244
+▁bigger	-5245
+ez	-5246
+▁boat	-5247
+iminal	-5248
+▁scale	-5249
+area	-5250
+chang	-5251
+▁spin	-5252
+▁label	-5253
+instead	-5254
+▁administration	-5255
+odies	-5256
+▁weap	-5257
+▁eventually	-5258
+umn	-5259
+▁icon	-5260
+▁cloud	-5261
+▁shoes	-5262
+▁aid	-5263
+▁boys	-5264
+random	-5265
+▁behavior	-5266
+▁remaining	-5267
+tick	-5268
+▁ris	-5269
+▁liqu	-5270
+germany	-5271
+▁arrived	-5272
+▁basket	-5273
+▁recovery	-5274
+▁cro	-5275
+ancer	-5276
+usion	-5277
+▁dollars	-5278
+ali	-5279
+atre	-5280
+▁pit	-5281
+ulate	-5282
+section	-5283
+▁pu	-5284
+abet	-5285
+osen	-5286
+▁evalu	-5287
+federal	-5288
+▁unless	-5289
+▁reliable	-5290
+astructure	-5291
+▁der	-5292
+▁pump	-5293
+lywood	-5294
+▁inches	-5295
+▁neighborhood	-5296
+rier	-5297
+▁fly	-5298
+▁compar	-5299
+▁revealed	-5300
+▁discovered	-5301
+nas	-5302
+georg	-5303
+williams	-5304
+▁consumer	-5305
+wish	-5306
+▁manufacturing	-5307
+kr	-5308
+▁thread	-5309
+▁volume	-5310
+wy	-5311
+morrow	-5312
+profit	-5313
+▁suitable	-5314
+bat	-5315
+▁fif	-5316
+thent	-5317
+training	-5318
+select	-5319
+history	-5320
+▁concent	-5321
+▁symptoms	-5322
+spec	-5323
+▁fro	-5324
+grees	-5325
+▁campus	-5326
+covery	-5327
+▁choices	-5328
+yard	-5329
+▁nurs	-5330
+▁doubt	-5331
+▁flight	-5332
+▁drivers	-5333
+ae	-5334
+bear	-5335
+▁realize	-5336
+idays	-5337
+▁dynam	-5338
+▁reput	-5339
+gov	-5340
+▁appreciate	-5341
+▁mine	-5342
+▁agent	-5343
+▁broken	-5344
+▁incorpor	-5345
+▁painting	-5346
+▁rub	-5347
+friendly	-5348
+▁answers	-5349
+▁cutting	-5350
+words	-5351
+▁blow	-5352
+▁quot	-5353
+▁components	-5354
+vd	-5355
+mean	-5356
+phys	-5357
+▁mel	-5358
+▁elig	-5359
+cking	-5360
+speed	-5361
+▁poll	-5362
+▁planned	-5363
+penn	-5364
+▁dish	-5365
+▁submit	-5366
+marketing	-5367
+▁restaurants	-5368
+oop	-5369
+irms	-5370
+▁election	-5371
+▁appointment	-5372
+▁sessions	-5373
+▁television	-5374
+ura	-5375
+▁corn	-5376
+▁pregn	-5377
+standard	-5378
+stat	-5379
+▁cabin	-5380
+▁compos	-5381
+carolina	-5382
+▁engineering	-5383
+ena	-5384
+rus	-5385
+▁emer	-5386
+▁rule	-5387
+▁protected	-5388
+mom	-5389
+nick	-5390
+supp	-5391
+poses	-5392
+orders	-5393
+italian	-5394
+members	-5395
+natural	-5396
+▁profit	-5397
+has	-5398
+acle	-5399
+iamond	-5400
+▁reserved	-5401
+vit	-5402
+give	-5403
+▁clo	-5404
+▁lived	-5405
+owntown	-5406
+▁earned	-5407
+covid	-5408
+▁load	-5409
+▁films	-5410
+▁twice	-5411
+install	-5412
+▁paying	-5413
+▁warrant	-5414
+▁considering	-5415
+mir	-5416
+▁cultural	-5417
+▁perfectly	-5418
+mos	-5419
+division	-5420
+cath	-5421
+ician	-5422
+still	-5423
+▁dates	-5424
+▁specifically	-5425
+▁athlet	-5426
+▁fields	-5427
+▁applied	-5428
+▁bedroom	-5429
+▁ages	-5430
+▁contest	-5431
+▁agencies	-5432
+▁internal	-5433
+▁strategies	-5434
+▁burn	-5435
+ternal	-5436
+▁dining	-5437
+▁seemed	-5438
+▁victory	-5439
+▁exclusive	-5440
+gage	-5441
+▁covers	-5442
+▁employee	-5443
+▁interior	-5444
+iat	-5445
+adel	-5446
+▁oven	-5447
+▁golf	-5448
+middle	-5449
+▁missed	-5450
+▁sad	-5451
+story	-5452
+▁license	-5453
+▁revenue	-5454
+ibilities	-5455
+poon	-5456
+▁forum	-5457
+▁loves	-5458
+rum	-5459
+▁edition	-5460
+▁housing	-5461
+▁university	-5462
+issue	-5463
+▁lights	-5464
+▁resource	-5465
+shire	-5466
+▁cere	-5467
+▁returns	-5468
+executive	-5469
+easy	-5470
+rance	-5471
+father	-5472
+▁labor	-5473
+▁moist	-5474
+▁automatically	-5475
+▁mir	-5476
+▁keeps	-5477
+▁league	-5478
+▁tomorrow	-5479
+ario	-5480
+farm	-5481
+tle	-5482
+base	-5483
+▁oppos	-5484
+▁confirmed	-5485
+stream	-5486
+▁thus	-5487
+▁proposed	-5488
+▁quant	-5489
+▁desire	-5490
+fs	-5491
+stra	-5492
+▁depos	-5493
+▁grade	-5494
+▁messages	-5495
+▁commission	-5496
+▁incredible	-5497
+▁die	-5498
+▁lit	-5499
+▁yes	-5500
+ospit	-5501
+▁issued	-5502
+▁ign	-5503
+ships	-5504
+proper	-5505
+▁prison	-5506
+▁aspects	-5507
+▁extract	-5508
+▁attorney	-5509
+andy	-5510
+mach	-5511
+▁surf	-5512
+▁chosen	-5513
+▁practical	-5514
+md	-5515
+▁debut	-5516
+ication	-5517
+richard	-5518
+▁supposed	-5519
+fection	-5520
+mt	-5521
+ano	-5522
+▁ben	-5523
+▁vit	-5524
+▁turns	-5525
+▁buildings	-5526
+▁surrounding	-5527
+paris	-5528
+▁cute	-5529
+▁ends	-5530
+account	-5531
+▁remote	-5532
+▁stretch	-5533
+▁purposes	-5534
+osh	-5535
+rep	-5536
+umes	-5537
+▁anc	-5538
+mexico	-5539
+▁shares	-5540
+ria	-5541
+andrew	-5542
+jackson	-5543
+▁injuries	-5544
+▁actions	-5545
+▁improved	-5546
+kay	-5547
+gment	-5548
+▁laws	-5549
+▁chain	-5550
+iest	-5551
+▁cake	-5552
+▁error	-5553
+xy	-5554
+▁spokes	-5555
+▁styles	-5556
+▁musical	-5557
+▁accurate	-5558
+▁standing	-5559
+kim	-5560
+▁wid	-5561
+▁tells	-5562
+▁circum	-5563
+▁negative	-5564
+▁industrial	-5565
+tax	-5566
+▁pod	-5567
+▁affected	-5568
+rac	-5569
+conf	-5570
+▁dra	-5571
+▁hang	-5572
+▁wheel	-5573
+instagram	-5574
+▁instructions	-5575
+mind	-5576
+▁pun	-5577
+▁apartment	-5578
+▁favourite	-5579
+▁registration	-5580
+▁cop	-5581
+should	-5582
+▁typically	-5583
+bad	-5584
+event	-5585
+flict	-5586
+▁pure	-5587
+comput	-5588
+▁speech	-5589
+▁valuable	-5590
+eas	-5591
+isa	-5592
+orable	-5593
+▁island	-5594
+▁moments	-5595
+▁atmosp	-5596
+ltd	-5597
+save	-5598
+▁rapid	-5599
+erce	-5600
+▁foods	-5601
+▁researchers	-5602
+iability	-5603
+▁replace	-5604
+fac	-5605
+dist	-5606
+daniel	-5607
+isters	-5608
+▁holds	-5609
+▁county	-5610
+▁failed	-5611
+cling	-5612
+pacific	-5613
+▁meetings	-5614
+.|	-5615
+▁une	-5616
+▁route	-5617
+▁presentation	-5618
+olymp	-5619
+michigan	-5620
+▁decades	-5621
+tes	-5622
+▁pleased	-5623
+▁greatest	-5624
+eler	-5625
+▁poly	-5626
+having	-5627
+ato	-5628
+▁debt	-5629
+▁usual	-5630
+▁influence	-5631
+glass	-5632
+track	-5633
+intage	-5634
+▁logo	-5635
+▁whom	-5636
+▁agreed	-5637
+▁stated	-5638
+husi	-5639
+iors	-5640
+▁inn	-5641
+donald	-5642
+▁element	-5643
+▁bound	-5644
+▁enable	-5645
+▁degrees	-5646
+▁recorded	-5647
+▁awareness	-5648
+due	-5649
+stop	-5650
+tele	-5651
+▁dial	-5652
+orough	-5653
+tay	-5654
+oosing	-5655
+nc	-5656
+pes	-5657
+▁rot	-5658
+sher	-5659
+stay	-5660
+zeal	-5661
+▁perman	-5662
+▁causes	-5663
+▁everyday	-5664
+▁processing	-5665
+member	-5666
+▁photograp	-5667
+image	-5668
+▁hero	-5669
+▁reference	-5670
+ache	-5671
+hour	-5672
+santa	-5673
+safety	-5674
+▁owned	-5675
+▁leather	-5676
+▁arm	-5677
+▁hall	-5678
+beat	-5679
+▁plate	-5680
+perfect	-5681
+▁concerned	-5682
+▁jack	-5683
+▁mini	-5684
+▁tack	-5685
+living	-5686
+▁emotional	-5687
+requ	-5688
+▁rough	-5689
+▁elimin	-5690
+oval	-5691
+cular	-5692
+▁trend	-5693
+model	-5694
+plate	-5695
+ocking	-5696
+vernor	-5697
+▁grown	-5698
+ulations	-5699
+▁officers	-5700
+▁supports	-5701
+adium	-5702
+canadian	-5703
+smart	-5704
+▁grad	-5705
+▁recy	-5706
+giving	-5707
+▁powder	-5708
+ucle	-5709
+▁newsp	-5710
+▁ongoing	-5711
+pod	-5712
+heat	-5713
+iano	-5714
+again	-5715
+▁adventure	-5716
+mbly	-5717
+▁sam	-5718
+itage	-5719
+sembly	-5720
+▁browser	-5721
+welry	-5722
+▁height	-5723
+▁designer	-5724
+fficient	-5725
+▁stopped	-5726
+orgeous	-5727
+virginia	-5728
+pon	-5729
+swe	-5730
+▁crazy	-5731
+despite	-5732
+▁premium	-5733
+!”	-5734
+cab	-5735
+▁rear	-5736
+jay	-5737
+▁phil	-5738
+subscri	-5739
+personal	-5740
+▁hardware	-5741
+scot	-5742
+oration	-5743
+▁organic	-5744
+▁bathroom	-5745
+▁database	-5746
+▁disappoint	-5747
+awards	-5748
+▁village	-5749
+ipl	-5750
+abled	-5751
+▁mode	-5752
+bag	-5753
+irty	-5754
+▁gro	-5755
+▁tro	-5756
+▁supplies	-5757
+▁functions	-5758
+▁effectively	-5759
+▁gear	-5760
+credit	-5761
+▁crowd	-5762
+▁calcul	-5763
+▁surprise	-5764
+broad	-5765
+▁knowing	-5766
+teen	-5767
+▁crime	-5768
+continue	-5769
+▁creation	-5770
+▁tradition	-5771
+▁candidates	-5772
+▁inspiration	-5773
+dun	-5774
+draw	-5775
+▁tut	-5776
+▁beer	-5777
+▁licens	-5778
+mountain	-5779
+▁intellig	-5780
+▁happening	-5781
+environment	-5782
+inal	-5783
+arks	-5784
+▁gam	-5785
+radio	-5786
+spanish	-5787
+▁distribution	-5788
+lan	-5789
+▁aw	-5790
+▁funny	-5791
+▁receiving	-5792
+▁electronic	-5793
+▁infrastructure	-5794
+upd	-5795
+▁mixed	-5796
+▁script	-5797
+▁improvement	-5798
+▁accepted	-5799
+▁employment	-5800
+▁identified	-5801
+fish	-5802
+▁tag	-5803
+▁talks	-5804
+▁weekly	-5805
+learning	-5806
+financial	-5807
+▁facing	-5808
+▁mixture	-5809
+▁providers	-5810
+pict	-5811
+advis	-5812
+clear	-5813
+▁belong	-5814
+▁treated	-5815
+francisco	-5816
+published	-5817
+▁violence	-5818
+plant	-5819
+▁harm	-5820
+▁serves	-5821
+vy	-5822
+mally	-5823
+polit	-5824
+▁advoc	-5825
+russian	-5826
+phas	-5827
+colorado	-5828
+shipping	-5829
+?"	-5830
+iated	-5831
+▁disp	-5832
+▁payments	-5833
+▁administr	-5834
+▁healthcare	-5835
+▁fixed	-5836
+▁flour	-5837
+▁silver	-5838
+clusion	-5839
+ilit	-5840
+▁sky	-5841
+▁sheet	-5842
+▁scholar	-5843
+▁losing	-5844
+▁taught	-5845
+▁lessons	-5846
+lay	-5847
+▁forces	-5848
+▁freedom	-5849
+▁extended	-5850
+▁determined	-5851
+▁horse	-5852
+iveness	-5853
+▁forest	-5854
+var	-5855
+erves	-5856
+senior	-5857
+▁worse	-5858
+▁stands	-5859
+▁trouble	-5860
+▁incident	-5861
+▁accessible	-5862
+▁ped	-5863
+▁teas	-5864
+▁arts	-5865
+▁intr	-5866
+▁closer	-5867
+crete	-5868
+graduate	-5869
+cra	-5870
+ele	-5871
+mov	-5872
+brid	-5873
+gate	-5874
+jersey	-5875
+▁decre	-5876
+cow	-5877
+▁symb	-5878
+▁kinds	-5879
+▁hoping	-5880
+▁successfully	-5881
+pen	-5882
+▁dol	-5883
+amber	-5884
+▁imagine	-5885
+▁mal	-5886
+▁coordin	-5887
+▁follows	-5888
+ref	-5889
+▁voc	-5890
+▁yellow	-5891
+▁alleged	-5892
+▁witness	-5893
+▁upper	-5894
+georgia	-5895
+▁papers	-5896
+▁patterns	-5897
+▁vacation	-5898
+▁enthusi	-5899
+▁matters	-5900
+webs	-5901
+etime	-5902
+▁telling	-5903
+▁significantly	-5904
+rors	-5905
+screen	-5906
+▁turning	-5907
+▁lift	-5908
+▁native	-5909
+▁therapy	-5910
+▁convenient	-5911
+▁efficiency	-5912
+▁impressive	-5913
+cro	-5914
+▁bat	-5915
+▁trading	-5916
+hp	-5917
+pool	-5918
+▁dram	-5919
+▁baking	-5920
+brow	-5921
+would	-5922
+kingdom	-5923
+▁stunning	-5924
+rim	-5925
+▁advis	-5926
+▁folks	-5927
+▁recipes	-5928
+phy	-5929
+create	-5930
+▁drugs	-5931
+▁guitar	-5932
+forcement	-5933
+▁conducted	-5934
+▁lifestyle	-5935
+dur	-5936
+rev	-5937
+▁pil	-5938
+▁male	-5939
+ansion	-5940
+▁recognized	-5941
+total	-5942
+large	-5943
+▁grab	-5944
+▁simpl	-5945
+finally	-5946
+▁clothes	-5947
+▁explained	-5948
+▁description	-5949
+▁expectations	-5950
+cop	-5951
+psy	-5952
+cers	-5953
+user	-5954
+▁spr	-5955
+ireland	-5956
+▁crisis	-5957
+▁dim	-5958
+▁entreprene	-5959
+illy	-5960
+orrow	-5961
+▁bless	-5962
+▁luxury	-5963
+inations	-5964
+wan	-5965
+corn	-5966
+pick	-5967
+software	-5968
+▁routine	-5969
+▁dangerous	-5970
+▁tall	-5971
+southern	-5972
+▁investors	-5973
+ansas	-5974
+▁houses	-5975
+▁trends	-5976
+▁chapter	-5977
+▁reduced	-5978
+▁procedures	-5979
+ava	-5980
+▁cock	-5981
+▁cable	-5982
+▁goods	-5983
+▁severe	-5984
+▁hits	-5985
+▁sequ	-5986
+ji	-5987
+tal	-5988
+▁channel	-5989
+▁tonight	-5990
+▁recording	-5991
+rated	-5992
+disney	-5993
+▁manner	-5994
+▁settings	-5995
+professional	-5996
+▁photography	-5997
+▁collaboration	-5998
+gun	-5999
+sales	-6000
+▁thin	-6001
+▁random	-6002
+▁context	-6003
+▁virtual	-6004
+▁conflict	-6005
+▁contemporary	-6006
+▁config	-6007
+▁forced	-6008
+iversary	-6009
+▁reporting	-6010
+▁il	-6011
+film	-6012
+▁accessories	-6013
+▁cups	-6014
+▁laugh	-6015
+▁challenging	-6016
+luc	-6017
+cuss	-6018
+italy	-6019
+▁vend	-6020
+▁gifts	-6021
+student	-6022
+▁council	-6023
+▁hosting	-6024
+▁graduate	-6025
+▁likes	-6026
+▁noticed	-6027
+▁authent	-6028
+▁attended	-6029
+▁clinical	-6030
+▁regional	-6031
+▁constantly	-6032
+▁perspective	-6033
+athe	-6034
+avenue	-6035
+▁printed	-6036
+folio	-6037
+▁adds	-6038
+think	-6039
+▁tiny	-6040
+baby	-6041
+nger	-6042
+▁depth	-6043
+▁excess	-6044
+mix	-6045
+royal	-6046
+▁govern	-6047
+must	-6048
+forest	-6049
+▁renew	-6050
+▁clothing	-6051
+▁foundation	-6052
+anne	-6053
+kids	-6054
+quir	-6055
+rition	-6056
+further	-6057
+writing	-6058
+▁involve	-6059
+▁domestic	-6060
+las	-6061
+▁:)	-6062
+lete	-6063
+utch	-6064
+express	-6065
+▁flavor	-6066
+▁retire	-6067
+▁univers	-6068
+zil	-6069
+amps	-6070
+iami	-6071
+▁hur	-6072
+▁stone	-6073
+ultural	-6074
+▁prompt	-6075
+▁innovation	-6076
+▁diverse	-6077
+▁searching	-6078
+▁bags	-6079
+ossible	-6080
+jam	-6081
+gard	-6082
+▁shots	-6083
+material	-6084
+▁circumst	-6085
+asts	-6086
+▁glad	-6087
+▁guard	-6088
+▁walls	-6089
+▁massive	-6090
+▁outstanding	-6091
+▁sustainable	-6092
+▁tested	-6093
+▁positions	-6094
+▁replacement	-6095
+lon	-6096
+such	-6097
+▁wet	-6098
+garden	-6099
+zealand	-6100
+five	-6101
+treat	-6102
+▁taxes	-6103
+▁carried	-6104
+rub	-6105
+ilst	-6106
+▁rose	-6107
+▁speaking	-6108
+gro	-6109
+amic	-6110
+enger	-6111
+▁candidate	-6112
+sle	-6113
+▁tou	-6114
+miami	-6115
+▁arrested	-6116
+▁millions	-6117
+▁construct	-6118
+▁resolution	-6119
+arab	-6120
+enced	-6121
+▁uplo	-6122
+▁examples	-6123
+dim	-6124
+unfortunately	-6125
+sav	-6126
+ugg	-6127
+▁ap	-6128
+eric	-6129
+▁jun	-6130
+▁wash	-6131
+update	-6132
+respond	-6133
+remember	-6134
+▁medicine	-6135
+jr	-6136
+▁garage	-6137
+▁figures	-6138
+▁carefully	-6139
+▁surprised	-6140
+▁integrated	-6141
+nee	-6142
+▁abuse	-6143
+▁smile	-6144
+▁assets	-6145
+▁programme	-6146
+stal	-6147
+viet	-6148
+ament	-6149
+▁deploy	-6150
+▁unable	-6151
+▁procedure	-6152
+cuit	-6153
+grow	-6154
+▁reward	-6155
+▁versions	-6156
+▁interests	-6157
+ampion	-6158
+▁cycle	-6159
+▁flood	-6160
+▁nearby	-6161
+▁pounds	-6162
+▁failure	-6163
+▁festival	-6164
+▁committee	-6165
+▁authorities	-6166
+irts	-6167
+asian	-6168
+location	-6169
+▁command	-6170
+▁availability	-6171
+dol	-6172
+onym	-6173
+raid	-6174
+comfort	-6175
+sd	-6176
+▁?	-6177
+sit	-6178
+▁tap	-6179
+aints	-6180
+short	-6181
+▁grew	-6182
+▁muse	-6183
+systems	-6184
+▁emphas	-6185
+▁residential	-6186
+pg	-6187
+clin	-6188
+oral	-6189
+iples	-6190
+▁iron	-6191
+▁prot	-6192
+▁criminal	-6193
+▁eleg	-6194
+russia	-6195
+▁bottle	-6196
+▁herself	-6197
+▁entirely	-6198
+▁seriously	-6199
+▁lad	-6200
+choose	-6201
+taylor	-6202
+feed	-6203
+talk	-6204
+▁fitness	-6205
+▁offices	-6206
+recent	-6207
+▁gorgeous	-6208
+▁anx	-6209
+▁pink	-6210
+▁tomat	-6211
+▁unlike	-6212
+▁household	-6213
+syl	-6214
+iana	-6215
+▁ordered	-6216
+▁scientific	-6217
+everyone	-6218
+▁creates	-6219
+▁dropped	-6220
+▁mountain	-6221
+dden	-6222
+▁bab	-6223
+▁ham	-6224
+▁tech	-6225
+common	-6226
+▁gre	-6227
+▁factor	-6228
+▁garlic	-6229
+▁eligible	-6230
+reek	-6231
+brook	-6232
+eries	-6233
+certain	-6234
+process	-6235
+▁distinct	-6236
+file	-6237
+▁acid	-6238
+official	-6239
+uro	-6240
+wik	-6241
+bury	-6242
+edge	-6243
+▁west	-6244
+▁layer	-6245
+▁occup	-6246
+▁concern	-6247
+yet	-6248
+▁weak	-6249
+photos	-6250
+▁awards	-6251
+▁detect	-6252
+▁trained	-6253
+▁thorough	-6254
+html	-6255
+▁shower	-6256
+▁atmosphere	-6257
+▁transportation	-6258
+fly	-6259
+zes	-6260
+▁cm	-6261
+wich	-6262
+proof	-6263
+▁risks	-6264
+▁spaces	-6265
+▁lighting	-6266
+eve	-6267
+▁eggs	-6268
+▁shift	-6269
+▁sick	-6270
+▁guidance	-6271
+▁historical	-6272
+adam	-6273
+▁quote	-6274
+▁wealth	-6275
+▁improving	-6276
+▁prop	-6277
+▁equal	-6278
+charles	-6279
+justice	-6280
+▁qualified	-6281
+▁transition	-6282
+vo	-6283
+edia	-6284
+anton	-6285
+▁)	-6286
+phones	-6287
+rip	-6288
+rug	-6289
+insp	-6290
+▁ease	-6291
+▁liquid	-6292
+▁rating	-6293
+▁capture	-6294
+professor	-6295
+kent	-6296
+▁san	-6297
+cloud	-6298
+▁none	-6299
+▁citizens	-6300
+▁identity	-6301
+▁contribute	-6302
+cue	-6303
+exp	-6304
+safe	-6305
+comment	-6306
+▁invited	-6307
+▁certified	-6308
+emic	-6309
+chall	-6310
+estate	-6311
+▁negot	-6312
+iti	-6313
+race	-6314
+▁skill	-6315
+▁hospit	-6316
+▁faculty	-6317
+tam	-6318
+porter	-6319
+return	-6320
+presents	-6321
+engineering	-6322
+▁originally	-6323
+hl	-6324
+llc	-6325
+mrs	-6326
+▁believes	-6327
+dead	-6328
+stan	-6329
+inson	-6330
+▁soul	-6331
+ificate	-6332
+▁tournament	-6333
+tel	-6334
+dale	-6335
+▁filed	-6336
+▁tasks	-6337
+▁joining	-6338
+▁singles	-6339
+’.	-6340
+sured	-6341
+▁legs	-6342
+▁judge	-6343
+▁fishing	-6344
+▁developers	-6345
+erts	-6346
+▁proven	-6347
+experience	-6348
+vas	-6349
+pers	-6350
+▁wra	-6351
+▁blend	-6352
+▁solar	-6353
+▁yards	-6354
+▁sudden	-6355
+▁termin	-6356
+▁listing	-6357
+athan	-6358
+jured	-6359
+vs	-6360
+eria	-6361
+sung	-6362
+▁rank	-6363
+axy	-6364
+adow	-6365
+queen	-6366
+▁empt	-6367
+▁entered	-6368
+▁mm	-6369
+onto	-6370
+▁susp	-6371
+▁tear	-6372
+player	-6373
+tell	-6374
+▁tip	-6375
+iously	-6376
+▁juice	-6377
+▁worst	-6378
+!"	-6379
+pn	-6380
+cou	-6381
+▁sole	-6382
+▁lucky	-6383
+release	-6384
+▁adapt	-6385
+▁broadcast	-6386
+▁increases	-6387
+bol	-6388
+▁||	-6389
+bass	-6390
+brad	-6391
+▁mut	-6392
+silver	-6393
+complete	-6394
+▁founded	-6395
+▁calendar	-6396
+▁networks	-6397
+izza	-6398
+agues	-6399
+brazil	-6400
+senate	-6401
+studies	-6402
+▁interface	-6403
+inar	-6404
+gallery	-6405
+ariz	-6406
+send	-6407
+▁frag	-6408
+igration	-6409
+ographic	-6410
+▁interpre	-6411
+autom	-6412
+▁faces	-6413
+▁cooper	-6414
+▁engagement	-6415
+motor	-6416
+▁spots	-6417
+▁conver	-6418
+▁protein	-6419
+▁younger	-6420
+aze	-6421
+alty	-6422
+▁sac	-6423
+▁firms	-6424
+▁pitch	-6425
+▁cotton	-6426
+omy	-6427
+▁rac	-6428
+▁hire	-6429
+icial	-6430
+▁arms	-6431
+▁root	-6432
+apping	-6433
+▁organized	-6434
+▁submitted	-6435
+aron	-6436
+ooks	-6437
+camer	-6438
+title	-6439
+course	-6440
+▁chose	-6441
+▁resol	-6442
+illiant	-6443
+reading	-6444
+▁refund	-6445
+▁counsel	-6446
+mers	-6447
+▁fle	-6448
+ership	-6449
+policy	-6450
+▁sample	-6451
+▁objects	-6452
+rot	-6453
+▁carbon	-6454
+commerce	-6455
+▁streets	-6456
+▁fighting	-6457
+republican	-6458
+▁capabilities	-6459
+ikes	-6460
+sweet	-6461
+▁zone	-6462
+▁aside	-6463
+▁appeal	-6464
+▁dental	-6465
+aturally	-6466
+▁awarded	-6467
+hur	-6468
+uling	-6469
+▁flag	-6470
+▁reputation	-6471
+▁implementation	-6472
+hire	-6473
+▁forever	-6474
+▁letters	-6475
+▁posting	-6476
+▁concert	-6477
+▁membership	-6478
+hous	-6479
+▁cos	-6480
+▁vul	-6481
+cycle	-6482
+▁vital	-6483
+hey	-6484
+iran	-6485
+▁moves	-6486
+seattle	-6487
+▁instance	-6488
+▁volunteers	-6489
+ku	-6490
+mess	-6491
+sound	-6492
+▁hole	-6493
+▁seed	-6494
+consult	-6495
+ensation	-6496
+together	-6497
+▁consequ	-6498
+▁drawing	-6499
+▁exhibition	-6500
+▁institutions	-6501
+sche	-6502
+uine	-6503
+▁hasn	-6504
+▁exposure	-6505
+▁strategic	-6506
+▁monitoring	-6507
+uster	-6508
+▁bonus	-6509
+catholic	-6510
+following	-6511
+▁downtown	-6512
+▁enjoying	-6513
+▁situations	-6514
+ala	-6515
+bible	-6516
+thetic	-6517
+▁spoke	-6518
+▁hosted	-6519
+▁output	-6520
+▁listening	-6521
+▁auto	-6522
+without	-6523
+▁tracks	-6524
+▁airport	-6525
+jord	-6526
+ennis	-6527
+forum	-6528
+northern	-6529
+▁assessment	-6530
+cool	-6531
+▁ult	-6532
+ateful	-6533
+▁sending	-6534
+▁choosing	-6535
+htt	-6536
+nda	-6537
+yers	-6538
+▁coc	-6539
+▁pulled	-6540
+customer	-6541
+▁privacy	-6542
+▁guidelines	-6543
+gary	-6544
+▁obl	-6545
+oured	-6546
+annels	-6547
+▁hopes	-6548
+▁ultimate	-6549
+oak	-6550
+ults	-6551
+▁attached	-6552
+aqu	-6553
+pin	-6554
+tail	-6555
+▁sight	-6556
+▁frequently	-6557
+lif	-6558
+say	-6559
+izer	-6560
+final	-6561
+▁flexible	-6562
+▁platforms	-6563
+▁suggested	-6564
+usb	-6565
+birth	-6566
+▁linked	-6567
+▁motion	-6568
+▁pleasure	-6569
+had	-6570
+bang	-6571
+▁vibr	-6572
+▁suffe	-6573
+▁causing	-6574
+▁workshop	-6575
+ucky	-6576
+wire	-6577
+▁hat	-6578
+▁performing	-6579
+▁regulations	-6580
+via	-6581
+sale	-6582
+dream	-6583
+▁delay	-6584
+atlanta	-6585
+therapy	-6586
+▁bodies	-6587
+▁closely	-6588
+bound	-6589
+square	-6590
+▁phase	-6591
+edition	-6592
+▁engage	-6593
+▁circumstances	-6594
+studio	-6595
+things	-6596
+▁input	-6597
+capital	-6598
+▁destroy	-6599
+hollywood	-6600
+▁destination	-6601
+fold	-6602
+▁inch	-6603
+▁warn	-6604
+version	-6605
+▁matches	-6606
+▁portion	-6607
+▁visited	-6608
+cit	-6609
+sole	-6610
+chand	-6611
+▁dreams	-6612
+▁reader	-6613
+▁operate	-6614
+▁informed	-6615
+dark	-6616
+inde	-6617
+▁bench	-6618
+▁trick	-6619
+▁terror	-6620
+including	-6621
+oz	-6622
+idi	-6623
+liam	-6624
+ially	-6625
+▁crew	-6626
+makers	-6627
+▁urban	-6628
+▁injured	-6629
+▁directed	-6630
+▁relatively	-6631
+allow	-6632
+piece	-6633
+nation	-6634
+server	-6635
+spirit	-6636
+▁meets	-6637
+▁massage	-6638
+▁copyright	-6639
+▁spiritual	-6640
+comb	-6641
+prime	-6642
+quick	-6643
+▁rice	-6644
+▁liked	-6645
+houston	-6646
+▁loving	-6647
+oft	-6648
+pak	-6649
+▁defend	-6650
+▁saving	-6651
+fat	-6652
+bird	-6653
+umps	-6654
+▁river	-6655
+isation	-6656
+▁manual	-6657
+▁promise	-6658
+insurance	-6659
+▁landscape	-6660
+.’	-6661
+esy	-6662
+burgh	-6663
+perhaps	-6664
+▁anticip	-6665
+▁historic	-6666
+.'	-6667
+rh	-6668
+neys	-6669
+osop	-6670
+▁bird	-6671
+▁lege	-6672
+▁upgrade	-6673
+▁memories	-6674
+▁encouraged	-6675
+▁treatments	-6676
+cru	-6677
+doors	-6678
+oster	-6679
+joseph	-6680
+winter	-6681
+▁filter	-6682
+past	-6683
+tics	-6684
+ounce	-6685
+▁jewelry	-6686
+▁holidays	-6687
+kevin	-6688
+▁hurt	-6689
+friends	-6690
+▁critic	-6691
+auto	-6692
+annual	-6693
+▁finds	-6694
+▁planet	-6695
+▁technique	-6696
+▁manufacturer	-6697
+ien	-6698
+brian	-6699
+▁butt	-6700
+▁tank	-6701
+uclear	-6702
+medicine	-6703
+▁gallery	-6704
+him	-6705
+obby	-6706
+▁tum	-6707
+charge	-6708
+▁boxes	-6709
+holders	-6710
+▁scheme	-6711
+▁authors	-6712
+▁capable	-6713
+championship	-6714
+▁publication	-6715
+▁km	-6716
+▁domin	-6717
+▁tooth	-6718
+era	-6719
+six	-6720
+army	-6721
+fred	-6722
+sport	-6723
+▁soil	-6724
+▁grass	-6725
+▁managing	-6726
+dall	-6727
+prim	-6728
+preme	-6729
+effect	-6730
+▁talked	-6731
+▁justice	-6732
+▁typical	-6733
+▁victims	-6734
+collection	-6735
+performance	-6736
+hay	-6737
+pir	-6738
+bath	-6739
+path	-6740
+tony	-6741
+asted	-6742
+▁attacks	-6743
+secretary	-6744
+ani	-6745
+lers	-6746
+▁flash	-6747
+▁smell	-6748
+iii	-6749
+spot	-6750
+▁rail	-6751
+minnes	-6752
+faction	-6753
+▁alongside	-6754
+▁resort	-6755
+aid	-6756
+trad	-6757
+▁subs	-6758
+▁spark	-6759
+▁rental	-6760
+products	-6761
+▁volunteer	-6762
+pi	-6763
+money	-6764
+▁savings	-6765
+▁suggestions	-6766
+▁($	-6767
+ella	-6768
+▁magic	-6769
+▁split	-6770
+samsung	-6771
+▁arrive	-6772
+▁debate	-6773
+▁wireless	-6774
+▁lawyer	-6775
+▁riding	-6776
+▁drinking	-6777
+amel	-6778
+below	-6779
+▁hate	-6780
+▁proof	-6781
+▁proceed	-6782
+ingly	-6783
+details	-6784
+▁blocks	-6785
+▁discipl	-6786
+izz	-6787
+▁cin	-6788
+▁stake	-6789
+▁relief	-6790
+dvd	-6791
+▁fairly	-6792
+▁seasons	-6793
+igs	-6794
+▁row	-6795
+▁depress	-6796
+▁consistent	-6797
+▁incredibly	-6798
+cha	-6799
+chi	-6800
+rible	-6801
+▁stir	-6802
+classic	-6803
+▁domain	-6804
+▁passing	-6805
+▁collected	-6806
+▁continuing	-6807
+ule	-6808
+▁mas	-6809
+▁cuts	-6810
+▁gent	-6811
+rovers	-6812
+private	-6813
+minnesota	-6814
+▁estimated	-6815
+▁initiative	-6816
+bru	-6817
+urb	-6818
+feel	-6819
+▁fas	-6820
+ingers	-6821
+▁prove	-6822
+▁anyway	-6823
+▁pocket	-6824
+▁password	-6825
+▁apparently	-6826
+dor	-6827
+hops	-6828
+agency	-6829
+▁sharp	-6830
+▁shops	-6831
+channel	-6832
+▁phones	-6833
+▁walked	-6834
+▁priority	-6835
+▁celebration	-6836
+▁tape	-6837
+columb	-6838
+histor	-6839
+▁newly	-6840
+▁shock	-6841
+▁chances	-6842
+bn	-6843
+davis	-6844
+golden	-6845
+▁reform	-6846
+▁approval	-6847
+▁equipped	-6848
+▁machines	-6849
+▁immediate	-6850
+▁solve	-6851
+ainless	-6852
+▁illness	-6853
+meanwhile	-6854
+▁external	-6855
+▁discussed	-6856
+nfl	-6857
+linux	-6858
+▁adopt	-6859
+▁storm	-6860
+eastern	-6861
+▁border	-6862
+magazine	-6863
+▁somewhere	-6864
+▁represents	-6865
+metro	-6866
+adelph	-6867
+▁crack	-6868
+▁forth	-6869
+▁mainly	-6870
+▁writers	-6871
+▁believed	-6872
+▁normally	-6873
+golf	-6874
+iron	-6875
+▁exh	-6876
+cancer	-6877
+▁belief	-6878
+▁classroom	-6879
+▁permission	-6880
+etic	-6881
+bring	-6882
+urning	-6883
+▁casino	-6884
+▁murder	-6885
+website	-6886
+▁muscle	-6887
+▁hook	-6888
+▁frustr	-6889
+▁vulner	-6890
+▁legislation	-6891
+ext	-6892
+years	-6893
+▁hell	-6894
+missions	-6895
+therefore	-6896
+song	-6897
+steel	-6898
+▁decade	-6899
+description	-6900
+nie	-6901
+▁scr	-6902
+▁colle	-6903
+▁instrument	-6904
+editor	-6905
+▁prize	-6906
+▁venue	-6907
+winning	-6908
+▁infect	-6909
+memorial	-6910
+▁covering	-6911
+▁realized	-6912
+)|	-6913
+ema	-6914
+▁stim	-6915
+▁recover	-6916
+▁clicking	-6917
+▁swimming	-6918
+▁industries	-6919
+irish	-6920
+stery	-6921
+▁banks	-6922
+counter	-6923
+▁casual	-6924
+partners	-6925
+▁chopped	-6926
+▁finance	-6927
+itionally	-6928
+▁vegetables	-6929
+▁differences	-6930
+dra	-6931
+icit	-6932
+ogue	-6933
+among	-6934
+apped	-6935
+consum	-6936
+valent	-6937
+▁feelings	-6938
+▁possibility	-6939
+mah	-6940
+shot	-6941
+▁tie	-6942
+▁east	-6943
+jewish	-6944
+▁meals	-6945
+finance	-6946
+▁ending	-6947
+▁dynamic	-6948
+▁breaking	-6949
+▁categories	-6950
+tar	-6951
+vac	-6952
+bits	-6953
+hero	-6954
+stor	-6955
+▁vacc	-6956
+▁facts	-6957
+▁shell	-6958
+content	-6959
+▁laptop	-6960
+▁permanent	-6961
+administration	-6962
+▁tab	-6963
+cript	-6964
+double	-6965
+▁theory	-6966
+adelphia	-6967
+▁shouldn	-6968
+▁bes	-6969
+ashes	-6970
+outer	-6971
+▁vari	-6972
+arizona	-6973
+written	-6974
+▁division	-6975
+▁handling	-6976
+cort	-6977
+▁dust	-6978
+▁agents	-6979
+▁replaced	-6980
+▁requests	-6981
+si	-6982
+▁fra	-6983
+▁raw	-6984
+▁ancient	-6985
+buff	-6986
+dave	-6987
+spain	-6988
+▁bear	-6989
+▁enables	-6990
+▁facilit	-6991
+▁portfolio	-6992
+ras	-6993
+vard	-6994
+▁upgr	-6995
+▁broke	-6996
+officer	-6997
+▁colours	-6998
+▁filling	-6999
+philadelphia	-7000
+▁mile	-7001
+▁mood	-7002
+▁index	-7003
+ione	-7004
+ridge	-7005
+▁elev	-7006
+▁breath	-7007
+▁formed	-7008
+▁anniversary	-7009
+▁dent	-7010
+compet	-7011
+stephen	-7012
+▁drinks	-7013
+▁headed	-7014
+everything	-7015
+▁religious	-7016
+▁attractive	-7017
+fix	-7018
+whel	-7019
+asant	-7020
+▁lets	-7021
+▁magn	-7022
+▁fifth	-7023
+collect	-7024
+▁buyers	-7025
+▁slowly	-7026
+▁printing	-7027
+▁brilliant	-7028
+▁fo	-7029
+vice	-7030
+▁sec	-7031
+jason	-7032
+arters	-7033
+▁giant	-7034
+▁enroll	-7035
+▁titles	-7036
+▁charity	-7037
+▁suffered	-7038
+ulpt	-7039
+shirt	-7040
+▁yard	-7041
+▁bunch	-7042
+▁recall	-7043
+▁surprising	-7044
+▁architecture	-7045
+fab	-7046
+▁bare	-7047
+ensity	-7048
+▁mortgage	-7049
+gypt	-7050
+▁adop	-7051
+▁coat	-7052
+miller	-7053
+▁hidden	-7054
+drop	-7055
+▁god	-7056
+woman	-7057
+▁fest	-7058
+single	-7059
+▁recognize	-7060
+▁traveling	-7061
+▁ast	-7062
+▁birds	-7063
+▁describe	-7064
+chel	-7065
+block	-7066
+vania	-7067
+▁noise	-7068
+▁aspect	-7069
+creative	-7070
+delivery	-7071
+subscribe	-7072
+▁accomplish	-7073
+fav	-7074
+▁adm	-7075
+▁pra	-7076
+prior	-7077
+galaxy	-7078
+register	-7079
+▁genuine	-7080
+solutions	-7081
+▁generate	-7082
+▁personally	-7083
+▁performances	-7084
+agu	-7085
+lad	-7086
+tree	-7087
+▁aims	-7088
+closed	-7089
+reviews	-7090
+▁doctors	-7091
+▁ec	-7092
+alse	-7093
+inton	-7094
+dallas	-7095
+▁roles	-7096
+▁packages	-7097
+half	-7098
+▁pin	-7099
+igate	-7100
+trail	-7101
+around	-7102
+▁empty	-7103
+▁compat	-7104
+▁defined	-7105
+construction	-7106
+▁recommendations	-7107
+%.	-7108
+owa	-7109
+roman	-7110
+staff	-7111
+trade	-7112
+▁loans	-7113
+amy	-7114
+▁acts	-7115
+toronto	-7116
+transport	-7117
+aren	-7118
+inos	-7119
+swed	-7120
+▁scan	-7121
+▁moder	-7122
+represent	-7123
+production	-7124
+dj	-7125
+dry	-7126
+jenn	-7127
+mini	-7128
+diego	-7129
+▁graphics	-7130
+hz	-7131
+▁bars	-7132
+ington	-7133
+▁habit	-7134
+patrick	-7135
+computer	-7136
+▁guaranteed	-7137
+sec	-7138
+▁evol	-7139
+▁gard	-7140
+▁orange	-7141
+▁conclud	-7142
+communications	-7143
+▁bid	-7144
+abeth	-7145
+apore	-7146
+otted	-7147
+poons	-7148
+wilson	-7149
+▁absor	-7150
+▁scenes	-7151
+▁assistant	-7152
+cart	-7153
+fresh	-7154
+▁blank	-7155
+▁recre	-7156
+▁acknow	-7157
+▁marked	-7158
+▁recognition	-7159
+ulf	-7160
+.....	-7161
+egypt	-7162
+orange	-7163
+▁heads	-7164
+▁inner	-7165
+columbia	-7166
+▁falling	-7167
+▁interactive	-7168
+suc	-7169
+▁et	-7170
+moon	-7171
+grant	-7172
+▁gall	-7173
+▁dishes	-7174
+▁presents	-7175
+▁producing	-7176
+▁specialist	-7177
+▁communications	-7178
+ati	-7179
+lane	-7180
+aware	-7181
+excell	-7182
+▁nights	-7183
+▁baseball	-7184
+▁channels	-7185
+anny	-7186
+ista	-7187
+wart	-7188
+▁trig	-7189
+writer	-7190
+explore	-7191
+▁singer	-7192
+▁crucial	-7193
+▁explains	-7194
+▁expressed	-7195
+▁rum	-7196
+▁substant	-7197
+pton	-7198
+▁fits	-7199
+ronics	-7200
+original	-7201
+info	-7202
+categ	-7203
+shore	-7204
+▁ensu	-7205
+▁speaker	-7206
+▁ceremony	-7207
+▁newspaper	-7208
+▁pharm	-7209
+▁cookies	-7210
+bott	-7211
+▁gym	-7212
+▁rig	-7213
+▁formal	-7214
+▁founder	-7215
+commercial	-7216
+▁authority	-7217
+kee	-7218
+url	-7219
+hart	-7220
+oker	-7221
+▁sor	-7222
+modern	-7223
+article	-7224
+▁compare	-7225
+▁proposal	-7226
+▁programming	-7227
+etry	-7228
+tense	-7229
+always	-7230
+events	-7231
+▁pride	-7232
+issions	-7233
+▁claimed	-7234
+gor	-7235
+▁ads	-7236
+▁alive	-7237
+▁stuck	-7238
+▁remark	-7239
+▁carrying	-7240
+▁expansion	-7241
+alo	-7242
+lady	-7243
+marc	-7244
+coach	-7245
+early	-7246
+sorry	-7247
+▁vintage	-7248
+▁achieved	-7249
+▁speakers	-7250
+▁whenever	-7251
+▁enforcement	-7252
+ran	-7253
+▁mand	-7254
+▁agric	-7255
+▁consists	-7256
+eliz	-7257
+shel	-7258
+inois	-7259
+▁enem	-7260
+abetes	-7261
+castle	-7262
+▁displ	-7263
+closure	-7264
+▁dollar	-7265
+electric	-7266
+▁reducing	-7267
+ifies	-7268
+sters	-7269
+▁vast	-7270
+celebr	-7271
+lessly	-7272
+muslim	-7273
+airport	-7274
+fashion	-7275
+▁centers	-7276
+▁percentage	-7277
+lah	-7278
+yan	-7279
+▁conv	-7280
+▁saved	-7281
+▁limits	-7282
+▁engaged	-7283
+coll	-7284
+wine	-7285
+▁fru	-7286
+stein	-7287
+▁boss	-7288
+▁embr	-7289
+▁clubs	-7290
+manchester	-7291
+▁connections	-7292
+▁preparation	-7293
+nik	-7294
+imize	-7295
+hallow	-7296
+▁seats	-7297
+▁hotels	-7298
+▁combine	-7299
+▁dealing	-7300
+▁desired	-7301
+▁grateful	-7302
+additional	-7303
+▁consultation	-7304
+▁manufacturers	-7305
+ilty	-7306
+▁rein	-7307
+▁vice	-7308
+itable	-7309
+▁pricing	-7310
+▁designers	-7311
+▁extension	-7312
+▁impossible	-7313
+eman	-7314
+gene	-7315
+pain	-7316
+along	-7317
+acious	-7318
+▁heating	-7319
+▁watched	-7320
+▁resulting	-7321
+▁improvements	-7322
+tro	-7323
+fant	-7324
+▁attending	-7325
+▁basically	-7326
+mate	-7327
+civil	-7328
+▁compr	-7329
+▁gender	-7330
+▁insight	-7331
+▁compliance	-7332
+▁purchasing	-7333
+gt	-7334
+wid	-7335
+▁ly	-7336
+anger	-7337
+youth	-7338
+osophy	-7339
+regional	-7340
+assistant	-7341
+singapore	-7342
+▁participating	-7343
+poly	-7344
+strong	-7345
+▁crash	-7346
+▁carpet	-7347
+▁narrow	-7348
+▁packed	-7349
+illinois	-7350
+sylvania	-7351
+▁winners	-7352
+▁shoulder	-7353
+▁enterprise	-7354
+▁introduction	-7355
+rab	-7356
+pitt	-7357
+zone	-7358
+stick	-7359
+engers	-7360
+▁liber	-7361
+▁marks	-7362
+manufact	-7363
+▁concrete	-7364
+▁contrast	-7365
+▁sensitive	-7366
+’,	-7367
+aka	-7368
+nin	-7369
+obe	-7370
+gift	-7371
+▁plot	-7372
+brother	-7373
+pending	-7374
+▁acting	-7375
+▁lesson	-7376
+▁default	-7377
+▁exceptional	-7378
+lam	-7379
+aska	-7380
+hind	-7381
+ingu	-7382
+▁clar	-7383
+▁deck	-7384
+▁lady	-7385
+▁strike	-7386
+planning	-7387
+questions	-7388
+sir	-7389
+impro	-7390
+▁consc	-7391
+designed	-7392
+ications	-7393
+too	-7394
+harry	-7395
+▁assum	-7396
+▁aggress	-7397
+▁suffering	-7398
+mun	-7399
+ulum	-7400
+jordan	-7401
+▁gaming	-7402
+▁everywhere	-7403
+alt	-7404
+moo	-7405
+oir	-7406
+zing	-7407
+▁dad	-7408
+orous	-7409
+▁opens	-7410
+▁acquis	-7411
+industry	-7412
+▁booking	-7413
+▁constant	-7414
+▁consumption	-7415
+fu	-7416
+mend	-7417
+creek	-7418
+girls	-7419
+▁ocean	-7420
+▁whilst	-7421
+▁succeed	-7422
+resources	-7423
+bush	-7424
+near	-7425
+▁legend	-7426
+▁healing	-7427
+▁opinions	-7428
+ila	-7429
+alle	-7430
+apol	-7431
+aund	-7432
+built	-7433
+close	-7434
+study	-7435
+▁bills	-7436
+▁criteria	-7437
+▁overwhel	-7438
+▁obviously	-7439
+usd	-7440
+onna	-7441
+▁hel	-7442
+arian	-7443
+index	-7444
+korea	-7445
+▁spray	-7446
+updated	-7447
+▁humans	-7448
+▁rising	-7449
+▁damaged	-7450
+▁heading	-7451
+▁functional	-7452
+▁resistance	-7453
+▁lie	-7454
+henry	-7455
+▁unus	-7456
+length	-7457
+theast	-7458
+▁actor	-7459
+ologist	-7460
+features	-7461
+▁preferred	-7462
+▁networking	-7463
+sha	-7464
+ghan	-7465
+▁mob	-7466
+ockey	-7467
+▁mold	-7468
+terior	-7469
+anthony	-7470
+itution	-7471
+▁driven	-7472
+▁gentle	-7473
+▁instant	-7474
+▁assault	-7475
+▁weapons	-7476
+▁experiment	-7477
+▁announcement	-7478
+qual	-7479
+mates	-7480
+▁fundra	-7481
+▁indeed	-7482
+▁literally	-7483
+pennsylvania	-7484
+)||	-7485
+▁ens	-7486
+dress	-7487
+▁lect	-7488
+▁mere	-7489
+▁pill	-7490
+univers	-7491
+▁attribut	-7492
+▁interact	-7493
+▁scientists	-7494
+jess	-7495
+▁bridge	-7496
+▁emails	-7497
+iders	-7498
+▁managers	-7499
+▁retirement	-7500
+nat	-7501
+▁pic	-7502
+analy	-7503
+kelly	-7504
+▁trim	-7505
+kitchen	-7506
+▁returning	-7507
+corporation	-7508
+reach	-7509
+▁loose	-7510
+▁seeds	-7511
+▁anymore	-7512
+▁checking	-7513
+▁entering	-7514
+▁displayed	-7515
+▁personality	-7516
+cad	-7517
+▁tor	-7518
+▁beef	-7519
+▁wins	-7520
+matthe	-7521
+novation	-7522
+▁podcast	-7523
+▁ensuring	-7524
+▁suggests	-7525
+▁talented	-7526
+api	-7527
+bab	-7528
+iac	-7529
+ints	-7530
+mann	-7531
+onut	-7532
+stars	-7533
+users	-7534
+▁solo	-7535
+▁intense	-7536
+▁admitted	-7537
+▁component	-7538
+eye	-7539
+▁plain	-7540
+▁deposit	-7541
+currently	-7542
+▁preparing	-7543
+▁basketball	-7544
+▁interviews	-7545
+boys	-7546
+cand	-7547
+dney	-7548
+▁gay	-7549
+▁tub	-7550
+icity	-7551
+intel	-7552
+langu	-7553
+sarah	-7554
+write	-7555
+better	-7556
+▁franch	-7557
+major	-7558
+integr	-7559
+▁yours	-7560
+onymous	-7561
+▁string	-7562
+▁farmers	-7563
+▁association	-7564
+itz	-7565
+sts	-7566
+holy	-7567
+liber	-7568
+cooper	-7569
+images	-7570
+▁forgot	-7571
+▁removal	-7572
+▁visible	-7573
+▁introduce	-7574
+▁packaging	-7575
+▁officially	-7576
+hem	-7577
+alan	-7578
+anti	-7579
+iraq	-7580
+hills	-7581
+vised	-7582
+▁boot	-7583
+▁trips	-7584
+▁accused	-7585
+▁matching	-7586
+▁politics	-7587
+▁wondering	-7588
+mw	-7589
+hub	-7590
+iowa	-7591
+activ	-7592
+▁rural	-7593
+jection	-7594
+▁acceler	-7595
+amm	-7596
+una	-7597
+▁hi	-7598
+athon	-7599
+ilton	-7600
+▁chart	-7601
+▁lemon	-7602
+▁lists	-7603
+▁olive	-7604
+▁salad	-7605
+▁invite	-7606
+▁closing	-7607
+▁graphic	-7608
+▁forecast	-7609
+protection	-7610
+▁creativity	-7611
+▁reasonable	-7612
+additionally	-7613
+etch	-7614
+focus	-7615
+former	-7616
+▁tired	-7617
+▁stable	-7618
+▁diseases	-7619
+▁lifetime	-7620
+▁increasingly	-7621
+fan	-7622
+▁batt	-7623
+▁pace	-7624
+aining	-7625
+bourne	-7626
+▁blind	-7627
+ression	-7628
+comments	-7629
+▁staying	-7630
+▁findings	-7631
+▁licensed	-7632
+▁professor	-7633
+▁puts	-7634
+acular	-7635
+▁focuses	-7636
+halloween	-7637
+ventional	-7638
+value	-7639
+taking	-7640
+▁recip	-7641
+success	-7642
+theatre	-7643
+▁trailer	-7644
+▁viewing	-7645
+▁stations	-7646
+▁correspond	-7647
+eration	-7648
+▁afraid	-7649
+▁teaspoon	-7650
+▁primarily	-7651
+▁potentially	-7652
+cn	-7653
+kh	-7654
+hemat	-7655
+stage	-7656
+▁brush	-7657
+▁nomin	-7658
+restaur	-7659
+▁suspect	-7660
+▁tracking	-7661
+▁developer	-7662
+▁temperatures	-7663
+oen	-7664
+teac	-7665
+defin	-7666
+eness	-7667
+▁widely	-7668
+beck	-7669
+ifer	-7670
+snow	-7671
+onder	-7672
+utter	-7673
+import	-7674
+address	-7675
+pection	-7676
+▁illegal	-7677
+compan	-7678
+▁chips	-7679
+▁attach	-7680
+▁naturally	-7681
+anes	-7682
+arth	-7683
+▁sail	-7684
+simply	-7685
+▁honey	-7686
+liament	-7687
+▁racing	-7688
+▁precise	-7689
+elizabeth	-7690
+jen	-7691
+ador	-7692
+imag	-7693
+▁ves	-7694
+▁faced	-7695
+▁stamp	-7696
+property	-7697
+blu	-7698
+bern	-7699
+andre	-7700
+icide	-7701
+maker	-7702
+friger	-7703
+▁intent	-7704
+▁visits	-7705
+▁texture	-7706
+▁stronger	-7707
+sac	-7708
+oked	-7709
+skin	-7710
+ented	-7711
+▁clock	-7712
+▁drama	-7713
+independ	-7714
+▁greatly	-7715
+▁expenses	-7716
+▁releases	-7717
+employment	-7718
+▁reduction	-7719
+uten	-7720
+psych	-7721
+▁yoga	-7722
+▁ultim	-7723
+supreme	-7724
+▁harder	-7725
+'.	-7726
+sq	-7727
+oses	-7728
+rober	-7729
+scale	-7730
+steven	-7731
+britain	-7732
+▁flying	-7733
+▁suppor	-7734
+tennes	-7735
+▁bacter	-7736
+▁behalf	-7737
+▁conserv	-7738
+▁strange	-7739
+▁obtained	-7740
+▁computers	-7741
+zzle	-7742
+ureau	-7743
+▁plug	-7744
+▁unex	-7745
+lights	-7746
+▁blogs	-7747
+pakistan	-7748
+▁checked	-7749
+▁warning	-7750
+rt	-7751
+glas	-7752
+▁fault	-7753
+▁yield	-7754
+nob	-7755
+kansas	-7756
+▁weren	-7757
+selling	-7758
+▁gained	-7759
+▁writes	-7760
+advanced	-7761
+▁careful	-7762
+▁warranty	-7763
+▁workshops	-7764
+▁delivering	-7765
+kong	-7766
+tips	-7767
+eland	-7768
+saint	-7769
+▁museum	-7770
+▁colleagues	-7771
+▁highlights	-7772
+gab	-7773
+rele	-7774
+ashed	-7775
+owner	-7776
+rency	-7777
+oregon	-7778
+owners	-7779
+rating	-7780
+▁param	-7781
+nothing	-7782
+▁principles	-7783
+▁nose	-7784
+iliate	-7785
+▁desktop	-7786
+▁superior	-7787
+▁employers	-7788
+uct	-7789
+unit	-7790
+vals	-7791
+▁vine	-7792
+simple	-7793
+andemic	-7794
+machine	-7795
+▁gotten	-7796
+sciences	-7797
+▁elegant	-7798
+▁raising	-7799
+application	-7800
+bull	-7801
+laure	-7802
+▁duty	-7803
+▁semi	-7804
+▁entit	-7805
+village	-7806
+▁controls	-7807
+▁communicate	-7808
+fm	-7809
+hello	-7810
+▁shapes	-7811
+▁struggling	-7812
+yl	-7813
+hav	-7814
+austin	-7815
+▁minimal	-7816
+▁electrical	-7817
+vc	-7818
+▁lip	-7819
+rants	-7820
+▁inventory	-7821
+▁promoting	-7822
+▁approaches	-7823
+barn	-7824
+hong	-7825
+vegas	-7826
+▁sour	-7827
+▁wave	-7828
+▁strict	-7829
+▁hanging	-7830
+▁shipped	-7831
+▁sponsor	-7832
+▁photographs	-7833
+aha	-7834
+tery	-7835
+▁twist	-7836
+▁wooden	-7837
+▁cameras	-7838
+deal	-7839
+math	-7840
+▁ther	-7841
+odge	-7842
+sels	-7843
+▁cher	-7844
+▁laid	-7845
+▁align	-7846
+▁scores	-7847
+victoria	-7848
+▁controvers	-7849
+▁ultimately	-7850
+▁till	-7851
+▁divid	-7852
+▁anxiety	-7853
+▁involves	-7854
+▁occurred	-7855
+▁signature	-7856
+▁participation	-7857
+bish	-7858
+▁emb	-7859
+frame	-7860
+princ	-7861
+stract	-7862
+getting	-7863
+▁resume	-7864
+▁struck	-7865
+▁thinks	-7866
+▁cancell	-7867
+▁involving	-7868
+▁	-7869
+e	-7870
+t	-7871
+a	-7872
+o	-7873
+i	-7874
+n	-7875
+s	-7876
+r	-7877
+h	-7878
+l	-7879
+d	-7880
+c	-7881
+u	-7882
+m	-7883
+p	-7884
+f	-7885
+g	-7886
+y	-7887
+w	-7888
+b	-7889
+.	-7890
+v	-7891
+,	-7892
+k	-7893
+-	-7894
+0	-7895
+1	-7896
+x	-7897
+2	-7898
+j	-7899
+’	-7900
+'	-7901
+:	-7902
+z	-7903
+q	-7904
+)	-7905
+3	-7906
+(	-7907
+5	-7908
+4	-7909
+"	-7910
+9	-7911
+6	-7912
+8	-7913
+!	-7914
+|	-7915
+7	-7916
+/	-7917
+?	-7918
+“	-7919
+”	-7920
+–	-7921
+;	-7922
+&	-7923
+$	-7924
+%	-7925
+—	-7926
+*	-7927
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_gpt.py b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_gpt.py
new file mode 100644
index 0000000000..6ea42d03e0
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_gpt.py
@@ -0,0 +1,3143 @@
+import base64, collections, copy, fcntl, glob, io, lzma, math, os
+from pathlib import Path
+import random, re, subprocess, time, uuid, numpy as np, sentencepiece as spm, torch, torch.distributed as dist, torch.nn.functional as F
+from torch import Tensor, nn
+from flash_attn_interface import (
+    flash_attn_func as flash_attn_3_func,
+    flash_attn_varlen_func,
+)
+from concurrent.futures import ThreadPoolExecutor
+import triton
+import triton.language as tl
+from triton.tools.tensor_descriptor import TensorDescriptor
+_FUSED_CE_LIBRARY = "pgsubmission1draft7fusedce"
+_FUSED_CE_BLOCK_SIZE = 1024
+_FUSED_CE_NUM_WARPS = 4
+@triton.jit
+def _softcapped_ce_fwd_kernel(
+    logits_ptr, losses_ptr, lse_ptr, targets_ptr,
+    stride_logits_n, stride_logits_v,
+    n_rows, n_cols, softcap,
+    block_size: tl.constexpr,
+):
+    row_idx = tl.program_id(0).to(tl.int64)
+    logits_row_ptr = logits_ptr + row_idx * stride_logits_n
+    max_val = -float("inf")
+    sum_exp = 0.0
+    A = 2.0 * softcap
+    inv_C = 2.0 / softcap
+    for off in range(0, n_cols, block_size):
+        cols = off + tl.arange(0, block_size)
+        mask = cols < n_cols
+        val = tl.load(
+            logits_row_ptr + cols * stride_logits_v,
+            mask=mask, other=-float("inf"),
+        ).to(tl.float32)
+        z = A * tl.sigmoid(val * inv_C)
+        z = tl.where(mask, z, -float("inf"))
+        curr_max = tl.max(z, axis=0)
+        new_max = tl.maximum(max_val, curr_max)
+        sum_exp = sum_exp * tl.exp(max_val - new_max) + tl.sum(tl.exp(z - new_max), axis=0)
+        max_val = new_max
+    lse = max_val + tl.log(sum_exp)
+    tl.store(lse_ptr + row_idx, lse)
+    target = tl.load(targets_ptr + row_idx).to(tl.int32)
+    target_val = tl.load(logits_row_ptr + target * stride_logits_v).to(tl.float32)
+    target_z = A * tl.sigmoid(target_val * inv_C)
+    tl.store(losses_ptr + row_idx, lse - target_z)
+@triton.jit
+def _softcapped_ce_bwd_kernel(
+    grad_logits_ptr, grad_losses_ptr, lse_ptr, logits_ptr, targets_ptr,
+    stride_logits_n, stride_logits_v,
+    stride_grad_n, stride_grad_v,
+    n_rows, n_cols, softcap,
+    block_size: tl.constexpr,
+):
+    row_idx = tl.program_id(0).to(tl.int64)
+    logits_row_ptr = logits_ptr + row_idx * stride_logits_n
+    grad_row_ptr = grad_logits_ptr + row_idx * stride_grad_n
+    lse = tl.load(lse_ptr + row_idx)
+    grad_loss = tl.load(grad_losses_ptr + row_idx).to(tl.float32)
+    target = tl.load(targets_ptr + row_idx).to(tl.int32)
+    A = 2.0 * softcap
+    inv_C = 2.0 / softcap
+    dz_dx_scale = A * inv_C
+    for off in range(0, n_cols, block_size):
+        cols = off + tl.arange(0, block_size)
+        mask = cols < n_cols
+        val = tl.load(
+            logits_row_ptr + cols * stride_logits_v,
+            mask=mask, other=0.0,
+        ).to(tl.float32)
+        sigmoid_u = tl.sigmoid(val * inv_C)
+        z = A * sigmoid_u
+        probs = tl.exp(z - lse)
+        grad_z = grad_loss * (probs - tl.where(cols == target, 1.0, 0.0))
+        grad_x = grad_z * (dz_dx_scale * sigmoid_u * (1.0 - sigmoid_u))
+        tl.store(grad_row_ptr + cols * stride_grad_v, grad_x, mask=mask)
+def _validate_softcapped_ce_inputs(
+    logits: Tensor, targets: Tensor, softcap: float,
+) -> tuple[Tensor, Tensor]:
+    if logits.ndim != 2:
+        raise ValueError(f"Expected logits.ndim=2, got {logits.ndim}")
+    if targets.ndim != 1:
+        raise ValueError(f"Expected targets.ndim=1, got {targets.ndim}")
+    if logits.shape[0] != targets.shape[0]:
+        raise ValueError(
+            f"Expected matching rows, got logits={tuple(logits.shape)} targets={tuple(targets.shape)}"
+        )
+    if not logits.is_cuda or not targets.is_cuda:
+        raise ValueError("softcapped_cross_entropy requires CUDA tensors")
+    if softcap <= 0.0:
+        raise ValueError(f"softcap must be positive, got {softcap}")
+    if logits.dtype not in (torch.float16, torch.bfloat16, torch.float32):
+        raise ValueError(f"Unsupported logits dtype: {logits.dtype}")
+    logits = logits.contiguous()
+    targets = targets.contiguous()
+    if targets.dtype != torch.int64:
+        targets = targets.to(dtype=torch.int64)
+    return logits, targets
+@torch.library.custom_op(f"{_FUSED_CE_LIBRARY}::softcapped_ce", mutates_args=())
+def softcapped_ce_op(logits: Tensor, targets: Tensor, softcap: float) -> tuple[Tensor, Tensor]:
+    logits, targets = _validate_softcapped_ce_inputs(logits, targets, float(softcap))
+    n_rows, n_cols = logits.shape
+    losses = torch.empty((n_rows,), device=logits.device, dtype=torch.float32)
+    lse = torch.empty((n_rows,), device=logits.device, dtype=torch.float32)
+    _softcapped_ce_fwd_kernel[(n_rows,)](
+        logits, losses, lse, targets,
+        logits.stride(0), logits.stride(1),
+        n_rows, n_cols, float(softcap),
+        block_size=_FUSED_CE_BLOCK_SIZE, num_warps=_FUSED_CE_NUM_WARPS,
+    )
+    return losses, lse
+@softcapped_ce_op.register_fake
+def _(logits: Tensor, targets: Tensor, softcap: float):
+    if logits.ndim != 2 or targets.ndim != 1:
+        raise ValueError("softcapped_ce fake impl expects 2D logits and 1D targets")
+    if logits.shape[0] != targets.shape[0]:
+        raise ValueError(
+            f"Expected matching rows, got logits={tuple(logits.shape)} targets={tuple(targets.shape)}"
+        )
+    n_rows = logits.shape[0]
+    return (
+        logits.new_empty((n_rows,), dtype=torch.float32),
+        logits.new_empty((n_rows,), dtype=torch.float32),
+    )
+@torch.library.custom_op(f"{_FUSED_CE_LIBRARY}::softcapped_ce_backward", mutates_args=())
+def softcapped_ce_backward_op(
+    logits: Tensor, targets: Tensor, lse: Tensor, grad_losses: Tensor, softcap: float,
+) -> Tensor:
+    logits, targets = _validate_softcapped_ce_inputs(logits, targets, float(softcap))
+    lse = lse.contiguous()
+    grad_losses = grad_losses.contiguous().to(dtype=torch.float32)
+    if lse.ndim != 1 or grad_losses.ndim != 1:
+        raise ValueError("Expected 1D lse and grad_losses")
+    if lse.shape[0] != logits.shape[0] or grad_losses.shape[0] != logits.shape[0]:
+        raise ValueError(
+            f"Expected row-aligned lse/grad_losses, got logits={tuple(logits.shape)} "
+            f"lse={tuple(lse.shape)} grad_losses={tuple(grad_losses.shape)}"
+        )
+    grad_logits = torch.empty_like(logits)
+    n_rows, n_cols = logits.shape
+    _softcapped_ce_bwd_kernel[(n_rows,)](
+        grad_logits, grad_losses, lse, logits, targets,
+        logits.stride(0), logits.stride(1),
+        grad_logits.stride(0), grad_logits.stride(1),
+        n_rows, n_cols, float(softcap),
+        block_size=_FUSED_CE_BLOCK_SIZE, num_warps=_FUSED_CE_NUM_WARPS,
+    )
+    return grad_logits
+@softcapped_ce_backward_op.register_fake
+def _(logits: Tensor, targets: Tensor, lse: Tensor, grad_losses: Tensor, softcap: float):
+    if logits.ndim != 2 or targets.ndim != 1 or lse.ndim != 1 or grad_losses.ndim != 1:
+        raise ValueError("softcapped_ce_backward fake impl expects 2D logits and 1D row tensors")
+    if (
+        logits.shape[0] != targets.shape[0]
+        or logits.shape[0] != lse.shape[0]
+        or logits.shape[0] != grad_losses.shape[0]
+    ):
+        raise ValueError("softcapped_ce_backward fake impl expects row-aligned tensors")
+    return logits.new_empty(logits.shape)
+def _softcapped_ce_setup_context(
+    ctx: torch.autograd.function.FunctionCtx, inputs, output,
+) -> None:
+    logits, targets, softcap = inputs
+    _losses, lse = output
+    ctx.save_for_backward(logits, targets, lse)
+    ctx.softcap = float(softcap)
+def _softcapped_ce_backward(
+    ctx: torch.autograd.function.FunctionCtx, grad_losses: Tensor, grad_lse: "Tensor | None",
+):
+    del grad_lse
+    logits, targets, lse = ctx.saved_tensors
+    grad_logits = torch.ops.pgsubmission1draft7fusedce.softcapped_ce_backward(
+        logits, targets, lse, grad_losses, ctx.softcap
+    )
+    return grad_logits, None, None
+softcapped_ce_op.register_autograd(
+    _softcapped_ce_backward, setup_context=_softcapped_ce_setup_context,
+)
+def softcapped_cross_entropy(
+    logits: Tensor, targets: Tensor, softcap: float, reduction: str = "mean",
+) -> Tensor:
+    losses, _lse = torch.ops.pgsubmission1draft7fusedce.softcapped_ce(
+        logits, targets, float(softcap)
+    )
+    if reduction == "none":
+        return losses
+    if reduction == "sum":
+        return losses.sum()
+    if reduction == "mean":
+        return losses.mean()
+    raise ValueError(f"Unsupported reduction={reduction!r}")
+class Hyperparameters:
+    data_dir = os.environ.get("DATA_DIR", "./data/")
+    seed = int(os.environ.get("SEED", 42))
+    run_id = os.environ.get("RUN_ID", str(uuid.uuid4()))
+    iterations = int(os.environ.get("ITERATIONS", 20000))
+    warmdown_frac = float(os.environ.get("WARMDOWN_FRAC", 0.75))
+    warmup_steps = int(os.environ.get("WARMUP_STEPS", 20))
+    train_batch_tokens = int(os.environ.get("TRAIN_BATCH_TOKENS", 786432))
+    fused_ce_enabled = bool(int(os.environ.get("FUSED_CE_ENABLED", "1")))
+    train_seq_len = int(os.environ.get("TRAIN_SEQ_LEN", 2048))
+    train_log_every = int(os.environ.get("TRAIN_LOG_EVERY", 500))
+    max_wallclock_seconds = float(os.environ.get("MAX_WALLCLOCK_SECONDS", 6e2))
+    val_batch_tokens = int(os.environ.get("VAL_BATCH_TOKENS", 524288))
+    eval_seq_len = int(os.environ.get("EVAL_SEQ_LEN", 2048))
+    val_loss_every = int(os.environ.get("VAL_LOSS_EVERY", 4000))
+    vocab_size = int(os.environ.get("VOCAB_SIZE", 8192))
+    num_layers = int(os.environ.get("NUM_LAYERS", 11))
+    xsa_last_n = int(os.environ.get("XSA_LAST_N", 11))
+    model_dim = int(os.environ.get("MODEL_DIM", 512))
+    num_kv_heads = int(os.environ.get("NUM_KV_HEADS", 4))
+    num_heads = int(os.environ.get("NUM_HEADS", 8))
+    mlp_mult = float(os.environ.get("MLP_MULT", 4.0))
+    skip_gates_enabled = bool(int(os.environ.get("SKIP_GATES_ENABLED", "1")))
+    logit_softcap = float(os.environ.get("LOGIT_SOFTCAP", 3e1))
+    rope_base = float(os.environ.get("ROPE_BASE", 1e4))
+    rope_dims = int(os.environ.get("ROPE_DIMS", 16))
+    rope_train_seq_len = int(os.environ.get("ROPE_TRAIN_SEQ_LEN", 2048))
+    rope_yarn = bool(int(os.environ.get("ROPE_YARN", "0")))
+    ln_scale = bool(int(os.environ.get("LN_SCALE", "1")))
+    qk_gain_init = float(os.environ.get("QK_GAIN_INIT", 5.0))
+    num_loops = int(os.environ.get("NUM_LOOPS", 2))
+    loop_start = int(os.environ.get("LOOP_START", 3))
+    loop_end = int(os.environ.get("LOOP_END", 5))
+    enable_looping_at = float(os.environ.get("ENABLE_LOOPING_AT", 0.35))
+    parallel_start_layer = int(os.environ.get("PARALLEL_START_LAYER", 8))
+    parallel_final_lane = os.environ.get("PARALLEL_FINAL_LANE", "mean")
+    min_lr = float(os.environ.get("MIN_LR", 0.1))
+    tied_embed_lr = float(os.environ.get("TIED_EMBED_LR", 0.03))
+    tied_embed_init_std = float(os.environ.get("TIED_EMBED_INIT_STD", 0.005))
+    matrix_lr = float(os.environ.get("MATRIX_LR", 0.026))
+    scalar_lr = float(os.environ.get("SCALAR_LR", 0.02))
+    muon_momentum = float(os.environ.get("MUON_MOMENTUM", 0.97))
+    muon_backend_steps = int(os.environ.get("MUON_BACKEND_STEPS", 5))
+    muon_momentum_warmup_start = float(
+        os.environ.get("MUON_MOMENTUM_WARMUP_START", 0.92)
+    )
+    muon_momentum_warmup_steps = int(os.environ.get("MUON_MOMENTUM_WARMUP_STEPS", 1500))
+    muon_row_normalize = bool(int(os.environ.get("MUON_ROW_NORMALIZE", "1")))
+    beta1 = float(os.environ.get("BETA1", 0.9))
+    beta2 = float(os.environ.get("BETA2", 0.95))
+    adam_eps = float(os.environ.get("ADAM_EPS", 1e-08))
+    grad_clip_norm = float(os.environ.get("GRAD_CLIP_NORM", 0.3))
+    eval_stride = int(os.environ.get("EVAL_STRIDE", 64))
+    adam_wd = float(os.environ.get("ADAM_WD", 0.02))
+    muon_wd = float(os.environ.get("MUON_WD", 0.095))
+    embed_wd = float(os.environ.get("EMBED_WD", 0.085))
+    ema_decay = float(os.environ.get("EMA_DECAY", 0.9965))
+    ttt_enabled = bool(int(os.environ.get("TTT_ENABLED", "1")))
+    ttt_lora_rank = int(os.environ.get("TTT_LORA_RANK", 96))
+    ttt_lora_lr = float(os.environ.get("TTT_LORA_LR", 0.0001))
+    ttt_chunk_size = int(os.environ.get("TTT_CHUNK_SIZE", 48))
+    ttt_eval_seq_len = int(os.environ.get("TTT_EVAL_SEQ_LEN", 2048))
+    ttt_batch_size = int(os.environ.get("TTT_BATCH_SIZE", 16))
+    ttt_grad_steps = int(os.environ.get("TTT_GRAD_STEPS", 1))
+    ttt_weight_decay = float(os.environ.get("TTT_WEIGHT_DECAY", 1.0))
+    ttt_beta1 = float(os.environ.get("TTT_BETA1", 0))
+    ttt_beta2 = float(os.environ.get("TTT_BETA2", 0.999))
+    ttt_k_lora = bool(int(os.environ.get("TTT_K_LORA", "1")))
+    ttt_mlp_lora = bool(int(os.environ.get("TTT_MLP_LORA", "1")))
+    ttt_o_lora = bool(int(os.environ.get("TTT_O_LORA", "1")))
+    ttt_optimizer = os.environ.get("TTT_OPTIMIZER", "adam")
+    ttt_eval_batches = os.environ.get("TTT_EVAL_BATCHES", "")
+    val_doc_fraction = float(os.environ.get("VAL_DOC_FRACTION", 1.0))
+    compressor = os.environ.get("COMPRESSOR", "brotli")
+    gptq_calibration_batches = int(os.environ.get("GPTQ_CALIBRATION_BATCHES", 16))
+    gptq_reserve_seconds = float(os.environ.get("GPTQ_RESERVE_SECONDS", 16.0))
+    phased_ttt_prefix_docs = int(os.environ.get("PHASED_TTT_PREFIX_DOCS", 2000))
+    phased_ttt_num_phases = int(os.environ.get("PHASED_TTT_NUM_PHASES", 3))
+    global_ttt_lr = float(os.environ.get("GLOBAL_TTT_LR", 0.001))
+    global_ttt_momentum = float(os.environ.get("GLOBAL_TTT_MOMENTUM", 0.9))
+    global_ttt_epochs = int(os.environ.get("GLOBAL_TTT_EPOCHS", 1))
+    global_ttt_chunk_tokens = int(os.environ.get("GLOBAL_TTT_CHUNK_TOKENS", 32768))
+    global_ttt_batch_seqs = int(os.environ.get("GLOBAL_TTT_BATCH_SEQS", 32))
+    global_ttt_warmup_start_lr = float(os.environ.get("GLOBAL_TTT_WARMUP_START_LR", 0.0))
+    global_ttt_warmup_chunks = int(os.environ.get("GLOBAL_TTT_WARMUP_CHUNKS", 0))
+    global_ttt_grad_clip = float(os.environ.get("GLOBAL_TTT_GRAD_CLIP", 1.0))
+    global_ttt_respect_doc_boundaries = bool(int(os.environ.get("GLOBAL_TTT_RESPECT_DOC_BOUNDARIES", "1")))
+    matrix_bits = int(os.environ.get("MATRIX_BITS", 6))
+    embed_bits = int(os.environ.get("EMBED_BITS", 7))
+    matrix_clip_sigmas = float(os.environ.get("MATRIX_CLIP_SIGMAS", 12.85))
+    embed_clip_sigmas = float(os.environ.get("EMBED_CLIP_SIGMAS", 15.0))
+    mlp_clip_sigmas = float(os.environ.get("MLP_CLIP_SIGMAS", 12.0))
+    attn_clip_sigmas = float(os.environ.get("ATTN_CLIP_SIGMAS", 13.0))
+    attn_out_gate_enabled = bool(int(os.environ.get("ATTN_OUT_GATE_ENABLED", "0")))
+    attn_out_gate_src = os.environ.get("ATTN_OUT_GATE_SRC", "proj")
+    smear_gate_enabled = bool(int(os.environ.get("SMEAR_GATE_ENABLED", "1")))
+    gate_window = int(os.environ.get("GATE_WINDOW", 12))
+    gated_attn_enabled = bool(int(os.environ.get("GATED_ATTN_ENABLED", "0")))
+    gated_attn_init_std = float(os.environ.get("GATED_ATTN_INIT_STD", 0.01))
+    gated_attn_quant_gate = bool(int(os.environ.get("GATED_ATTN_QUANT_GATE", "1")))
+    sparse_attn_gate_enabled = bool(int(os.environ.get("SPARSE_ATTN_GATE_ENABLED", "1")))
+    sparse_attn_gate_init_std = float(os.environ.get("SPARSE_ATTN_GATE_INIT_STD", 0.0))
+    sparse_attn_gate_scale = float(os.environ.get("SPARSE_ATTN_GATE_SCALE", 1.0))
+    lqer_enabled = bool(int(os.environ.get("LQER_ENABLED", "1")))
+    lqer_rank = int(os.environ.get("LQER_RANK", 4))
+    lqer_top_k = int(os.environ.get("LQER_TOP_K", 1))
+    lqer_factor_bits = int(os.environ.get("LQER_FACTOR_BITS", 4))
+    lqer_asym_enabled = bool(int(os.environ.get("LQER_ASYM_ENABLED", "1")))
+    lqer_asym_group = int(os.environ.get("LQER_ASYM_GROUP", "64"))
+    distributed = "RANK" in os.environ and "WORLD_SIZE" in os.environ
+    rank = int(os.environ.get("RANK", "0"))
+    world_size = int(os.environ.get("WORLD_SIZE", "1"))
+    local_rank = int(os.environ.get("LOCAL_RANK", "0"))
+    is_main_process = rank == 0
+    grad_accum_steps = 8 // world_size
+    caseops_enabled = bool(int(os.environ.get("CASEOPS_ENABLED", "1")))
+    _default_caseops_data = os.path.join(
+        data_dir,
+        "datasets",
+        "fineweb10B_sp8192_caseops",
+        "datasets",
+        "datasets",
+        "fineweb10B_sp8192_lossless_caps_caseops_v1_reserved",
+    )
+    _default_caseops_tok = os.path.join(
+        data_dir,
+        "datasets",
+        "fineweb10B_sp8192_caseops",
+        "datasets",
+        "tokenizers",
+        "fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model",
+    )
+    if caseops_enabled:
+        datasets_dir = os.environ.get("DATA_PATH", _default_caseops_data)
+        tokenizer_path = os.environ.get("TOKENIZER_PATH", _default_caseops_tok)
+    else:
+        datasets_dir = os.environ.get(
+            "DATA_PATH",
+            os.path.join(data_dir, "datasets", f"fineweb10B_sp{vocab_size}"),
+        )
+        tokenizer_path = os.environ.get(
+            "TOKENIZER_PATH",
+            os.path.join(data_dir, "tokenizers", f"fineweb_{vocab_size}_bpe.model"),
+        )
+    train_files = os.path.join(datasets_dir, "fineweb_train_*.bin")
+    val_files = os.path.join(datasets_dir, "fineweb_val_*.bin")
+    val_bytes_files = os.path.join(datasets_dir, "fineweb_val_bytes_*.bin")
+    artifact_dir = os.environ.get("ARTIFACT_DIR", "")
+    logfile = (
+        os.path.join(artifact_dir, f"{run_id}.txt")
+        if artifact_dir
+        else f"logs/{run_id}.txt"
+    )
+    model_path = os.environ.get(
+        "MODEL_PATH",
+        os.path.join(artifact_dir, "final_model.pt")
+        if artifact_dir
+        else f"models/{run_id}.pt",
+    )
+    load_model_path = os.environ.get("LOAD_MODEL_PATH", model_path)
+    quantized_model_path = os.environ.get(
+        "QUANTIZED_MODEL_PATH",
+        os.path.join(artifact_dir, "final_model.int6.ptz")
+        if artifact_dir
+        else f"models/{run_id}.int{matrix_bits}.ptz",
+    )
+_logger_hparams = None
+def set_logging_hparams(h):
+    global _logger_hparams
+    _logger_hparams = h
+def log(msg, console=True):
+    if _logger_hparams is None:
+        print(msg)
+        return
+    if _logger_hparams.is_main_process:
+        if console:
+            print(msg)
+        if _logger_hparams.logfile is not None:
+            with open(_logger_hparams.logfile, "a", encoding="utf-8") as f:
+                print(msg, file=f)
+class ValidationData:
+    def __init__(self, h, device):
+        self.sp = spm.SentencePieceProcessor(model_file=h.tokenizer_path)
+        if int(self.sp.vocab_size()) != h.vocab_size:
+            raise ValueError(
+                f"VOCAB_SIZE={h.vocab_size} does not match tokenizer vocab_size={int(self.sp.vocab_size())}"
+            )
+        self.val_tokens = load_validation_tokens(h.val_files, h.eval_seq_len)
+        (
+            self.base_bytes_lut,
+            self.has_leading_space_lut,
+            self.is_boundary_token_lut,
+        ) = build_sentencepiece_luts(self.sp, h.vocab_size, device)
+        self.caseops_enabled = bool(getattr(h, "caseops_enabled", False))
+        self.val_bytes = None
+        if self.caseops_enabled:
+            self.val_bytes = load_validation_byte_sidecar(
+                h.val_bytes_files, h.eval_seq_len, self.val_tokens.numel()
+            )
+def build_sentencepiece_luts(sp, vocab_size, device):
+    sp_vocab_size = int(sp.vocab_size())
+    assert (
+        sp.piece_to_id("▁") != sp.unk_id()
+    ), "Tokenizer must have '▁' (space) as its own token for correct BPB byte counting"
+    table_size = max(sp_vocab_size, vocab_size)
+    base_bytes_np = np.zeros((table_size,), dtype=np.int16)
+    has_leading_space_np = np.zeros((table_size,), dtype=np.bool_)
+    is_boundary_token_np = np.ones((table_size,), dtype=np.bool_)
+    for token_id in range(sp_vocab_size):
+        if sp.is_control(token_id) or sp.is_unknown(token_id) or sp.is_unused(token_id):
+            continue
+        is_boundary_token_np[token_id] = False
+        if sp.is_byte(token_id):
+            base_bytes_np[token_id] = 1
+            continue
+        piece = sp.id_to_piece(token_id)
+        if piece.startswith("▁"):
+            has_leading_space_np[token_id] = True
+            piece = piece[1:]
+        base_bytes_np[token_id] = len(piece.encode("utf-8"))
+    return (
+        torch.tensor(base_bytes_np, dtype=torch.int16, device=device),
+        torch.tensor(has_leading_space_np, dtype=torch.bool, device=device),
+        torch.tensor(is_boundary_token_np, dtype=torch.bool, device=device),
+    )
+def load_validation_tokens(pattern, seq_len):
+    files = [
+        Path(p)
+        for p in sorted(glob.glob(pattern))
+        if "_bytes_" not in Path(p).name
+    ]
+    if not files:
+        raise FileNotFoundError(f"No files found for pattern: {pattern}")
+    tokens = torch.cat([load_data_shard(file) for file in files]).contiguous()
+    usable = (tokens.numel() - 1) // seq_len * seq_len
+    if usable <= 0:
+        raise ValueError(f"Validation split is too short for TRAIN_SEQ_LEN={seq_len}")
+    return tokens[: usable + 1]
+def load_validation_byte_sidecar(pattern, seq_len, expected_len):
+    files = [Path(p) for p in sorted(glob.glob(pattern))]
+    if not files:
+        raise FileNotFoundError(f"No byte sidecar files for pattern: {pattern}")
+    shards = [load_data_shard(file) for file in files]
+    bytes_full = torch.cat(shards).contiguous()
+    if bytes_full.numel() < expected_len:
+        raise ValueError(
+            f"Byte sidecar too short: {bytes_full.numel()} < val_tokens {expected_len}"
+        )
+    return bytes_full[:expected_len].to(torch.int32)
+def load_data_shard(file):
+    header_bytes = 256 * np.dtype("<i4").itemsize
+    token_bytes = np.dtype("<u2").itemsize
+    header = np.fromfile(file, dtype="<i4", count=256)
+    if header.size != 256 or int(header[0]) != 20240520 or int(header[1]) != 1:
+        raise ValueError(f"Unexpected shard header for {file}")
+    num_tokens = int(header[2])
+    expected_size = header_bytes + num_tokens * token_bytes
+    if file.stat().st_size != expected_size:
+        raise ValueError(
+            f"Shard size mismatch for {file}: expected {expected_size} bytes"
+        )
+    tokens_np = np.fromfile(file, dtype="<u2", count=num_tokens, offset=header_bytes)
+    if tokens_np.size != num_tokens:
+        raise ValueError(f"Short read for {file}")
+    return torch.from_numpy(tokens_np.astype(np.uint16, copy=False))
+_SHARD_HEADER_BYTES = 256 * np.dtype("<i4").itemsize
+_SHARD_NTOKENS_CACHE = {}
+_MMAP_CACHE = {}
+def _read_num_tokens(file):
+    key = str(file)
+    cached = _SHARD_NTOKENS_CACHE.get(key)
+    if cached is not None:
+        return cached
+    header = np.fromfile(file, dtype="<i4", count=256)
+    if header.size != 256 or int(header[0]) != 20240520 or int(header[1]) != 1:
+        raise ValueError(f"Unexpected shard header for {file}")
+    n = int(header[2])
+    _SHARD_NTOKENS_CACHE[key] = n
+    return n
+def _get_shard_memmap(file):
+    key = str(file)
+    mm = _MMAP_CACHE.get(key)
+    if mm is not None:
+        return mm
+    n = _read_num_tokens(file)
+    mm = np.memmap(file, mode="r", dtype="<u2", offset=_SHARD_HEADER_BYTES, shape=(n,))
+    _MMAP_CACHE[key] = mm
+    return mm
+BOS_ID = None
+def get_next_multiple_of_n(v, n):
+    return ((v + n - 1) // n) * n
+def _build_cu_seqlens(bos_pos, total_len, device, max_doc_len=0, bucket_size=64):
+    if not bos_pos or bos_pos[0] != 0:
+        bos_pos = [0] + bos_pos
+    seg_starts = []
+    starts_with_end = bos_pos + [total_len]
+    for i in range(len(starts_with_end) - 1):
+        start = starts_with_end[i]
+        end = starts_with_end[i + 1]
+        if max_doc_len > 0:
+            pos = start
+            while pos < end:
+                seg_starts.append(pos)
+                pos += max_doc_len
+        else:
+            seg_starts.append(start)
+    boundaries = seg_starts + [total_len]
+    padded_len = get_next_multiple_of_n(len(boundaries), bucket_size)
+    cu = torch.full((padded_len,), total_len, dtype=torch.int32, device=device)
+    cu[: len(boundaries)] = torch.tensor(boundaries, dtype=torch.int32, device=device)
+    seg_ends = seg_starts[1:] + [total_len]
+    max_seqlen = max(end - start for start, end in zip(seg_starts, seg_ends))
+    return cu, max_seqlen
+class DocumentPackingLoader:
+    _shard_pool = ThreadPoolExecutor(1)
+    def __init__(self, h, device, cu_bucket_size=64):
+        self.rank = h.rank
+        self.world_size = h.world_size
+        self.device = device
+        self.cu_bucket_size = cu_bucket_size
+        self.max_seq_len = h.train_seq_len
+        all_files = [Path(p) for p in sorted(glob.glob(h.train_files))]
+        if not all_files:
+            raise FileNotFoundError(f"No files found for pattern: {h.train_files}")
+        self.files = all_files
+        self.file_iter = iter(self.files)
+        self._init_shard(load_data_shard(next(self.file_iter)))
+        self._next_shard = self._submit_next_shard()
+        self._batch_pool = ThreadPoolExecutor(1)
+        self._next_batch = None
+    def _init_shard(self, tokens):
+        global BOS_ID
+        self.tokens = tokens
+        self.shard_size = tokens.numel()
+        if BOS_ID is None:
+            BOS_ID = 1
+        self.bos_idx = (
+            (tokens == BOS_ID).nonzero(as_tuple=True)[0].to(torch.int64).cpu().numpy()
+        )
+        self.cursor = int(self.bos_idx[0])
+    def _submit_next_shard(self):
+        try:
+            path = next(self.file_iter)
+            return self._shard_pool.submit(load_data_shard, path)
+        except StopIteration:
+            return None
+    def _advance_shard(self):
+        if self._next_shard is None:
+            self.file_iter = iter(self.files)
+            self._next_shard = self._shard_pool.submit(
+                load_data_shard, next(self.file_iter)
+            )
+        self._init_shard(self._next_shard.result())
+        self._next_shard = self._submit_next_shard()
+    def _local_doc_starts(self, local_start, total_len):
+        lo = np.searchsorted(self.bos_idx, local_start, side="left")
+        hi = np.searchsorted(self.bos_idx, local_start + total_len, side="left")
+        return (self.bos_idx[lo:hi] - local_start).tolist()
+    def _prepare_batch(self, num_tokens_local, max_seq_len):
+        per_rank_span = num_tokens_local + 1
+        global_span = per_rank_span * self.world_size
+        while self.cursor + global_span > self.shard_size:
+            self._advance_shard()
+        local_start = self.cursor + self.rank * per_rank_span
+        buf = self.tokens[local_start : local_start + per_rank_span]
+        inputs = buf[:-1].to(dtype=torch.int64).pin_memory()
+        targets = buf[1:].to(dtype=torch.int64).pin_memory()
+        starts = self._local_doc_starts(local_start, inputs.numel())
+        cu_seqlens, max_seqlen = _build_cu_seqlens(
+            starts, inputs.numel(), inputs.device, max_seq_len, self.cu_bucket_size
+        )
+        cu_seqlens = cu_seqlens.pin_memory()
+        self.cursor += global_span
+        return inputs, targets, cu_seqlens, max_seqlen
+    def next_batch(self, global_tokens, grad_accum_steps):
+        num_tokens_local = global_tokens // (self.world_size * grad_accum_steps)
+        if self._next_batch is not None:
+            inputs, targets, cu_seqlens, max_seqlen = self._next_batch.result()
+        else:
+            inputs, targets, cu_seqlens, max_seqlen = self._prepare_batch(
+                num_tokens_local, self.max_seq_len
+            )
+        self._next_batch = self._batch_pool.submit(
+            self._prepare_batch, num_tokens_local, self.max_seq_len
+        )
+        return (
+            inputs[None].to(self.device, non_blocking=True),
+            targets[None].to(self.device, non_blocking=True),
+            cu_seqlens.to(self.device, non_blocking=True),
+            max_seqlen,
+        )
+class ShuffledSequenceLoader:
+    def __init__(self, h, device):
+        self.world_size = h.world_size
+        self.seq_len = h.train_seq_len
+        self.device = device
+        all_files = [Path(p) for p in sorted(glob.glob(h.train_files))]
+        if not all_files:
+            raise FileNotFoundError(f"No files found for pattern: {h.train_files}")
+        self.files = all_files[h.rank :: h.world_size]
+        self.rng = np.random.Generator(np.random.PCG64(h.rank))
+        self.num_tokens = [_read_num_tokens(f) for f in self.files]
+        self.start_inds = [[] for _ in self.files]
+        for si in range(len(self.files)):
+            self._reset_shard(si)
+    def _reset_shard(self, si):
+        max_phase = min(
+            self.seq_len - 1, max(0, self.num_tokens[si] - self.seq_len - 1)
+        )
+        phase = int(self.rng.integers(max_phase + 1)) if max_phase > 0 else 0
+        num_sequences = (self.num_tokens[si] - 1 - phase) // self.seq_len
+        sequence_order = self.rng.permutation(num_sequences)
+        self.start_inds[si] = (phase + sequence_order * self.seq_len).tolist()
+    def next_batch(self, global_tokens, grad_accum_steps):
+        device_tokens = global_tokens // (self.world_size * grad_accum_steps)
+        device_batch_size = device_tokens // self.seq_len
+        remaining = np.array([len(s) for s in self.start_inds], dtype=np.float64)
+        x = torch.empty((device_batch_size, self.seq_len), dtype=torch.int64)
+        y = torch.empty((device_batch_size, self.seq_len), dtype=torch.int64)
+        for bi in range(device_batch_size):
+            total = remaining.sum()
+            if total <= 0:
+                for si in range(len(self.files)):
+                    self._reset_shard(si)
+                remaining = np.array(
+                    [len(s) for s in self.start_inds], dtype=np.float64
+                )
+                total = remaining.sum()
+            probs = remaining / total
+            si = int(self.rng.choice(len(self.files), p=probs))
+            start_ind = self.start_inds[si].pop()
+            remaining[si] -= 1
+            mm = _get_shard_memmap(self.files[si])
+            window = torch.as_tensor(
+                np.array(mm[start_ind : start_ind + self.seq_len + 1], dtype=np.int64)
+            )
+            x[bi] = window[:-1]
+            y[bi] = window[1:]
+        return x.to(self.device, non_blocking=True), y.to(
+            self.device, non_blocking=True
+        )
+class RMSNorm(nn.Module):
+    def __init__(self, eps=None):
+        super().__init__()
+        self.eps = eps
+    def forward(self, x):
+        return F.rms_norm(x, (x.size(-1),), eps=self.eps)
+class CastedLinear(nn.Linear):
+    def forward(self, x):
+        w = self.weight.to(x.dtype)
+        bias = self.bias.to(x.dtype) if self.bias is not None else None
+        return F.linear(x, w, bias)
+@triton.jit
+def linear_leaky_relu_square_kernel(
+    a_desc,
+    b_desc,
+    c_desc,
+    aux_desc,
+    M,
+    N,
+    K,
+    BLOCK_SIZE_M: tl.constexpr,
+    BLOCK_SIZE_N: tl.constexpr,
+    BLOCK_SIZE_K: tl.constexpr,
+    NUM_SMS: tl.constexpr,
+    FORWARD: tl.constexpr,
+):
+    dtype = tl.bfloat16
+    start_pid = tl.program_id(axis=0)
+    num_pid_m = tl.cdiv(M, BLOCK_SIZE_M)
+    num_pid_n = tl.cdiv(N, BLOCK_SIZE_N)
+    k_tiles = tl.cdiv(K, BLOCK_SIZE_K)
+    num_tiles = num_pid_m * num_pid_n
+    tile_id_c = start_pid - NUM_SMS
+    for tile_id in tl.range(start_pid, num_tiles, NUM_SMS, flatten=True):
+        pid_m = tile_id // num_pid_n
+        pid_n = tile_id % num_pid_n
+        offs_am = pid_m * BLOCK_SIZE_M
+        offs_bn = pid_n * BLOCK_SIZE_N
+        accumulator = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N), dtype=tl.float32)
+        for ki in range(k_tiles):
+            offs_k = ki * BLOCK_SIZE_K
+            a = a_desc.load([offs_am, offs_k])
+            b = b_desc.load([offs_bn, offs_k])
+            accumulator = tl.dot(a, b.T, accumulator)
+        tile_id_c += NUM_SMS
+        offs_am_c = offs_am
+        offs_bn_c = offs_bn
+        acc = tl.reshape(accumulator, (BLOCK_SIZE_M, 2, BLOCK_SIZE_N // 2))
+        acc = tl.permute(acc, (0, 2, 1))
+        acc0, acc1 = tl.split(acc)
+        c0 = acc0.to(dtype)
+        c1 = acc1.to(dtype)
+        if not FORWARD:
+            pre0 = aux_desc.load([offs_am_c, offs_bn_c])
+            pre1 = aux_desc.load([offs_am_c, offs_bn_c + BLOCK_SIZE_N // 2])
+            c0 = c0 * tl.where(pre0 > 0, 2.0 * pre0, 0.18 * pre0)
+            c1 = c1 * tl.where(pre1 > 0, 2.0 * pre1, 0.18 * pre1)
+        c_desc.store([offs_am_c, offs_bn_c], c0)
+        c_desc.store([offs_am_c, offs_bn_c + BLOCK_SIZE_N // 2], c1)
+        if FORWARD:
+            aux0 = tl.where(c0 > 0, c0, 0.3 * c0)
+            aux1 = tl.where(c1 > 0, c1, 0.3 * c1)
+            aux_desc.store([offs_am_c, offs_bn_c], aux0 * aux0)
+            aux_desc.store([offs_am_c, offs_bn_c + BLOCK_SIZE_N // 2], aux1 * aux1)
+def linear_leaky_relu_square(a, b, aux=None):
+    M, K = a.shape
+    N, K2 = b.shape
+    assert K == K2
+    c = torch.empty((M, N), device=a.device, dtype=a.dtype)
+    forward = aux is None
+    if aux is None:
+        aux = torch.empty((M, N), device=a.device, dtype=a.dtype)
+    num_sms = torch.cuda.get_device_properties(a.device).multi_processor_count
+    BLOCK_SIZE_M, BLOCK_SIZE_N, BLOCK_SIZE_K = 128, 256, 64
+    num_stages = 4 if forward else 3
+    a_desc = TensorDescriptor.from_tensor(a, [BLOCK_SIZE_M, BLOCK_SIZE_K])
+    b_desc = TensorDescriptor.from_tensor(b, [BLOCK_SIZE_N, BLOCK_SIZE_K])
+    c_desc = TensorDescriptor.from_tensor(c, [BLOCK_SIZE_M, BLOCK_SIZE_N // 2])
+    aux_desc = TensorDescriptor.from_tensor(aux, [BLOCK_SIZE_M, BLOCK_SIZE_N // 2])
+    grid = lambda _meta: (
+        min(num_sms, triton.cdiv(M, BLOCK_SIZE_M) * triton.cdiv(N, BLOCK_SIZE_N)),
+    )
+    linear_leaky_relu_square_kernel[grid](
+        a_desc,
+        b_desc,
+        c_desc,
+        aux_desc,
+        M,
+        N,
+        K,
+        BLOCK_SIZE_M=BLOCK_SIZE_M,
+        BLOCK_SIZE_N=BLOCK_SIZE_N,
+        BLOCK_SIZE_K=BLOCK_SIZE_K,
+        NUM_SMS=num_sms,
+        FORWARD=forward,
+        num_stages=num_stages,
+        num_warps=8,
+    )
+    if forward:
+        return c, aux
+    return c
+class FusedLinearLeakyReLUSquareFunction(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, x, w1, w2):
+        x_flat = x.reshape(-1, x.shape[-1])
+        pre, post = linear_leaky_relu_square(x_flat, w1)
+        out = F.linear(post, w2)
+        ctx.save_for_backward(x, w1, w2, pre, post)
+        return out.view(*x.shape[:-1], out.shape[-1])
+    @staticmethod
+    def backward(ctx, grad_output):
+        x, w1, w2, pre, post = ctx.saved_tensors
+        x_flat = x.reshape(-1, x.shape[-1])
+        grad_output_flat = grad_output.reshape(-1, grad_output.shape[-1])
+        dw2 = grad_output_flat.T @ post
+        dpre = linear_leaky_relu_square(grad_output_flat, w2.T.contiguous(), aux=pre)
+        dw1 = dpre.T @ x_flat
+        dx = dpre @ w1
+        return dx.view_as(x), dw1, dw2
+FusedLeakyReLUSquareMLP = FusedLinearLeakyReLUSquareFunction.apply
+class Rotary(nn.Module):
+    def __init__(self, dim, base=1e4, train_seq_len=1024, rope_dims=0, yarn=True):
+        super().__init__()
+        self.dim = dim
+        self.base = base
+        self.train_seq_len = train_seq_len
+        self.yarn = yarn
+        self.rope_dims = rope_dims if rope_dims > 0 else dim
+        inv_freq = 1.0 / base ** (
+            torch.arange(0, self.rope_dims, 2, dtype=torch.float32) / self.rope_dims
+        )
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self._seq_len_cached = 0
+        self._cos_cached = None
+        self._sin_cached = None
+    def forward(self, seq_len, device, dtype):
+        if (
+            self._cos_cached is None
+            or self._sin_cached is None
+            or self._seq_len_cached < seq_len
+            or self._cos_cached.device != device
+        ):
+            rd = self.rope_dims
+            if self.yarn and seq_len > self.train_seq_len:
+                scale = seq_len / self.train_seq_len
+                new_base = self.base * scale ** (rd / (rd - 2))
+                inv_freq = 1.0 / new_base ** (
+                    torch.arange(0, rd, 2, dtype=torch.float32, device=device) / rd
+                )
+            else:
+                inv_freq = self.inv_freq.float().to(device)
+            t = torch.arange(seq_len, device=device, dtype=torch.float32)
+            freqs = torch.outer(t, inv_freq)
+            self._cos_cached = freqs.cos()[None, :, None, :]
+            self._sin_cached = freqs.sin()[None, :, None, :]
+            self._seq_len_cached = seq_len
+        return self._cos_cached[:, :seq_len].to(dtype=dtype), self._sin_cached[:, :seq_len].to(dtype=dtype)
+def apply_rotary_emb(x, cos, sin, rope_dims=0):
+    if rope_dims > 0 and rope_dims < x.size(-1):
+        x_rope, x_pass = x[..., :rope_dims], x[..., rope_dims:]
+        half = rope_dims // 2
+        x1, x2 = x_rope[..., :half], x_rope[..., half:]
+        x_rope = torch.cat((x1 * cos + x2 * sin, x1 * -sin + x2 * cos), dim=-1)
+        return torch.cat((x_rope, x_pass), dim=-1)
+    half = x.size(-1) // 2
+    x1, x2 = x[..., :half], x[..., half:]
+    return torch.cat((x1 * cos + x2 * sin, x1 * -sin + x2 * cos), dim=-1)
+class CausalSelfAttention(nn.Module):
+    def __init__(
+        self, dim, num_heads, num_kv_heads, rope_base, qk_gain_init, train_seq_len, yarn=True,
+        attn_out_gate=False, attn_out_gate_src="proj", gate_window=12,
+        gated_attn=False, gated_attn_init_std=0.01,
+        sparse_attn_gate=False, sparse_attn_gate_init_std=0.0, sparse_attn_gate_scale=1.0,
+    ):
+        super().__init__()
+        if dim % num_heads != 0:
+            raise ValueError("model_dim must be divisible by num_heads")
+        if num_heads % num_kv_heads != 0:
+            raise ValueError("num_heads must be divisible by num_kv_heads")
+        if int(attn_out_gate) + int(gated_attn) + int(sparse_attn_gate) > 1:
+            raise ValueError(
+                "attn_out_gate, gated_attn, and sparse_attn_gate are mutually exclusive"
+            )
+        self.num_heads = num_heads
+        self.num_kv_heads = num_kv_heads
+        self.head_dim = dim // num_heads
+        if self.head_dim % 2 != 0:
+            raise ValueError("head_dim must be even for RoPE")
+        self.q_gain = nn.Parameter(
+            torch.full((num_heads,), qk_gain_init, dtype=torch.float32)
+        )
+        self.rope_dims = 0
+        self.rotary = Rotary(self.head_dim, base=rope_base, train_seq_len=train_seq_len, yarn=yarn)
+        self.use_xsa = False
+        self.attn_out_gate = attn_out_gate
+        self.attn_out_gate_src = attn_out_gate_src
+        self.gate_window = gate_window
+        if attn_out_gate:
+            self.attn_gate_proj = CastedLinear(gate_window, num_heads, bias=False)
+            self.attn_gate_proj._zero_init = True
+        self.gated_attn = gated_attn
+        if gated_attn:
+            W = torch.empty(num_heads, dim, dtype=torch.float32)
+            nn.init.normal_(W, mean=0.0, std=gated_attn_init_std)
+            self.attn_gate_w = nn.Parameter(W)
+        self.sparse_attn_gate = sparse_attn_gate
+        self.sparse_attn_gate_scale = sparse_attn_gate_scale
+        if sparse_attn_gate:
+            W = torch.empty(num_heads, gate_window, dtype=torch.float32)
+            if sparse_attn_gate_init_std > 0:
+                nn.init.normal_(W, mean=0.0, std=sparse_attn_gate_init_std)
+            else:
+                nn.init.zeros_(W)
+            self.attn_gate_w = nn.Parameter(W)
+    def _xsa_efficient(self, y, v):
+        B, T, H, D = y.shape
+        Hkv = v.size(-2)
+        group = H // Hkv
+        y_g = y.reshape(B, T, Hkv, group, D)
+        vn = F.normalize(v, dim=-1).unsqueeze(-2)
+        proj = (y_g * vn).sum(dim=-1, keepdim=True) * vn
+        return (y_g - proj).reshape(B, T, H, D)
+    def forward(self, x, q_w, k_w, v_w, out_w, cu_seqlens=None, max_seqlen=0):
+        bsz, seqlen, dim = x.shape
+        q_raw = F.linear(x, q_w.to(x.dtype))
+        q = q_raw.reshape(bsz, seqlen, self.num_heads, self.head_dim)
+        k = F.linear(x, k_w.to(x.dtype)).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim)
+        v = F.linear(x, v_w.to(x.dtype)).reshape(bsz, seqlen, self.num_kv_heads, self.head_dim)
+        q = F.rms_norm(q, (q.size(-1),))
+        k = F.rms_norm(k, (k.size(-1),))
+        cos, sin = self.rotary(seqlen, x.device, q.dtype)
+        q = apply_rotary_emb(q, cos, sin, self.rope_dims)
+        k = apply_rotary_emb(k, cos, sin, self.rope_dims)
+        q = q * self.q_gain.to(dtype=q.dtype)[None, None, :, None]
+        if cu_seqlens is not None:
+            y = flash_attn_varlen_func(
+                q[0],
+                k[0],
+                v[0],
+                cu_seqlens_q=cu_seqlens,
+                cu_seqlens_k=cu_seqlens,
+                max_seqlen_q=max_seqlen,
+                max_seqlen_k=max_seqlen,
+                causal=True,
+                window_size=(-1, -1),
+            )[None]
+        else:
+            y = flash_attn_3_func(q, k, v, causal=True)
+        if self.use_xsa:
+            y = self._xsa_efficient(y, v)
+        if self.attn_out_gate:
+            gate_src = q_raw if self.attn_out_gate_src == "q" else x
+            gate_in = gate_src[..., : self.gate_window].contiguous()
+            g = 2.0 * torch.sigmoid(self.attn_gate_proj(gate_in))
+            y = y * g[..., None]
+        if self.gated_attn:
+            x_c = x.contiguous()
+            g = torch.sigmoid(F.linear(x_c, self.attn_gate_w.to(x.dtype)))
+            y = y * g[..., None]
+        if self.sparse_attn_gate:
+            gate_in = x[..., : self.gate_window].contiguous()
+            g = torch.sigmoid(
+                self.sparse_attn_gate_scale
+                * F.linear(gate_in, self.attn_gate_w.to(x.dtype))
+            )
+            y = y * g[..., None]
+        y = y.reshape(bsz, seqlen, dim)
+        self._last_proj_input = y.detach() if getattr(self, "_calib", False) else None
+        return F.linear(y, out_w.to(x.dtype))
+class MLP(nn.Module):
+    def __init__(self, dim, mlp_mult):
+        super().__init__()
+        self.use_fused = True
+    def forward(self, x, up_w, down_w):
+        if self.training and self.use_fused:
+            return FusedLeakyReLUSquareMLP(x, up_w.to(x.dtype), down_w.to(x.dtype))
+        hidden = F.leaky_relu(F.linear(x, up_w.to(x.dtype)), negative_slope=0.3).square()
+        self._last_down_input = hidden.detach() if getattr(self, "_calib", False) else None
+        return F.linear(hidden, down_w.to(x.dtype))
+class Block(nn.Module):
+    def __init__(
+        self,
+        dim,
+        num_heads,
+        num_kv_heads,
+        mlp_mult,
+        rope_base,
+        qk_gain_init,
+        train_seq_len,
+        layer_idx=0,
+        ln_scale=False,
+        yarn=True,
+        attn_out_gate=False,
+        attn_out_gate_src="proj",
+        gate_window=12,
+        gated_attn=False,
+        gated_attn_init_std=0.01,
+        sparse_attn_gate=False,
+        sparse_attn_gate_init_std=0.0,
+        sparse_attn_gate_scale=1.0,
+    ):
+        super().__init__()
+        self.attn_norm = RMSNorm()
+        self.mlp_norm = RMSNorm()
+        self.attn = CausalSelfAttention(
+            dim, num_heads, num_kv_heads, rope_base, qk_gain_init, train_seq_len, yarn=yarn,
+            attn_out_gate=attn_out_gate, attn_out_gate_src=attn_out_gate_src, gate_window=gate_window,
+            gated_attn=gated_attn, gated_attn_init_std=gated_attn_init_std,
+            sparse_attn_gate=sparse_attn_gate,
+            sparse_attn_gate_init_std=sparse_attn_gate_init_std,
+            sparse_attn_gate_scale=sparse_attn_gate_scale,
+        )
+        self.mlp = MLP(dim, mlp_mult)
+        self.attn_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32))
+        self.mlp_scale = nn.Parameter(torch.ones(dim, dtype=torch.float32))
+        self.resid_mix = nn.Parameter(
+            torch.stack((torch.ones(dim), torch.zeros(dim))).float()
+        )
+        self.ln_scale_factor = 1.0 / math.sqrt(layer_idx + 1) if ln_scale else 1.0
+    def forward(self, x, x0, q_w, k_w, v_w, out_w, up_w, down_w, cu_seqlens=None, max_seqlen=0):
+        mix = self.resid_mix.to(dtype=x.dtype)
+        x_in = mix[0][None, None, :] * x + mix[1][None, None, :] * x0
+        attn_out = self.attn(
+            self.attn_norm(x_in) * self.ln_scale_factor,
+            q_w, k_w, v_w, out_w,
+            cu_seqlens=cu_seqlens,
+            max_seqlen=max_seqlen,
+        )
+        x_out = x_in + self.attn_scale.to(dtype=x_in.dtype)[None, None, :] * attn_out
+        x_out = x_out + self.mlp_scale.to(dtype=x_out.dtype)[
+            None, None, :
+        ] * self.mlp(self.mlp_norm(x_out) * self.ln_scale_factor, up_w, down_w)
+        return x_out
+class GPT(nn.Module):
+    def __init__(self, h):
+        super().__init__()
+        if h.logit_softcap <= 0.0:
+            raise ValueError(f"logit_softcap must be positive, got {h.logit_softcap}")
+        self.tied_embed_init_std = h.tied_embed_init_std
+        self.logit_softcap = h.logit_softcap
+        self.fused_ce_enabled = bool(h.fused_ce_enabled)
+        self.tok_emb = nn.Embedding(h.vocab_size, h.model_dim)
+        self.num_layers = h.num_layers
+        head_dim = h.model_dim // h.num_heads
+        kv_dim = h.num_kv_heads * head_dim
+        hidden_dim = int(h.mlp_mult * h.model_dim)
+        self.qo_bank = nn.Parameter(torch.empty(2 * h.num_layers, h.model_dim, h.model_dim))
+        self.kv_bank = nn.Parameter(torch.empty(2 * h.num_layers, kv_dim, h.model_dim))
+        self.mlp_up_bank = nn.Parameter(torch.empty(h.num_layers, hidden_dim, h.model_dim))
+        self.mlp_down_bank = nn.Parameter(torch.empty(h.num_layers, h.model_dim, hidden_dim))
+        self.num_encoder_layers = h.num_layers // 2
+        self.num_decoder_layers = h.num_layers - self.num_encoder_layers
+        self.blocks = nn.ModuleList(
+            [
+                Block(
+                    h.model_dim,
+                    h.num_heads,
+                    h.num_kv_heads,
+                    h.mlp_mult,
+                    h.rope_base,
+                    h.qk_gain_init,
+                    h.train_seq_len,
+                    layer_idx=i,
+                    ln_scale=h.ln_scale,
+                    yarn=h.rope_yarn,
+                    attn_out_gate=h.attn_out_gate_enabled,
+                    attn_out_gate_src=h.attn_out_gate_src,
+                    gate_window=h.gate_window,
+                    gated_attn=h.gated_attn_enabled,
+                    gated_attn_init_std=h.gated_attn_init_std,
+                    sparse_attn_gate=h.sparse_attn_gate_enabled,
+                    sparse_attn_gate_init_std=h.sparse_attn_gate_init_std,
+                    sparse_attn_gate_scale=h.sparse_attn_gate_scale,
+                )
+                for i in range(h.num_layers)
+            ]
+        )
+        if h.rope_dims > 0:
+            head_dim = h.model_dim // h.num_heads
+            for block in self.blocks:
+                block.attn.rope_dims = h.rope_dims
+                block.attn.rotary = Rotary(
+                    head_dim,
+                    base=h.rope_base,
+                    train_seq_len=h.train_seq_len,
+                    rope_dims=h.rope_dims,
+                    yarn=h.rope_yarn,
+                )
+        self.final_norm = RMSNorm()
+        if h.xsa_last_n > 0:
+            for i in range(max(0, h.num_layers - h.xsa_last_n), h.num_layers):
+                self.blocks[i].attn.use_xsa = True
+        self.looping_active = False
+        if h.num_loops > 0:
+            loop_seg = list(range(h.loop_start, h.loop_end + 1))
+            all_indices = list(range(h.loop_start))
+            for _ in range(h.num_loops + 1):
+                all_indices.extend(loop_seg)
+            all_indices.extend(range(h.loop_end + 1, h.num_layers))
+            num_enc = len(all_indices) // 2
+            self.encoder_indices = all_indices[:num_enc]
+            self.decoder_indices = all_indices[num_enc:]
+        else:
+            self.encoder_indices = list(range(self.num_encoder_layers))
+            self.decoder_indices = list(range(self.num_encoder_layers, h.num_layers))
+        self.num_skip_weights = min(
+            len(self.encoder_indices), len(self.decoder_indices)
+        )
+        self.skip_weights = nn.Parameter(
+            torch.ones(self.num_skip_weights, h.model_dim, dtype=torch.float32)
+        )
+        self.skip_gates = (
+            nn.Parameter(
+                torch.zeros(self.num_skip_weights, h.model_dim, dtype=torch.float32)
+            )
+            if h.skip_gates_enabled
+            else None
+        )
+        self.parallel_start_layer = h.parallel_start_layer
+        self.parallel_final_lane = h.parallel_final_lane.lower()
+        self.parallel_post_lambdas = nn.Parameter(
+            torch.ones(h.num_layers, 2, 2, dtype=torch.float32)
+        )
+        self.parallel_resid_lambdas = nn.Parameter(
+            torch.full((h.num_layers, 2), 1.1, dtype=torch.float32)
+        )
+        self.smear_gate_enabled = h.smear_gate_enabled
+        if self.smear_gate_enabled:
+            self.smear_window = h.gate_window
+            self.smear_gate = CastedLinear(self.smear_window, 1, bias=False)
+            self.smear_gate._zero_init = True
+            self.smear_lambda = nn.Parameter(torch.zeros(1, dtype=torch.float32))
+        self._init_weights()
+    def _init_weights(self):
+        nn.init.normal_(self.tok_emb.weight, mean=0.0, std=self.tied_embed_init_std)
+        n = self.num_layers
+        proj_scale = 1.0 / math.sqrt(2 * n)
+        for i in range(n):
+            nn.init.orthogonal_(self.qo_bank.data[i], gain=1.0)
+            nn.init.zeros_(self.qo_bank.data[n + i])
+            self.qo_bank.data[n + i].mul_(proj_scale)
+            nn.init.orthogonal_(self.kv_bank.data[i], gain=1.0)
+            nn.init.orthogonal_(self.kv_bank.data[n + i], gain=1.0)
+        for i in range(n):
+            nn.init.orthogonal_(self.mlp_up_bank.data[i], gain=1.0)
+            nn.init.zeros_(self.mlp_down_bank.data[i])
+            self.mlp_down_bank.data[i].mul_(proj_scale)
+        for name, module in self.named_modules():
+            if isinstance(module, nn.Linear):
+                if getattr(module, "_zero_init", False):
+                    nn.init.zeros_(module.weight)
+                elif (
+                    module.weight.ndim == 2
+                    and module.weight.shape[0] >= 64
+                    and module.weight.shape[1] >= 64
+                ):
+                    nn.init.orthogonal_(module.weight, gain=1.0)
+    def _bank_weights(self, i):
+        n = self.num_layers
+        return (
+            self.qo_bank[i],
+            self.kv_bank[i],
+            self.kv_bank[n + i],
+            self.qo_bank[n + i],
+            self.mlp_up_bank[i],
+            self.mlp_down_bank[i],
+        )
+    def _parallel_block(
+        self, block_idx, lane0, lane1, x0,
+        q_w, k_w, v_w, out_w, up_w, down_w,
+        cu_seqlens=None, max_seqlen=0,
+    ):
+        block = self.blocks[block_idx]
+        mix = block.resid_mix.to(dtype=lane0.dtype)
+        attn_read = mix[0][None, None, :] * lane0 + mix[1][None, None, :] * x0
+        attn_out = block.attn(
+            block.attn_norm(attn_read) * block.ln_scale_factor,
+            q_w, k_w, v_w, out_w,
+            cu_seqlens=cu_seqlens, max_seqlen=max_seqlen,
+        )
+        attn_out = block.attn_scale.to(dtype=attn_out.dtype)[None, None, :] * attn_out
+        mlp_read = lane1
+        mlp_out = block.mlp_scale.to(dtype=lane1.dtype)[None, None, :] * block.mlp(
+            block.mlp_norm(mlp_read) * block.ln_scale_factor, up_w, down_w
+        )
+        attn_resid = self.parallel_resid_lambdas[block_idx, 0].to(dtype=lane0.dtype)
+        attn_post = self.parallel_post_lambdas[block_idx, 0].to(dtype=lane0.dtype)
+        mlp_resid = self.parallel_resid_lambdas[block_idx, 1].to(dtype=lane0.dtype)
+        mlp_post = self.parallel_post_lambdas[block_idx, 1].to(dtype=lane0.dtype)
+        lane0 = attn_resid * lane0 + attn_post[0] * attn_out + mlp_post[0] * mlp_out
+        lane1 = mlp_resid * lane1 + attn_post[1] * attn_out + mlp_post[1] * mlp_out
+        return lane0, lane1
+    def _final_parallel_hidden(self, lane0, lane1):
+        if self.parallel_final_lane == "mlp":
+            return lane1
+        if self.parallel_final_lane == "attn":
+            return lane0
+        return 0.5 * (lane0 + lane1)
+    def _forward_hidden(self, input_ids, cu_seqlens=None, max_seqlen=0):
+        x = self.tok_emb(input_ids)
+        if self.smear_gate_enabled:
+            sl = self.smear_lambda.to(dtype=x.dtype)
+            gate_in = x[:, 1:, : self.smear_window].contiguous()
+            g = sl * torch.sigmoid(self.smear_gate(gate_in))
+            bos_mask = (input_ids[:, 1:] == 1).unsqueeze(-1)
+            g = g.masked_fill(bos_mask, 0.0)
+            x = torch.cat([x[:, :1], x[:, 1:] + g * x[:, :-1]], dim=1)
+        x = F.rms_norm(x, (x.size(-1),))
+        x0 = x
+        skips = []
+        enc_iter = (
+            self.encoder_indices
+            if self.looping_active
+            else range(self.num_encoder_layers)
+        )
+        dec_iter = (
+            self.decoder_indices
+            if self.looping_active
+            else range(
+                self.num_encoder_layers,
+                self.num_encoder_layers + self.num_decoder_layers,
+            )
+        )
+        for i in enc_iter:
+            q_w, k_w, v_w, out_w, up_w, down_w = self._bank_weights(i)
+            x = self.blocks[i](x, x0, q_w, k_w, v_w, out_w, up_w, down_w, cu_seqlens=cu_seqlens, max_seqlen=max_seqlen)
+            skips.append(x)
+        psl = self.parallel_start_layer
+        lane0 = None
+        lane1 = None
+        for skip_idx, i in enumerate(dec_iter):
+            q_w, k_w, v_w, out_w, up_w, down_w = self._bank_weights(i)
+            if i >= psl and psl > 0:
+                if lane0 is None:
+                    lane0 = x
+                    lane1 = x
+                if skip_idx < self.num_skip_weights and skips:
+                    skip = skips.pop()
+                    w = self.skip_weights[skip_idx].to(dtype=lane0.dtype)[None, None, :]
+                    if self.skip_gates is not None:
+                        g = torch.sigmoid(self.skip_gates[skip_idx].to(dtype=lane0.dtype))[None, None, :]
+                        lane0 = torch.lerp(w * skip, lane0, g)
+                    else:
+                        lane0 = lane0 + w * skip
+                lane0, lane1 = self._parallel_block(
+                    i, lane0, lane1, x0, q_w, k_w, v_w, out_w, up_w, down_w,
+                    cu_seqlens=cu_seqlens, max_seqlen=max_seqlen,
+                )
+            else:
+                if skip_idx < self.num_skip_weights and skips:
+                    scaled_skip = (
+                        self.skip_weights[skip_idx].to(dtype=x.dtype)[None, None, :]
+                        * skips.pop()
+                    )
+                    if self.skip_gates is not None:
+                        g = torch.sigmoid(self.skip_gates[skip_idx].to(dtype=x.dtype))[None, None, :]
+                        x = torch.lerp(scaled_skip, x, g)
+                    else:
+                        x = x + scaled_skip
+                x = self.blocks[i](x, x0, q_w, k_w, v_w, out_w, up_w, down_w, cu_seqlens=cu_seqlens, max_seqlen=max_seqlen)
+        if lane0 is not None:
+            x = self._final_parallel_hidden(lane0, lane1)
+        x = self.final_norm(x)
+        return x
+    def _project_logits(self, hidden):
+        return F.linear(hidden, self.tok_emb.weight)
+    def forward_logits(self, input_ids, cu_seqlens=None, max_seqlen=0):
+        hidden = self._forward_hidden(input_ids, cu_seqlens=cu_seqlens, max_seqlen=max_seqlen)
+        logits_proj = self._project_logits(hidden)
+        return self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap)
+    def forward(self, input_ids, target_ids, cu_seqlens=None, max_seqlen=0):
+        hidden = self._forward_hidden(input_ids, cu_seqlens=cu_seqlens, max_seqlen=max_seqlen)
+        logits_proj = self._project_logits(hidden)
+        flat_targets = target_ids.reshape(-1)
+        if self.fused_ce_enabled:
+            return softcapped_cross_entropy(
+                logits_proj.reshape(-1, logits_proj.size(-1)),
+                flat_targets,
+                self.logit_softcap,
+                reduction="mean",
+            )
+        logits = self.logit_softcap * torch.tanh(logits_proj / self.logit_softcap)
+        return F.cross_entropy(
+            logits.reshape(-1, logits.size(-1)).float(),
+            flat_targets,
+            reduction="mean",
+        )
+    def forward_ttt(self, input_ids, target_ids, lora):
+        x = self.tok_emb(input_ids)
+        if self.smear_gate_enabled:
+            sl = self.smear_lambda.to(dtype=x.dtype)
+            gate_in = x[:, 1:, : self.smear_window].contiguous()
+            g = sl * torch.sigmoid(self.smear_gate(gate_in))
+            bos_mask = (input_ids[:, 1:] == 1).unsqueeze(-1)
+            g = g.masked_fill(bos_mask, 0.0)
+            x = torch.cat([x[:, :1], x[:, 1:] + g * x[:, :-1]], dim=1)
+        x = F.rms_norm(x, (x.size(-1),))
+        x0 = x
+        skips = []
+        enc_iter = (
+            self.encoder_indices
+            if self.looping_active
+            else list(range(self.num_encoder_layers))
+        )
+        dec_iter = (
+            self.decoder_indices
+            if self.looping_active
+            else list(
+                range(
+                    self.num_encoder_layers,
+                    self.num_encoder_layers + self.num_decoder_layers,
+                )
+            )
+        )
+        slot = 0
+        for i in enc_iter:
+            q_w, k_w, v_w, out_w, up_w, down_w = self._bank_weights(i)
+            x = self._block_with_lora(self.blocks[i], x, x0, lora, slot, q_w, k_w, v_w, out_w, up_w, down_w)
+            slot += 1
+            skips.append(x)
+        psl = self.parallel_start_layer
+        lane0 = None
+        lane1 = None
+        for skip_idx, i in enumerate(dec_iter):
+            q_w, k_w, v_w, out_w, up_w, down_w = self._bank_weights(i)
+            if i >= psl and psl > 0:
+                if lane0 is None:
+                    lane0 = x
+                    lane1 = x
+                if skip_idx < self.num_skip_weights and skips:
+                    skip = skips.pop()
+                    w = self.skip_weights[skip_idx].to(dtype=lane0.dtype)[None, None, :]
+                    if self.skip_gates is not None:
+                        g = torch.sigmoid(self.skip_gates[skip_idx].to(dtype=lane0.dtype))[None, None, :]
+                        lane0 = torch.lerp(w * skip, lane0, g)
+                    else:
+                        lane0 = lane0 + w * skip
+                lane0, lane1 = self._parallel_block_with_lora(
+                    i, lane0, lane1, x0, lora, slot,
+                    q_w, k_w, v_w, out_w, up_w, down_w,
+                )
+            else:
+                if skip_idx < self.num_skip_weights and skips:
+                    scaled_skip = (
+                        self.skip_weights[skip_idx].to(dtype=x.dtype)[None, None, :]
+                        * skips.pop()
+                    )
+                    if self.skip_gates is not None:
+                        g = torch.sigmoid(self.skip_gates[skip_idx].to(dtype=x.dtype))[None, None, :]
+                        x = torch.lerp(scaled_skip, x, g)
+                    else:
+                        x = x + scaled_skip
+                x = self._block_with_lora(self.blocks[i], x, x0, lora, slot, q_w, k_w, v_w, out_w, up_w, down_w)
+            slot += 1
+        if lane0 is not None:
+            x = self._final_parallel_hidden(lane0, lane1)
+        x = self.final_norm(x)
+        logits = F.linear(x, self.tok_emb.weight) + lora.lm_head_lora(x)
+        logits = self.logit_softcap * torch.tanh(logits / self.logit_softcap)
+        bsz, sl, V = logits.shape
+        return F.cross_entropy(
+            logits.float().reshape(-1, V), target_ids.reshape(-1), reduction="none"
+        ).reshape(bsz, sl)
+    def _block_with_lora(self, block, x, x0, lora, slot, q_w, k_w, v_w, out_w, up_w, down_w):
+        mix = block.resid_mix.to(dtype=x.dtype)
+        x_in = mix[0][None, None, :] * x + mix[1][None, None, :] * x0
+        n = block.attn_norm(x_in) * block.ln_scale_factor
+        attn = block.attn
+        bsz, seqlen, dim = n.shape
+        q_raw = F.linear(n, q_w.to(n.dtype)) + lora.q_loras[slot](n)
+        q = q_raw.reshape(bsz, seqlen, attn.num_heads, attn.head_dim)
+        k = F.linear(n, k_w.to(n.dtype))
+        if lora.k_loras is not None:
+            k = k + lora.k_loras[slot](n)
+        k = k.reshape(bsz, seqlen, attn.num_kv_heads, attn.head_dim)
+        v = (F.linear(n, v_w.to(n.dtype)) + lora.v_loras[slot](n)).reshape(
+            bsz, seqlen, attn.num_kv_heads, attn.head_dim
+        )
+        q = F.rms_norm(q, (q.size(-1),))
+        k = F.rms_norm(k, (k.size(-1),))
+        cos, sin = attn.rotary(seqlen, n.device, q.dtype)
+        q = apply_rotary_emb(q, cos, sin, attn.rope_dims)
+        k = apply_rotary_emb(k, cos, sin, attn.rope_dims)
+        q = q * attn.q_gain.to(dtype=q.dtype)[None, None, :, None]
+        y = flash_attn_3_func(q, k, v, causal=True)
+        if attn.use_xsa:
+            y = attn._xsa_efficient(y, v)
+        if attn.attn_out_gate:
+            gate_src = q_raw if attn.attn_out_gate_src == "q" else n
+            gate_in = gate_src[..., : attn.gate_window].contiguous()
+            g = 2.0 * torch.sigmoid(attn.attn_gate_proj(gate_in))
+            y = y * g[..., None]
+        if attn.gated_attn:
+            n_c = n.contiguous()
+            g = torch.sigmoid(F.linear(n_c, attn.attn_gate_w.to(n.dtype)))
+            y = y * g[..., None]
+        if attn.sparse_attn_gate:
+            gate_in = n[..., : attn.gate_window].contiguous()
+            g = torch.sigmoid(
+                attn.sparse_attn_gate_scale
+                * F.linear(gate_in, attn.attn_gate_w.to(n.dtype))
+            )
+            y = y * g[..., None]
+        y = y.reshape(bsz, seqlen, dim)
+        attn_out = F.linear(y, out_w.to(n.dtype))
+        if lora.o_loras is not None:
+            attn_out = attn_out + lora.o_loras[slot](n)
+        x_out = x_in + block.attn_scale.to(dtype=x_in.dtype)[None, None, :] * attn_out
+        mlp_n = block.mlp_norm(x_out) * block.ln_scale_factor
+        mlp_out = block.mlp(mlp_n, up_w, down_w)
+        if lora.mlp_loras is not None:
+            mlp_out = mlp_out + lora.mlp_loras[slot](mlp_n)
+        x_out = x_out + block.mlp_scale.to(dtype=x_out.dtype)[None, None, :] * mlp_out
+        return x_out
+    def _parallel_block_with_lora(
+        self, block_idx, lane0, lane1, x0, lora, slot,
+        q_w, k_w, v_w, out_w, up_w, down_w,
+    ):
+        block = self.blocks[block_idx]
+        mix = block.resid_mix.to(dtype=lane0.dtype)
+        attn_read = mix[0][None, None, :] * lane0 + mix[1][None, None, :] * x0
+        n = block.attn_norm(attn_read) * block.ln_scale_factor
+        attn = block.attn
+        bsz, seqlen, dim = n.shape
+        q_raw = F.linear(n, q_w.to(n.dtype)) + lora.q_loras[slot](n)
+        q = q_raw.reshape(bsz, seqlen, attn.num_heads, attn.head_dim)
+        k = F.linear(n, k_w.to(n.dtype))
+        if lora.k_loras is not None:
+            k = k + lora.k_loras[slot](n)
+        k = k.reshape(bsz, seqlen, attn.num_kv_heads, attn.head_dim)
+        v = (F.linear(n, v_w.to(n.dtype)) + lora.v_loras[slot](n)).reshape(
+            bsz, seqlen, attn.num_kv_heads, attn.head_dim
+        )
+        q = F.rms_norm(q, (q.size(-1),))
+        k = F.rms_norm(k, (k.size(-1),))
+        cos, sin = attn.rotary(seqlen, n.device, q.dtype)
+        q = apply_rotary_emb(q, cos, sin, attn.rope_dims)
+        k = apply_rotary_emb(k, cos, sin, attn.rope_dims)
+        q = q * attn.q_gain.to(dtype=q.dtype)[None, None, :, None]
+        y = flash_attn_3_func(q, k, v, causal=True)
+        if attn.use_xsa:
+            y = attn._xsa_efficient(y, v)
+        if attn.attn_out_gate:
+            gate_src = q_raw if attn.attn_out_gate_src == "q" else n
+            gate_in = gate_src[..., : attn.gate_window].contiguous()
+            g = 2.0 * torch.sigmoid(attn.attn_gate_proj(gate_in))
+            y = y * g[..., None]
+        if attn.gated_attn:
+            n_c = n.contiguous()
+            g = torch.sigmoid(F.linear(n_c, attn.attn_gate_w.to(n.dtype)))
+            y = y * g[..., None]
+        if attn.sparse_attn_gate:
+            gate_in = n[..., : attn.gate_window].contiguous()
+            g = torch.sigmoid(
+                attn.sparse_attn_gate_scale
+                * F.linear(gate_in, attn.attn_gate_w.to(n.dtype))
+            )
+            y = y * g[..., None]
+        y = y.reshape(bsz, seqlen, dim)
+        attn_out = F.linear(y, out_w.to(n.dtype))
+        if lora.o_loras is not None:
+            attn_out = attn_out + lora.o_loras[slot](n)
+        attn_out = block.attn_scale.to(dtype=attn_out.dtype)[None, None, :] * attn_out
+        mlp_read = lane1
+        mlp_n = block.mlp_norm(mlp_read) * block.ln_scale_factor
+        mlp_out = block.mlp(mlp_n, up_w, down_w)
+        if lora.mlp_loras is not None:
+            mlp_out = mlp_out + lora.mlp_loras[slot](mlp_n)
+        mlp_out = block.mlp_scale.to(dtype=lane1.dtype)[None, None, :] * mlp_out
+        attn_resid = self.parallel_resid_lambdas[block_idx, 0].to(dtype=lane0.dtype)
+        attn_post = self.parallel_post_lambdas[block_idx, 0].to(dtype=lane0.dtype)
+        mlp_resid = self.parallel_resid_lambdas[block_idx, 1].to(dtype=lane0.dtype)
+        mlp_post = self.parallel_post_lambdas[block_idx, 1].to(dtype=lane0.dtype)
+        lane0 = attn_resid * lane0 + attn_post[0] * attn_out + mlp_post[0] * mlp_out
+        lane1 = mlp_resid * lane1 + attn_post[1] * attn_out + mlp_post[1] * mlp_out
+        return lane0, lane1
+class BatchedLinearLoRA(nn.Module):
+    _ALPHA = float(os.environ.get("TTT_LORA_ALPHA", "144"))
+    _WARM_START_A = bool(int(os.environ.get("TTT_WARM_START_A", "1")))
+    def __init__(self, bsz, in_features, out_features, rank):
+        super().__init__()
+        self._bound = 1.0 / math.sqrt(in_features)
+        self._scale = self._ALPHA / rank
+        self.A = nn.Parameter(
+            torch.empty(bsz, rank, in_features).uniform_(-self._bound, self._bound)
+        )
+        self.B = nn.Parameter(torch.zeros(bsz, out_features, rank))
+    def reset(self):
+        with torch.no_grad():
+            if not self._WARM_START_A:
+                self.A.uniform_(-self._bound, self._bound)
+            self.B.zero_()
+    def forward(self, x):
+        return ((x @ self.A.transpose(1, 2)) @ self.B.transpose(1, 2)) * self._scale
+class BatchedTTTLoRA(nn.Module):
+    def __init__(self, bsz, model, rank, k_lora=True, mlp_lora=True, o_lora=True):
+        super().__init__()
+        self.bsz = bsz
+        dim = model.qo_bank.shape[-1]
+        vocab = model.tok_emb.num_embeddings
+        if getattr(model, "looping_active", False):
+            num_slots = len(model.encoder_indices) + len(model.decoder_indices)
+        else:
+            num_slots = len(model.blocks)
+        kv_dim = model.blocks[0].attn.num_kv_heads * (
+            dim // model.blocks[0].attn.num_heads
+        )
+        embed_dim = model.tok_emb.embedding_dim
+        self.lm_head_lora = BatchedLinearLoRA(bsz, embed_dim, vocab, rank)
+        self.q_loras = nn.ModuleList(
+            [BatchedLinearLoRA(bsz, dim, dim, rank) for _ in range(num_slots)]
+        )
+        self.v_loras = nn.ModuleList(
+            [BatchedLinearLoRA(bsz, dim, kv_dim, rank) for _ in range(num_slots)]
+        )
+        self.k_loras = (
+            nn.ModuleList(
+                [BatchedLinearLoRA(bsz, dim, kv_dim, rank) for _ in range(num_slots)]
+            )
+            if k_lora
+            else None
+        )
+        self.mlp_loras = (
+            nn.ModuleList(
+                [BatchedLinearLoRA(bsz, dim, dim, rank) for _ in range(num_slots)]
+            )
+            if mlp_lora
+            else None
+        )
+        self.o_loras = (
+            nn.ModuleList(
+                [BatchedLinearLoRA(bsz, dim, dim, rank) for _ in range(num_slots)]
+            )
+            if o_lora
+            else None
+        )
+    def reset(self):
+        with torch.no_grad():
+            self.lm_head_lora.reset()
+            for loras in [self.q_loras, self.v_loras, self.k_loras,
+                          self.mlp_loras, self.o_loras]:
+                if loras is not None:
+                    for lora in loras:
+                        lora.reset()
+_PE_COEFFS = (
+    (8.156554524902461, -22.48329292557795, 15.878769915207462),
+    (4.042929935166739, -2.808917465908714, 0.5000178451051316),
+    (3.8916678022926607, -2.772484153217685, 0.5060648178503393),
+    (3.285753657755655, -2.3681294933425376, 0.46449024233003106),
+    (2.3465413258596377, -1.7097828382687081, 0.42323551169305323),
+)
+@torch.compile
+def zeropower_via_newtonschulz5(G, steps=10, eps=1e-07):
+    was_2d = G.ndim == 2
+    if was_2d:
+        G = G.unsqueeze(0)
+    X = G.bfloat16()
+    transposed = X.size(-2) > X.size(-1)
+    if transposed:
+        X = X.mT
+    X = X / (X.norm(dim=(-2, -1), keepdim=True) + eps)
+    coeffs = _PE_COEFFS[:steps] if steps <= len(_PE_COEFFS) else _PE_COEFFS
+    for a, b, c in coeffs:
+        A = X @ X.mT
+        B = b * A + c * (A @ A)
+        X = a * X + B @ X
+    if transposed:
+        X = X.mT
+    if was_2d:
+        X = X.squeeze(0)
+    return X
+class Muon(torch.optim.Optimizer):
+    def __init__(
+        self,
+        params,
+        lr,
+        momentum,
+        backend_steps,
+        nesterov=True,
+        weight_decay=0.0,
+        row_normalize=False,
+    ):
+        super().__init__(
+            params,
+            dict(
+                lr=lr,
+                momentum=momentum,
+                backend_steps=backend_steps,
+                nesterov=nesterov,
+                weight_decay=weight_decay,
+                row_normalize=row_normalize,
+            ),
+        )
+        self._built = False
+    def _build(self):
+        self._distributed = dist.is_available() and dist.is_initialized()
+        self._world_size = dist.get_world_size() if self._distributed else 1
+        self._rank = dist.get_rank() if self._distributed else 0
+        ws = self._world_size
+        self._bank_meta = []
+        for group in self.param_groups:
+            for p in group["params"]:
+                B = p.shape[0]
+                padded_B = ((B + ws - 1) // ws) * ws
+                shard_B = padded_B // ws
+                tail = p.shape[1:]
+                dev = p.device
+                self._bank_meta.append({
+                    "p": p,
+                    "B": B,
+                    "padded_grad": torch.zeros(padded_B, *tail, device=dev, dtype=torch.bfloat16),
+                    "shard": torch.zeros(shard_B, *tail, device=dev, dtype=torch.bfloat16),
+                    "shard_mom": torch.zeros(shard_B, *tail, device=dev, dtype=torch.bfloat16),
+                    "full_update": torch.zeros(padded_B, *tail, device=dev, dtype=torch.bfloat16),
+                    "scale": max(1, p.shape[-2] / p.shape[-1]) ** 0.5,
+                })
+        self._bank_meta.sort(key=lambda m: -m["p"].numel())
+        self._built = True
+    def launch_reduce_scatters(self):
+        if not self._built:
+            self._build()
+        if not self._distributed:
+            return
+        self._rs_futures = []
+        for m in self._bank_meta:
+            p = m["p"]
+            if p.grad is None:
+                self._rs_futures.append(None)
+                continue
+            pg = m["padded_grad"]
+            pg[: m["B"]].copy_(p.grad.bfloat16())
+            if pg.shape[0] > m["B"]:
+                pg[m["B"] :].zero_()
+            fut = dist.reduce_scatter_tensor(
+                m["shard"], pg, op=dist.ReduceOp.AVG, async_op=True
+            )
+            self._rs_futures.append(fut)
+    @torch.no_grad()
+    def step(self, closure=None):
+        loss = None
+        if closure is not None:
+            with torch.enable_grad():
+                loss = closure()
+        if not self._built:
+            self._build()
+        for group in self.param_groups:
+            lr = group["lr"]
+            momentum = group["momentum"]
+            backend_steps = group["backend_steps"]
+            nesterov = group["nesterov"]
+            wd = group.get("weight_decay", 0.0)
+            row_normalize = group.get("row_normalize", False)
+            prev_ag_handle = None
+            prev_m = None
+            sharded = self._distributed and hasattr(self, "_rs_futures")
+            for idx, m in enumerate(self._bank_meta):
+                p = m["p"]
+                if p.grad is None:
+                    continue
+                if prev_ag_handle is not None:
+                    prev_ag_handle.wait()
+                    pp = prev_m["p"]
+                    upd = prev_m["full_update"][: prev_m["B"]]
+                    if wd > 0.0:
+                        pp.data.mul_(1.0 - lr * wd)
+                    pp.add_(upd.to(dtype=pp.dtype), alpha=-lr * prev_m["scale"])
+                if sharded and self._rs_futures[idx] is not None:
+                    self._rs_futures[idx].wait()
+                    g = m["shard"]
+                    buf = m["shard_mom"]
+                else:
+                    g = p.grad.bfloat16()
+                    state = self.state[p]
+                    if "momentum_buffer" not in state:
+                        state["momentum_buffer"] = torch.zeros_like(g)
+                    buf = state["momentum_buffer"]
+                buf.mul_(momentum).add_(g)
+                if nesterov:
+                    update = g.add(buf, alpha=momentum)
+                else:
+                    update = buf
+                if row_normalize:
+                    rn = update.float().norm(dim=-1, keepdim=True).clamp_min(1e-07)
+                    update = update / rn.to(update.dtype)
+                update = zeropower_via_newtonschulz5(update, steps=backend_steps)
+                if sharded:
+                    prev_ag_handle = dist.all_gather_into_tensor(
+                        m["full_update"], update, async_op=True
+                    )
+                    prev_m = m
+                else:
+                    if wd > 0.0:
+                        p.data.mul_(1.0 - lr * wd)
+                    p.add_(update.to(dtype=p.dtype), alpha=-lr * m["scale"])
+            if prev_ag_handle is not None:
+                prev_ag_handle.wait()
+                pp = prev_m["p"]
+                upd = prev_m["full_update"][: prev_m["B"]]
+                if wd > 0.0:
+                    pp.data.mul_(1.0 - lr * wd)
+                pp.add_(upd.to(dtype=pp.dtype), alpha=-lr * prev_m["scale"])
+            if hasattr(self, "_rs_futures"):
+                del self._rs_futures
+        return loss
+CONTROL_TENSOR_NAME_PATTERNS = tuple(
+    pattern
+    for pattern in os.environ.get(
+        "CONTROL_TENSOR_NAME_PATTERNS",
+        "attn_scale,attn_scales,mlp_scale,mlp_scales,resid_mix,resid_mixes,q_gain,skip_weight,skip_weights,skip_gates,parallel_post_lambdas,parallel_resid_lambdas,attn_gate_proj,attn_gate_w,smear_gate,smear_lambda",
+    ).split(",")
+    if pattern
+)
+PACKED_REPLICATED_GRAD_MAX_NUMEL = 1 << 15
+class Optimizers:
+    def __init__(self, h, base_model):
+        matrix_params = [
+            base_model.qo_bank,
+            base_model.kv_bank,
+            base_model.mlp_up_bank,
+            base_model.mlp_down_bank,
+        ]
+        block_named_params = list(base_model.blocks.named_parameters())
+        scalar_params = [
+            p
+            for (name, p) in block_named_params
+            if p.ndim < 2
+            or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)
+        ]
+        if base_model.skip_weights.numel() > 0:
+            scalar_params.append(base_model.skip_weights)
+        if base_model.skip_gates is not None and base_model.skip_gates.numel() > 0:
+            scalar_params.append(base_model.skip_gates)
+        if base_model.parallel_post_lambdas is not None:
+            scalar_params.append(base_model.parallel_post_lambdas)
+        if base_model.parallel_resid_lambdas is not None:
+            scalar_params.append(base_model.parallel_resid_lambdas)
+        if getattr(base_model, "smear_gate_enabled", False):
+            scalar_params.append(base_model.smear_gate.weight)
+            scalar_params.append(base_model.smear_lambda)
+        token_lr = h.tied_embed_lr
+        tok_params = [
+            {"params": [base_model.tok_emb.weight], "lr": token_lr, "base_lr": token_lr}
+        ]
+        self.optimizer_tok = torch.optim.AdamW(
+            tok_params,
+            betas=(h.beta1, h.beta2),
+            eps=h.adam_eps,
+            weight_decay=h.embed_wd,
+            fused=True,
+        )
+        self.optimizer_muon = Muon(
+            matrix_params,
+            lr=h.matrix_lr,
+            momentum=h.muon_momentum,
+            backend_steps=h.muon_backend_steps,
+            weight_decay=h.muon_wd,
+            row_normalize=h.muon_row_normalize,
+        )
+        for group in self.optimizer_muon.param_groups:
+            group["base_lr"] = h.matrix_lr
+        self.optimizer_scalar = torch.optim.AdamW(
+            [{"params": scalar_params, "lr": h.scalar_lr, "base_lr": h.scalar_lr}],
+            betas=(h.beta1, h.beta2),
+            eps=h.adam_eps,
+            weight_decay=h.adam_wd,
+            fused=True,
+        )
+        self.optimizers = [
+            self.optimizer_tok,
+            self.optimizer_muon,
+            self.optimizer_scalar,
+        ]
+        self.replicated_params = list(tok_params[0]["params"])
+        self.replicated_params.extend(scalar_params)
+        self.replicated_large_params = []
+        self.replicated_packed_params = []
+        for p in self.replicated_params:
+            if p.numel() <= PACKED_REPLICATED_GRAD_MAX_NUMEL:
+                self.replicated_packed_params.append(p)
+            else:
+                self.replicated_large_params.append(p)
+    def __iter__(self):
+        return iter(self.optimizers)
+    def zero_grad_all(self):
+        for opt in self.optimizers:
+            opt.zero_grad(set_to_none=True)
+    def _all_reduce_packed_grads(self):
+        grads_by_key = collections.defaultdict(list)
+        for p in self.replicated_packed_params:
+            if p.grad is not None:
+                grads_by_key[(p.grad.device, p.grad.dtype)].append(p.grad)
+        for grads in grads_by_key.values():
+            flat = torch.empty(
+                sum(g.numel() for g in grads),
+                device=grads[0].device,
+                dtype=grads[0].dtype,
+            )
+            offset = 0
+            for g in grads:
+                n = g.numel()
+                flat[offset : offset + n].copy_(g.contiguous().view(-1))
+                offset += n
+            dist.all_reduce(flat, op=dist.ReduceOp.AVG)
+            offset = 0
+            for g in grads:
+                n = g.numel()
+                g.copy_(flat[offset : offset + n].view_as(g))
+                offset += n
+    def step(self, distributed=False):
+        self.optimizer_muon.launch_reduce_scatters()
+        if distributed:
+            reduce_handles = [
+                dist.all_reduce(p.grad, op=dist.ReduceOp.AVG, async_op=True)
+                for p in self.replicated_large_params
+                if p.grad is not None
+            ]
+            self._all_reduce_packed_grads()
+            for handle in reduce_handles:
+                handle.wait()
+        self.optimizer_tok.step()
+        self.optimizer_scalar.step()
+        self.optimizer_muon.step()
+        self.zero_grad_all()
+def restore_fp32_params(model):
+    for module in model.modules():
+        if isinstance(module, CastedLinear):
+            module.float()
+    for name, param in model.named_parameters():
+        if (
+            param.ndim < 2
+            or any(pattern in name for pattern in CONTROL_TENSOR_NAME_PATTERNS)
+        ) and param.dtype != torch.float32:
+            param.data = param.data.float()
+    if hasattr(model, "qo_bank") and model.qo_bank is not None:
+        model.qo_bank.data = model.qo_bank.data.float()
+        model.kv_bank.data = model.kv_bank.data.float()
+    model.mlp_up_bank.data = model.mlp_up_bank.data.float()
+    model.mlp_down_bank.data = model.mlp_down_bank.data.float()
+def collect_hessians(model, train_loader, h, device, n_calibration_batches=64):
+    hessians = {}
+    hooks = []
+    for i, block in enumerate(model.blocks):
+        block.attn._calib = True
+        block.mlp._calib = True
+        block.mlp.use_fused = False
+    def make_attn_hook(layer_idx):
+        def hook_fn(module, inp, out):
+            x = inp[0].detach().float()
+            if x.ndim == 3:
+                x = x.reshape(-1, x.shape[-1])
+            for suffix in ["c_q", "c_k", "c_v"]:
+                name = f"blocks.{layer_idx}.attn.{suffix}.weight"
+                if name not in hessians:
+                    hessians[name] = torch.zeros(
+                        x.shape[1], x.shape[1], dtype=torch.float32, device=device
+                    )
+                hessians[name].addmm_(x.T, x)
+            y = module._last_proj_input
+            if y is not None:
+                y = y.float()
+                if y.ndim == 3:
+                    y = y.reshape(-1, y.shape[-1])
+                name = f"blocks.{layer_idx}.attn.proj.weight"
+                if name not in hessians:
+                    hessians[name] = torch.zeros(
+                        y.shape[1], y.shape[1], dtype=torch.float32, device=device
+                    )
+                hessians[name].addmm_(y.T, y)
+        return hook_fn
+    def make_mlp_hook(layer_idx):
+        def hook_fn(module, inp, out):
+            x = inp[0].detach().float()
+            if x.ndim == 3:
+                x = x.reshape(-1, x.shape[-1])
+            name = f"blocks.{layer_idx}.mlp.fc.weight"
+            if name not in hessians:
+                hessians[name] = torch.zeros(
+                    x.shape[1], x.shape[1], dtype=torch.float32, device=device
+                )
+            hessians[name].addmm_(x.T, x)
+            h_act = module._last_down_input
+            if h_act is not None:
+                h_act = h_act.float()
+                if h_act.ndim == 3:
+                    h_act = h_act.reshape(-1, h_act.shape[-1])
+                name = f"blocks.{layer_idx}.mlp.proj.weight"
+                if name not in hessians:
+                    hessians[name] = torch.zeros(
+                        h_act.shape[1], h_act.shape[1], dtype=torch.float32, device=device
+                    )
+                hessians[name].addmm_(h_act.T, h_act)
+        return hook_fn
+    for i, block in enumerate(model.blocks):
+        hooks.append(block.attn.register_forward_hook(make_attn_hook(i)))
+        hooks.append(block.mlp.register_forward_hook(make_mlp_hook(i)))
+    def make_output_hook(name):
+        def hook_fn(module, inp, out):
+            x = out.detach().float()
+            if x.ndim == 3:
+                x = x.reshape(-1, x.shape[-1])
+            if name not in hessians:
+                hessians[name] = torch.zeros(
+                    x.shape[1], x.shape[1], dtype=torch.float32, device=device
+                )
+            hessians[name].addmm_(x.T, x)
+        return hook_fn
+    hooks.append(
+        model.final_norm.register_forward_hook(make_output_hook("tok_emb.weight"))
+    )
+    model.eval()
+    with torch.no_grad():
+        for _ in range(n_calibration_batches):
+            x, _ = train_loader.next_batch(h.train_batch_tokens, h.grad_accum_steps)
+            model.forward_logits(x)
+    for hook in hooks:
+        hook.remove()
+    for i, block in enumerate(model.blocks):
+        block.attn._calib = False
+        block.mlp._calib = False
+        block.mlp.use_fused = True
+    for name in hessians:
+        hessians[name] = hessians[name].cpu() / n_calibration_batches
+    return hessians
+def gptq_quantize_weight(w, H, clip_sigmas=3.0, clip_range=63, block_size=128):
+    W_orig = w.float().clone()
+    rows, cols = W_orig.shape
+    H = H.float().clone()
+    dead = torch.diag(H) == 0
+    H[dead, dead] = 1
+    damp = 0.01 * H.diag().mean()
+    H.diagonal().add_(damp)
+    perm = torch.argsort(H.diag(), descending=True)
+    invperm = torch.argsort(perm)
+    W_perm = W_orig[:, perm].clone()
+    W_perm[:, dead[perm]] = 0
+    H = H[perm][:, perm]
+    H_flip = torch.flip(H, dims=(0, 1))
+    L_flip = torch.linalg.cholesky(H_flip)
+    U = torch.flip(L_flip, dims=(0, 1))
+    eye = torch.eye(cols, dtype=H.dtype, device=H.device)
+    Hinv = torch.linalg.solve_triangular(U, eye, upper=True)
+    row_std = W_orig.std(dim=1)
+    s = (clip_sigmas * row_std / clip_range).clamp_min(1e-10).to(torch.float16)
+    sf = s.float()
+    Q = torch.zeros(rows, cols, dtype=torch.int8)
+    W_work = W_perm.clone()
+    for i1 in range(0, cols, block_size):
+        i2 = min(i1 + block_size, cols)
+        W_block = W_work[:, i1:i2].clone()
+        Hinv_block = Hinv[i1:i2, i1:i2]
+        Err = torch.zeros(rows, i2 - i1)
+        for j in range(i2 - i1):
+            w_col = W_block[:, j]
+            d = Hinv_block[j, j]
+            q_col = torch.clamp(torch.round(w_col / sf), -clip_range, clip_range)
+            Q[:, i1 + j] = q_col.to(torch.int8)
+            err = (w_col - q_col.float() * sf) / d
+            Err[:, j] = err
+            W_block[:, j:] -= err.unsqueeze(1) * Hinv_block[j, j:].unsqueeze(0)
+        if i2 < cols:
+            W_work[:, i2:] -= Err @ Hinv[i1:i2, i2:]
+    return Q[:, invperm], s
+def _quantize_gate_int8_row(w):
+    W = w.float().contiguous()
+    row_max = W.abs().amax(dim=1).clamp_min(1e-10)
+    s = (row_max / 127.0).to(torch.float16)
+    sf = s.float().view(-1, 1)
+    q = torch.clamp(torch.round(W / sf), -127, 127).to(torch.int8)
+    return q, s
+def _lqer_pack(A, B, bits):
+    rng = 2 ** (bits - 1) - 1
+    sA = (A.abs().amax(dim=1).clamp_min(1e-10) / rng).to(torch.float16)
+    sB = (B.abs().amax(dim=1).clamp_min(1e-10) / rng).to(torch.float16)
+    qA = torch.clamp(torch.round(A / sA.float().view(-1, 1)), -rng, rng).to(torch.int8)
+    qB = torch.clamp(torch.round(B / sB.float().view(-1, 1)), -rng, rng).to(torch.int8)
+    return qA, sA, qB, sB
+def _lqer_pack_asym(A, B, g=64):
+    sA = (A.abs().amax().clamp_min(1e-10) / 1.5).to(torch.float16)
+    qA = torch.clamp(torch.round(A / sA.float()), -2, 1).to(torch.int8)
+    Bf = B.reshape(-1, g)
+    Bmax = Bf.abs().amax(dim=-1, keepdim=True).clamp_min(1e-10)
+    sB = (Bmax / 7.5).to(torch.float16).reshape(-1)
+    qB = torch.clamp(torch.round(Bf / sB.float().reshape(-1, 1)), -8, 7).to(
+        torch.int8
+    ).reshape(B.shape)
+    return qA, sA, qB, sB
+def gptq_mixed_quantize(state_dict, hessians, h):
+    result = {}
+    meta = {}
+    quant_gate = bool(getattr(h, "gated_attn_quant_gate", False))
+    lqer_on = bool(getattr(h, "lqer_enabled", False))
+    lqer_cands = {}
+    for (name, tensor) in state_dict.items():
+        t = tensor.detach().cpu().contiguous()
+        if (
+            quant_gate
+            and t.is_floating_point()
+            and t.ndim == 2
+            and name.endswith(".attn_gate_w")
+            and 32 <= t.numel() <= 8192
+        ):
+            gq, gs = _quantize_gate_int8_row(t)
+            result[name + ".gq"] = gq
+            result[name + ".gs"] = gs
+            meta[name] = "gate_int8_row"
+            continue
+        if not t.is_floating_point() or t.numel() <= 65536:
+            result[name] = t.to(torch.float16) if t.is_floating_point() else t
+            meta[name] = "passthrough (float16)"
+            continue
+        if "tok_emb" in name:
+            cs = h.embed_clip_sigmas
+        elif ".mlp." in name:
+            cs = h.mlp_clip_sigmas
+        elif ".attn." in name:
+            cs = h.attn_clip_sigmas
+        else:
+            cs = h.matrix_clip_sigmas
+        bits = h.embed_bits if "tok_emb" in name else h.matrix_bits
+        clip_range = 2 ** (bits - 1) - 1
+        ret = gptq_quantize_weight(
+            t, hessians[name], clip_sigmas=cs, clip_range=clip_range
+        )
+        q, s = ret
+        result[name + ".q"] = q
+        result[name + ".scale"] = s
+        meta[name] = f"gptq (int{bits})"
+        if lqer_on:
+            W_q = q.float() * s.float().view(-1, 1)
+            E = t.float() - W_q
+            lqer_cands[name] = (E, float(E.norm()))
+    if lqer_on and lqer_cands:
+        top = sorted(lqer_cands.items(), key=lambda kv: -kv[1][1])[: h.lqer_top_k]
+        asym_on = bool(getattr(h, "lqer_asym_enabled", False))
+        asym_g = int(getattr(h, "lqer_asym_group", 64))
+        for (name, (E, _)) in top:
+            U, S, Vh = torch.linalg.svd(E, full_matrices=False)
+            r = min(h.lqer_rank, S.numel())
+            A = (U[:, :r] * S[:r]).contiguous()
+            B = Vh[:r, :].contiguous()
+            if asym_on and B.numel() % asym_g == 0:
+                qA, sA, qB, sB = _lqer_pack_asym(A, B, asym_g)
+                result[name + ".lqA_a"] = qA
+                result[name + ".lqAs_a"] = sA
+                result[name + ".lqB_a"] = qB
+                result[name + ".lqBs_a"] = sB
+                meta[name] = meta[name] + "+lqer_asym"
+            else:
+                qA, sA, qB, sB = _lqer_pack(A, B, h.lqer_factor_bits)
+                result[name + ".lqA"] = qA
+                result[name + ".lqAs"] = sA
+                result[name + ".lqB"] = qB
+                result[name + ".lqBs"] = sB
+                meta[name] = meta[name] + "+lqer"
+    categories = collections.defaultdict(set)
+    for (name, cat) in meta.items():
+        short = re.sub("\\.\\d+$", "", re.sub("blocks\\.\\d+", "blocks", name))
+        categories[cat].add(short)
+    log("Quantized weights:")
+    for cat in sorted(categories):
+        log(f"  {cat}: {', '.join(sorted(categories[cat]))}")
+    return result, meta
+def dequantize_mixed(result, meta, template_sd):
+    out = {}
+    for (name, orig) in template_sd.items():
+        info = meta.get(name)
+        if info is None:
+            continue
+        orig_dtype = orig.dtype
+        if "passthrough" in info:
+            t = result[name]
+            if t.dtype == torch.float16 and orig_dtype in (
+                torch.float32,
+                torch.bfloat16,
+            ):
+                t = t.to(orig_dtype)
+            out[name] = t
+            continue
+        if info == "gate_int8_row":
+            gq = result[name + ".gq"]
+            gs = result[name + ".gs"]
+            out[name] = (gq.float() * gs.float().view(-1, 1)).to(orig_dtype)
+            continue
+        q, s = result[name + ".q"], result[name + ".scale"]
+        if s.ndim > 0:
+            W = q.float() * s.float().view(q.shape[0], *[1] * (q.ndim - 1))
+        else:
+            W = q.float() * float(s.item())
+        if "lqer_asym" in info:
+            qA_t = result[name + ".lqA_a"]
+            sA_t = result[name + ".lqAs_a"]
+            qB_t = result[name + ".lqB_a"]
+            sB_t = result[name + ".lqBs_a"]
+            qA = qA_t.float() * float(sA_t)
+            g_sz = qB_t.numel() // sB_t.numel()
+            qB = (qB_t.reshape(-1, g_sz).float() * sB_t.float().view(-1, 1)).reshape(
+                qB_t.shape
+            )
+            W = W + qA @ qB
+        elif "lqer" in info:
+            qA = result[name + ".lqA"].float() * result[name + ".lqAs"].float().view(-1, 1)
+            qB = result[name + ".lqB"].float() * result[name + ".lqBs"].float().view(-1, 1)
+            W = W + qA @ qB
+        out[name] = W.to(orig_dtype)
+    return out
+_BSHF_MAGIC = b"BSHF"
+def _byte_shuffle(data, stride=2):
+    if stride <= 1 or len(data) < stride:
+        return data
+    src = np.frombuffer(data, dtype=np.uint8)
+    n = len(src)
+    out = np.empty(n, dtype=np.uint8)
+    dest_off = 0
+    for pos in range(stride):
+        chunk = src[pos::stride]
+        out[dest_off : dest_off + len(chunk)] = chunk
+        dest_off += len(chunk)
+    return _BSHF_MAGIC + bytes([stride]) + out.tobytes()
+def _byte_unshuffle(data):
+    if len(data) < 5 or data[:4] != _BSHF_MAGIC:
+        return data
+    stride = data[4]
+    if stride < 2:
+        return data[5:]
+    payload = np.frombuffer(data, dtype=np.uint8, offset=5)
+    n = len(payload)
+    out = np.empty(n, dtype=np.uint8)
+    src_off = 0
+    for pos in range(stride):
+        chunk_len = n // stride + (1 if pos < n % stride else 0)
+        out[pos::stride][:chunk_len] = payload[src_off : src_off + chunk_len]
+        src_off += chunk_len
+    return out.tobytes()
+def _compress(data, compressor):
+    data = _byte_shuffle(data)
+    if compressor == "lzma":
+        return lzma.compress(data, preset=6)
+    elif compressor == "brotli":
+        import brotli
+        return brotli.compress(data, quality=11)
+    raise ValueError(f"Unknown compressor: {compressor!r}")
+def _decompress(data, compressor):
+    if compressor == "lzma":
+        raw = lzma.decompress(data)
+    elif compressor == "brotli":
+        import brotli
+        raw = brotli.decompress(data)
+    else:
+        raise ValueError(f"Unknown compressor: {compressor!r}")
+    raw = _byte_unshuffle(raw)
+    return raw
+def _unbank_state_dict(state_dict, num_layers):
+    sd = {}
+    n = num_layers
+    for k, v in state_dict.items():
+        t = v.detach().cpu() if v is not None else None
+        if k == "qo_bank":
+            for i in range(n):
+                sd[f"blocks.{i}.attn.c_q.weight"] = t[i]
+                sd[f"blocks.{i}.attn.proj.weight"] = t[n + i]
+        elif k == "kv_bank":
+            for i in range(n):
+                sd[f"blocks.{i}.attn.c_k.weight"] = t[i]
+                sd[f"blocks.{i}.attn.c_v.weight"] = t[n + i]
+        elif k == "mlp_up_bank":
+            for i in range(n):
+                sd[f"blocks.{i}.mlp.fc.weight"] = t[i]
+        elif k == "mlp_down_bank":
+            for i in range(n):
+                sd[f"blocks.{i}.mlp.proj.weight"] = t[i]
+        else:
+            if t is not None:
+                sd[k] = t
+    return sd
+def _rebank_state_dict(flat_sd, num_layers, model_dim, kv_dim, hidden_dim):
+    sd = {}
+    n = num_layers
+    sd["qo_bank"] = torch.zeros(2 * n, model_dim, model_dim)
+    sd["kv_bank"] = torch.zeros(2 * n, kv_dim, model_dim)
+    for i in range(n):
+        sd["qo_bank"][i] = flat_sd[f"blocks.{i}.attn.c_q.weight"]
+        sd["qo_bank"][n + i] = flat_sd[f"blocks.{i}.attn.proj.weight"]
+        sd["kv_bank"][i] = flat_sd[f"blocks.{i}.attn.c_k.weight"]
+        sd["kv_bank"][n + i] = flat_sd[f"blocks.{i}.attn.c_v.weight"]
+    sd["mlp_up_bank"] = torch.zeros(n, hidden_dim, model_dim)
+    sd["mlp_down_bank"] = torch.zeros(n, model_dim, hidden_dim)
+    for i in range(n):
+        sd["mlp_up_bank"][i] = flat_sd[f"blocks.{i}.mlp.fc.weight"]
+        sd["mlp_down_bank"][i] = flat_sd[f"blocks.{i}.mlp.proj.weight"]
+    for k, v in flat_sd.items():
+        if not (
+            k.startswith("blocks.")
+            and any(
+                p in k
+                for p in [
+                    ".attn.c_q.", ".attn.c_k.", ".attn.c_v.",
+                    ".attn.proj.", ".mlp.fc.", ".mlp.proj.",
+                ]
+            )
+        ):
+            sd[k] = v
+    return sd
+def _compressed_code_size(code):
+    code_raw = code.encode("utf-8")
+    minified = subprocess.run(
+        ["pyminify", "--no-rename-locals", "--no-hoist-literals", "--remove-literal-statements", "-"],
+        input=code_raw, capture_output=True, check=True,
+    ).stdout
+    compressed = lzma.compress(minified)
+    encoded = base64.b85encode(compressed)
+    wrapper = b'import lzma as L,base64 as B\nexec(L.decompress(B.b85decode("' + encoded + b'")))\n'
+    return len(code_raw), len(wrapper)
+def serialize(h, base_model, code):
+    try:
+        code_bytes_uncompressed, code_bytes = _compressed_code_size(code)
+    except (FileNotFoundError, subprocess.CalledProcessError) as e:
+        code_bytes_uncompressed, code_bytes = len(code.encode("utf-8")), 0
+        log(f"pyminify unavailable ({type(e).__name__}); skipping compressed-code-size measurement")
+    if h.is_main_process:
+        torch.save(base_model.state_dict(), h.model_path)
+        model_bytes = os.path.getsize(h.model_path)
+        log(f"Serialized model: {model_bytes} bytes")
+        log(f"Code size (uncompressed): {code_bytes_uncompressed} bytes")
+        log(f"Code size (compressed): {code_bytes} bytes")
+    sd_cpu = _unbank_state_dict(base_model.state_dict(), h.num_layers)
+    device = torch.device("cuda", h.local_rank)
+    t0 = time.perf_counter()
+    calib_loader = ShuffledSequenceLoader(h, device)
+    log("GPTQ:collecting Hessians from calibration data...")
+    hessians = collect_hessians(
+        base_model,
+        calib_loader,
+        h,
+        device,
+        n_calibration_batches=h.gptq_calibration_batches,
+    )
+    log(f"GPTQ:collected {len(hessians)} Hessians in {time.perf_counter()-t0:.1f}s")
+    t_quant = time.perf_counter()
+    quant_result, quant_meta = gptq_mixed_quantize(sd_cpu, hessians, h)
+    log(f"GPTQ:quantized weights in {time.perf_counter()-t_quant:.2f}s")
+    quant_buf = io.BytesIO()
+    torch.save({"w": quant_result, "m": quant_meta}, quant_buf)
+    quant_raw = quant_buf.getvalue()
+    quant_blob = _compress(quant_raw, h.compressor)
+    quant_file_bytes = len(quant_blob)
+    bytes_total = quant_file_bytes + code_bytes
+    if h.is_main_process:
+        with open(h.quantized_model_path, "wb") as f:
+            f.write(quant_blob)
+        log(f"Serialized model quantized+{h.compressor}: {quant_file_bytes} bytes")
+        log(f"Total submission size quantized+{h.compressor}: {bytes_total} bytes")
+    return bytes_total, quant_file_bytes
+def deserialize(h, device):
+    eval_model = GPT(h).to(device).bfloat16()
+    restore_fp32_params(eval_model)
+    flat_template = _unbank_state_dict(eval_model.state_dict(), h.num_layers)
+    with open(h.quantized_model_path, "rb") as f:
+        quant_blob_disk = f.read()
+    quant_state = torch.load(
+        io.BytesIO(_decompress(quant_blob_disk, h.compressor)), map_location="cpu"
+    )
+    deq_flat = dequantize_mixed(quant_state["w"], quant_state["m"], flat_template)
+    head_dim = h.model_dim // h.num_heads
+    kv_dim = h.num_kv_heads * head_dim
+    hidden_dim = int(h.mlp_mult * h.model_dim)
+    deq_state = _rebank_state_dict(deq_flat, h.num_layers, h.model_dim, kv_dim, hidden_dim)
+    eval_model.load_state_dict(deq_state, strict=True)
+    return eval_model
+def _loss_bpb(loss_sum, token_count, byte_count):
+    val_loss = (loss_sum / token_count).item()
+    val_bpb = val_loss / math.log(2.0) * (token_count.item() / byte_count.item())
+    return val_loss, val_bpb
+def eval_val(h, device, val_data, model, forward_logits_fn=None):
+    seq_len = h.eval_seq_len
+    local_batch_tokens = h.val_batch_tokens // (h.world_size * h.grad_accum_steps)
+    if local_batch_tokens < seq_len:
+        raise ValueError(
+            f"VAL_BATCH_SIZE must provide at least one sequence per rank; got VAL_BATCH_SIZE={h.val_batch_tokens}, WORLD_SIZE={h.world_size}, GRAD_ACCUM_STEPS={h.grad_accum_steps}, seq_len={seq_len}"
+        )
+    local_batch_seqs = local_batch_tokens // seq_len
+    total_seqs = (val_data.val_tokens.numel() - 1) // seq_len
+    seq_start = total_seqs * h.rank // h.world_size
+    seq_end = total_seqs * (h.rank + 1) // h.world_size
+    seq_end = seq_start + ((seq_end - seq_start) // local_batch_seqs) * local_batch_seqs
+    val_loss_sum = torch.zeros((), device=device, dtype=torch.float64)
+    val_token_count = torch.zeros((), device=device, dtype=torch.float64)
+    val_byte_count = torch.zeros((), device=device, dtype=torch.float64)
+    run_forward_logits = (
+        (model.module.forward_logits if hasattr(model, "module") else model.forward_logits)
+        if forward_logits_fn is None
+        else forward_logits_fn
+    )
+    model.eval()
+    global BOS_ID
+    if BOS_ID is None:
+        BOS_ID = 1
+    with torch.no_grad():
+        for batch_seq_start in range(seq_start, seq_end, local_batch_seqs):
+            batch_seq_end = min(batch_seq_start + local_batch_seqs, seq_end)
+            raw_start = batch_seq_start * seq_len
+            raw_end = batch_seq_end * seq_len + 1
+            local = val_data.val_tokens[raw_start:raw_end].to(
+                device=device, dtype=torch.int64, non_blocking=True
+            )
+            x = local[:-1]
+            y = local[1:]
+            bos_pos = (x == BOS_ID).nonzero(as_tuple=True)[0].tolist()
+            cu_seqlens, max_seqlen = _build_cu_seqlens(
+                bos_pos, x.numel(), x.device, h.eval_seq_len, 64
+            )
+            with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True):
+                logits = run_forward_logits(
+                    x[None], cu_seqlens=cu_seqlens, max_seqlen=max_seqlen
+                ).detach()
+            per_token_loss = F.cross_entropy(
+                logits.reshape(-1, logits.size(-1)).float(),
+                y.reshape(-1),
+                reduction="none",
+            )
+            val_loss_sum += per_token_loss.to(torch.float64).sum()
+            val_token_count += float(y.numel())
+            prev_ids = x
+            tgt_ids = y
+            if val_data.caseops_enabled and val_data.val_bytes is not None:
+                sidecar_slice = val_data.val_bytes[raw_start + 1 : raw_end].to(
+                    device=device, dtype=torch.int32, non_blocking=True
+                )
+                val_byte_count += sidecar_slice.to(torch.float64).sum()
+            else:
+                token_bytes = val_data.base_bytes_lut[tgt_ids].to(dtype=torch.int16)
+                token_bytes += (
+                    val_data.has_leading_space_lut[tgt_ids]
+                    & ~val_data.is_boundary_token_lut[prev_ids]
+                ).to(dtype=torch.int16)
+                val_byte_count += token_bytes.to(torch.float64).sum()
+    if dist.is_available() and dist.is_initialized():
+        dist.all_reduce(val_loss_sum, op=dist.ReduceOp.SUM)
+        dist.all_reduce(val_token_count, op=dist.ReduceOp.SUM)
+        dist.all_reduce(val_byte_count, op=dist.ReduceOp.SUM)
+    model.train()
+    return _loss_bpb(val_loss_sum, val_token_count, val_byte_count)
+def _find_docs(all_tokens):
+    bos_positions = (all_tokens == BOS_ID).nonzero(as_tuple=True)[0].numpy()
+    docs = []
+    for i in range(len(bos_positions)):
+        start = int(bos_positions[i])
+        end = (
+            int(bos_positions[i + 1])
+            if i + 1 < len(bos_positions)
+            else all_tokens.numel()
+        )
+        if i + 1 < len(bos_positions):
+            end += 1
+        assert end - start >= 2
+        docs.append((start, end - start))
+    return docs
+def _build_ttt_global_batches(doc_entries, h, ascending=False):
+    batch_size = h.ttt_batch_size
+    global_doc_entries = sorted(doc_entries, key=lambda x: x[1][1])
+    global_batches = [
+        global_doc_entries[i : i + batch_size]
+        for i in range(0, len(global_doc_entries), batch_size)
+    ]
+    indexed = list(enumerate(global_batches))
+    if not ascending:
+        indexed.sort(key=lambda ib: -max(dl for _, (_, dl) in ib[1]))
+    return indexed
+def _init_batch_counter(path):
+    with open(path, "wb") as f:
+        f.write((0).to_bytes(4, "little"))
+def _claim_next_batch(counter_path, queue_len):
+    try:
+        with open(counter_path, "r+b") as f:
+            fcntl.flock(f, fcntl.LOCK_EX)
+            idx = int.from_bytes(f.read(4), "little")
+            f.seek(0)
+            f.write((idx + 1).to_bytes(4, "little"))
+            f.flush()
+    except FileNotFoundError:
+        return queue_len
+    return idx
+def _compute_chunk_window(ci, pred_len, num_chunks, chunk_size, eval_seq_len):
+    chunk_end = pred_len if ci == num_chunks - 1 else (ci + 1) * chunk_size
+    win_start = max(0, chunk_end - eval_seq_len)
+    win_len = chunk_end - win_start
+    chunk_start = ci * chunk_size
+    chunk_offset = chunk_start - win_start
+    chunk_len = chunk_end - chunk_start
+    return win_start, win_len, chunk_offset, chunk_len
+def _accumulate_bpb(
+    ptl,
+    x,
+    y,
+    chunk_offsets,
+    chunk_lens,
+    pos_idx,
+    base_bytes_lut,
+    has_leading_space_lut,
+    is_boundary_token_lut,
+    loss_sum,
+    byte_sum,
+    token_count,
+    y_bytes=None,
+):
+    pos = pos_idx[: x.size(1)].unsqueeze(0)
+    mask = (
+        (chunk_lens.unsqueeze(1) > 0)
+        & (pos >= chunk_offsets.unsqueeze(1))
+        & (pos < (chunk_offsets + chunk_lens).unsqueeze(1))
+    )
+    mask_f64 = mask.to(torch.float64)
+    if y_bytes is not None:
+        tok_bytes = y_bytes.to(torch.float64)
+    else:
+        tok_bytes = base_bytes_lut[y].to(torch.float64)
+        tok_bytes += (has_leading_space_lut[y] & ~is_boundary_token_lut[x]).to(
+            torch.float64
+        )
+    loss_sum += (ptl.to(torch.float64) * mask_f64).sum()
+    byte_sum += (tok_bytes * mask_f64).sum()
+    token_count += chunk_lens.to(torch.float64).sum()
+def _add_to_counter(path, delta):
+    try:
+        with open(path, "r+b") as f:
+            fcntl.flock(f, fcntl.LOCK_EX)
+            cur = int.from_bytes(f.read(8), "little", signed=True)
+            cur += int(delta)
+            f.seek(0)
+            f.write(int(cur).to_bytes(8, "little", signed=True))
+            f.flush()
+            return cur
+    except FileNotFoundError:
+        return int(delta)
+def _init_int64_counter(path):
+    with open(path, "wb") as f:
+        f.write((0).to_bytes(8, "little", signed=True))
+def _select_ttt_doc_entries(docs, h):
+    doc_entries = list(enumerate(docs))
+    if h.val_doc_fraction < 1.0:
+        sample_n = max(1, int(round(len(docs) * h.val_doc_fraction)))
+        sampled_indices = sorted(
+            random.Random(h.seed).sample(range(len(docs)), sample_n)
+        )
+        return [(i, docs[i]) for i in sampled_indices]
+    return doc_entries
+def train_val_ttt_global_sgd_distributed(h, device, val_data, base_model, val_tokens, batch_seqs=None):
+    global BOS_ID
+    if BOS_ID is None:
+        BOS_ID = 1
+    base_model.eval()
+    seq_len = h.eval_seq_len
+    total_tokens = val_tokens.numel() - 1
+    ttt_chunk = h.global_ttt_chunk_tokens
+    batch_seqs = h.global_ttt_batch_seqs if batch_seqs is None else batch_seqs
+    num_chunks = (total_tokens + ttt_chunk - 1) // ttt_chunk
+    ttt_params = [p for p in base_model.parameters()]
+    for p in ttt_params:
+        p.requires_grad_(True)
+    optimizer = torch.optim.SGD(
+        ttt_params, lr=h.global_ttt_lr, momentum=h.global_ttt_momentum
+    )
+    t_start = time.perf_counter()
+    for ci in range(num_chunks):
+        chunk_start = ci * ttt_chunk
+        chunk_end = min((ci + 1) * ttt_chunk, total_tokens)
+        is_last_chunk = ci == num_chunks - 1
+        if is_last_chunk or h.global_ttt_epochs <= 0:
+            continue
+        base_model.train()
+        chunk_seqs = (chunk_end - chunk_start) // seq_len
+        if chunk_seqs <= 0:
+            continue
+        warmup_chunks = max(0, min(h.global_ttt_warmup_chunks, num_chunks - 1))
+        if warmup_chunks > 0 and ci < warmup_chunks:
+            warmup_denom = max(warmup_chunks - 1, 1)
+            warmup_t = ci / warmup_denom
+            lr_now = (
+                h.global_ttt_warmup_start_lr
+                + (h.global_ttt_lr - h.global_ttt_warmup_start_lr) * warmup_t
+            )
+        else:
+            decay_steps = max(num_chunks - 1 - warmup_chunks, 1)
+            decay_ci = max(ci - warmup_chunks, 0)
+            lr_now = h.global_ttt_lr * 0.5 * (
+                1.0 + math.cos(math.pi * decay_ci / decay_steps)
+            )
+        for pg in optimizer.param_groups:
+            pg["lr"] = lr_now
+        my_seq_s = chunk_seqs * h.rank // h.world_size
+        my_seq_e = chunk_seqs * (h.rank + 1) // h.world_size
+        my_chunk_seqs = my_seq_e - my_seq_s
+        for _ in range(h.global_ttt_epochs):
+            for bs in range(0, my_chunk_seqs, batch_seqs):
+                be = min(bs + batch_seqs, my_chunk_seqs)
+                actual_bs = my_seq_s + bs
+                start_tok = chunk_start + actual_bs * seq_len
+                end_tok = chunk_start + (my_seq_s + be) * seq_len + 1
+                if end_tok > val_tokens.numel():
+                    continue
+                local = val_tokens[start_tok:end_tok].to(device=device, dtype=torch.int64)
+                x_flat = local[:-1]
+                y_flat = local[1:]
+                optimizer.zero_grad(set_to_none=True)
+                with torch.enable_grad():
+                    with torch.autocast(device_type="cuda", dtype=torch.bfloat16):
+                        if h.global_ttt_respect_doc_boundaries:
+                            bos_pos = (x_flat == BOS_ID).nonzero(as_tuple=True)[0].tolist()
+                            cu_seqlens, max_seqlen = _build_cu_seqlens(
+                                bos_pos, x_flat.numel(), x_flat.device, h.eval_seq_len, 64
+                            )
+                            loss = base_model(
+                                x_flat[None],
+                                y_flat[None],
+                                cu_seqlens=cu_seqlens,
+                                max_seqlen=max_seqlen,
+                            )
+                        else:
+                            x = x_flat.reshape(-1, seq_len)
+                            y = y_flat.reshape(-1, seq_len)
+                            loss = base_model(x, y)
+                loss.backward()
+                if dist.is_available() and dist.is_initialized():
+                    for p in ttt_params:
+                        if p.grad is not None:
+                            dist.all_reduce(p.grad, op=dist.ReduceOp.SUM)
+                            p.grad.mul_(1.0 / h.world_size)
+                if h.global_ttt_grad_clip > 0:
+                    torch.nn.utils.clip_grad_norm_(ttt_params, h.global_ttt_grad_clip)
+                optimizer.step()
+        base_model.eval()
+        if h.rank == 0:
+            elapsed = time.perf_counter() - t_start
+            log(
+                f"tttg: c{ci+1}/{num_chunks} lr:{lr_now:.6f} t:{elapsed:.1f}s"
+            )
+    for p in base_model.parameters():
+        p.requires_grad_(True)
+    base_model.eval()
+def eval_val_ttt_phased(h, base_model, device, val_data, forward_ttt_train):
+    global BOS_ID
+    if BOS_ID is None:
+        BOS_ID = 1
+    base_model.eval()
+    for p in base_model.parameters():
+        p.requires_grad_(False)
+    all_tokens = val_data.val_tokens
+    all_tokens_idx = all_tokens.to(torch.int32)
+    docs = _find_docs(all_tokens)
+    doc_entries = _select_ttt_doc_entries(docs, h)
+    prefix_doc_limit = max(0, min(len(doc_entries), int(h.phased_ttt_prefix_docs)))
+    num_phases = max(1, int(h.phased_ttt_num_phases))
+    phase_boundaries = []
+    for pi in range(num_phases):
+        boundary = prefix_doc_limit * (pi + 1) // num_phases
+        phase_boundaries.append(boundary)
+    current_phase = 0
+    current_phase_boundary = phase_boundaries[0]
+    log(
+        "ttt_phased:"
+        f" total_docs:{len(doc_entries)} prefix_docs:{prefix_doc_limit} "
+        f"suffix_docs:{len(doc_entries) - prefix_doc_limit}"
+        f" num_phases:{num_phases} boundaries:{phase_boundaries}"
+    )
+    chunk_size, eval_seq_len = h.ttt_chunk_size, h.ttt_eval_seq_len
+    eval_batch_set = None
+    if h.ttt_eval_batches:
+        eval_batch_set = set(int(x) for x in h.ttt_eval_batches.split(",") if x.strip())
+    use_ascending = eval_batch_set is not None
+    global_batches_sorted = _build_ttt_global_batches(
+        doc_entries, h, ascending=use_ascending
+    )
+    queue_len = len(global_batches_sorted)
+    counter_path = f"/tmp/ttt_counter_{h.run_id}"
+    prefix_counter_path = f"/tmp/ttt_prefix_counter_{h.run_id}"
+    pause_flag_path = f"/tmp/ttt_pause_flag_{h.run_id}"
+    if h.rank == 0:
+        _init_batch_counter(counter_path)
+        _init_int64_counter(prefix_counter_path)
+        try:
+            os.remove(pause_flag_path)
+        except FileNotFoundError:
+            pass
+    if dist.is_available() and dist.is_initialized():
+        path_list = [counter_path, prefix_counter_path, pause_flag_path]
+        dist.broadcast_object_list(path_list, src=0)
+        counter_path, prefix_counter_path, pause_flag_path = path_list
+        dist.barrier()
+    loss_sum = torch.zeros((), device=device, dtype=torch.float64)
+    byte_sum = torch.zeros((), device=device, dtype=torch.float64)
+    token_count = torch.zeros((), device=device, dtype=torch.float64)
+    t_start = time.perf_counter()
+    reusable_lora = BatchedTTTLoRA(
+        h.ttt_batch_size, base_model, h.ttt_lora_rank,
+        k_lora=h.ttt_k_lora, mlp_lora=h.ttt_mlp_lora, o_lora=h.ttt_o_lora,
+    ).to(device)
+    def _build_opt(lora):
+        return torch.optim.AdamW(
+            lora.parameters(), lr=h.ttt_lora_lr,
+            betas=(h.ttt_beta1, h.ttt_beta2),
+            eps=1e-10, weight_decay=h.ttt_weight_decay, fused=True,
+        )
+    reusable_opt = _build_opt(reusable_lora)
+    local_scored_docs = []
+    global_ttt_done = prefix_doc_limit == 0
+    while True:
+      queue_idx = _claim_next_batch(counter_path, queue_len)
+      if queue_idx >= queue_len:
+          break
+      orig_batch_idx, batch_entries = global_batches_sorted[queue_idx]
+      batch = [doc for _, doc in batch_entries]
+      bsz = len(batch)
+      prev_loss = loss_sum.item()
+      prev_bytes = byte_sum.item()
+      prev_tokens = token_count.item()
+      if bsz == reusable_lora.bsz:
+          reusable_lora.reset()
+          for s in reusable_opt.state.values():
+              for k, v in s.items():
+                  if isinstance(v, torch.Tensor):
+                      v.zero_()
+                  elif k == "step":
+                      s[k] = 0
+          cur_lora = reusable_lora
+          cur_opt = reusable_opt
+      else:
+          cur_lora = BatchedTTTLoRA(
+              bsz, base_model, h.ttt_lora_rank,
+              k_lora=h.ttt_k_lora, mlp_lora=h.ttt_mlp_lora, o_lora=h.ttt_o_lora,
+          ).to(device)
+          cur_opt = _build_opt(cur_lora)
+      pred_lens = [doc_len - 1 for _, doc_len in batch]
+      num_chunks = [(pl + chunk_size - 1) // chunk_size for pl in pred_lens]
+      max_nc = max(num_chunks)
+      num_chunks_t = torch.tensor(num_chunks, dtype=torch.int64, device=device)
+      for ci in range(max_nc):
+          active = [ci < nc for nc in num_chunks]
+          needs_train = any(ci < nc - 1 for nc in num_chunks)
+          tok_starts = torch.zeros(bsz, dtype=torch.int64)
+          tok_wls = torch.zeros(bsz, dtype=torch.int64)
+          chunk_offsets_cpu = torch.zeros(bsz, dtype=torch.int64)
+          chunk_lens_cpu = torch.zeros(bsz, dtype=torch.int64)
+          for b in range(bsz):
+              if not active[b]:
+                  continue
+              doc_start, doc_len = batch[b]
+              win_start, win_len, chunk_offset, chunk_len = _compute_chunk_window(
+                  ci, pred_lens[b], num_chunks[b], chunk_size, eval_seq_len
+              )
+              tok_starts[b] = doc_start + win_start
+              tok_wls[b] = win_len
+              chunk_offsets_cpu[b] = chunk_offset
+              chunk_lens_cpu[b] = chunk_len
+          _, context_size, chunk_offset, _ = _compute_chunk_window(
+              ci, (ci + 1) * chunk_size, ci + 1, chunk_size, eval_seq_len
+          )
+          col_idx = torch.arange(context_size + 1)
+          idx = tok_starts.unsqueeze(1) + col_idx.unsqueeze(0)
+          idx.clamp_(max=all_tokens.numel() - 1)
+          gathered_gpu = all_tokens_idx[idx].to(
+              device=device, dtype=torch.int64, non_blocking=True
+          )
+          valid = (col_idx[:context_size].unsqueeze(0) < tok_wls.unsqueeze(1)).to(
+              device, non_blocking=True
+          )
+          chunk_offsets = chunk_offsets_cpu.to(device, non_blocking=True)
+          chunk_lens = chunk_lens_cpu.to(device, non_blocking=True)
+          x = torch.where(valid, gathered_gpu[:, :context_size], 0)
+          y = torch.where(valid, gathered_gpu[:, 1 : context_size + 1], 0)
+          ctx_pos = torch.arange(context_size, device=device, dtype=torch.int64)
+          with torch.autocast(device_type="cuda", dtype=torch.bfloat16):
+              per_tok_loss = forward_ttt_train(x, y, lora=cur_lora)
+          y_bytes_arg = None
+          if val_data.caseops_enabled and val_data.val_bytes is not None:
+              y_idx = (
+                  tok_starts.unsqueeze(1)
+                  + 1
+                  + col_idx[:context_size].unsqueeze(0)
+              )
+              y_idx = y_idx.clamp_(max=val_data.val_bytes.numel() - 1)
+              y_bytes_arg = val_data.val_bytes[y_idx].to(
+                  device=device, dtype=torch.int32, non_blocking=True
+              )
+              y_bytes_arg = torch.where(
+                  valid, y_bytes_arg, torch.zeros_like(y_bytes_arg)
+              )
+          with torch.no_grad():
+              _accumulate_bpb(
+                  per_tok_loss,
+                  x,
+                  y,
+                  chunk_offsets,
+                  chunk_lens,
+                  ctx_pos,
+                  val_data.base_bytes_lut,
+                  val_data.has_leading_space_lut,
+                  val_data.is_boundary_token_lut,
+                  loss_sum,
+                  byte_sum,
+                  token_count,
+                  y_bytes=y_bytes_arg,
+              )
+          if needs_train:
+              activate_chunk_mask = (num_chunks_t - 1 > ci).float()
+              for gi in range(h.ttt_grad_steps):
+                  if gi > 0:
+                      with torch.autocast(device_type="cuda", dtype=torch.bfloat16):
+                          per_tok_loss = forward_ttt_train(x, y, lora=cur_lora)
+                  per_doc = per_tok_loss[
+                      :, chunk_offset : chunk_offset + chunk_size
+                  ].mean(dim=-1)
+                  cur_opt.zero_grad(set_to_none=True)
+                  (per_doc * activate_chunk_mask).sum().backward()
+                  cur_opt.step()
+          else:
+              del per_tok_loss
+      batch_num = orig_batch_idx + 1
+      doc_lens = [dl for _, dl in batch]
+      should_report = batch_num in eval_batch_set if eval_batch_set is not None else True
+      if should_report:
+          cur_tokens = token_count.item()
+          cur_loss_val = loss_sum.item()
+          cur_bytes_val = byte_sum.item()
+          dt = cur_tokens - prev_tokens
+          db = cur_bytes_val - prev_bytes
+          if dt > 0 and db > 0:
+              b_loss = (cur_loss_val - prev_loss) / dt
+              b_bpb = b_loss / math.log(2.0) * (dt / db)
+          else:
+              b_loss = b_bpb = 0.0
+          r_loss = cur_loss_val / max(cur_tokens, 1)
+          r_bpb = r_loss / math.log(2.0) * (cur_tokens / max(cur_bytes_val, 1))
+          elapsed = time.perf_counter() - t_start
+          log(
+              f"ttp: b{batch_num}/{queue_len} bl:{b_loss:.4f} bb:{b_bpb:.4f} "
+              f"rl:{r_loss:.4f} rb:{r_bpb:.4f} dl:{min(doc_lens)}-{max(doc_lens)} "
+              f"gd:{int(global_ttt_done)}"
+          )
+      if not global_ttt_done:
+          local_scored_docs.extend(
+              (orig_batch_idx, pos, doc_start, doc_len)
+              for pos, (doc_start, doc_len) in enumerate(batch)
+          )
+          prefix_done = _add_to_counter(prefix_counter_path, len(batch_entries))
+          if prefix_done >= current_phase_boundary:
+              try:
+                  with open(pause_flag_path, "x"):
+                      pass
+              except FileExistsError:
+                  pass
+          should_pause = os.path.exists(pause_flag_path)
+          if should_pause:
+              if dist.is_available() and dist.is_initialized():
+                  dist.barrier()
+              gathered_scored_docs = [None] * h.world_size
+              if dist.is_available() and dist.is_initialized():
+                  dist.all_gather_object(gathered_scored_docs, local_scored_docs)
+              else:
+                  gathered_scored_docs = [local_scored_docs]
+              scored_docs_for_global = []
+              for rank_docs in gathered_scored_docs:
+                  if rank_docs:
+                      scored_docs_for_global.extend(rank_docs)
+              scored_docs_for_global.sort(key=lambda x: (x[0], x[1]))
+              scored_docs_for_global = scored_docs_for_global[:current_phase_boundary]
+              scored_token_chunks = [
+                  val_data.val_tokens[doc_start : doc_start + doc_len]
+                  for _, _, doc_start, doc_len in scored_docs_for_global
+              ]
+              if scored_token_chunks:
+                  global_ttt_tokens = torch.cat(scored_token_chunks)
+              else:
+                  global_ttt_tokens = val_data.val_tokens[:0]
+              if h.rank == 0:
+                  prefix_done = 0
+                  try:
+                      with open(prefix_counter_path, "rb") as f:
+                          prefix_done = int.from_bytes(
+                              f.read(8), "little", signed=True
+                          )
+                  except FileNotFoundError:
+                      pass
+                  log(
+                      f"ttpp: phase:{current_phase + 1}/{num_phases} pd:{prefix_done} "
+                      f"gd:{len(scored_docs_for_global)} "
+                      f"t:{time.perf_counter() - t_start:.1f}s"
+                  )
+              train_val_ttt_global_sgd_distributed(
+                  h, device, val_data, base_model, global_ttt_tokens
+              )
+              for p in base_model.parameters():
+                  p.requires_grad_(False)
+              reusable_lora = BatchedTTTLoRA(
+                  h.ttt_batch_size, base_model, h.ttt_lora_rank,
+                  k_lora=h.ttt_k_lora, mlp_lora=h.ttt_mlp_lora, o_lora=h.ttt_o_lora,
+              ).to(device)
+              reusable_opt = _build_opt(reusable_lora)
+              current_phase += 1
+              if current_phase >= num_phases:
+                  global_ttt_done = True
+              else:
+                  current_phase_boundary = phase_boundaries[current_phase]
+                  if h.rank == 0:
+                      try:
+                          os.remove(pause_flag_path)
+                      except FileNotFoundError:
+                          pass
+              if dist.is_available() and dist.is_initialized():
+                  dist.barrier()
+              if h.rank == 0:
+                  log(f"ttpr: phase:{current_phase}/{num_phases} t:{time.perf_counter() - t_start:.1f}s")
+      del cur_lora, cur_opt
+    if dist.is_available() and dist.is_initialized():
+        dist.all_reduce(loss_sum, op=dist.ReduceOp.SUM)
+        dist.all_reduce(byte_sum, op=dist.ReduceOp.SUM)
+        dist.all_reduce(token_count, op=dist.ReduceOp.SUM)
+    for p in base_model.parameters():
+        p.requires_grad_(True)
+    base_model.train()
+    return _loss_bpb(loss_sum, token_count, byte_sum)
+def timed_eval(label, fn, *args, **kwargs):
+    torch.cuda.synchronize()
+    t0 = time.perf_counter()
+    val_loss, val_bpb = fn(*args, **kwargs)
+    torch.cuda.synchronize()
+    elapsed_ms = 1e3 * (time.perf_counter() - t0)
+    log(
+        f"{label} val_loss:{val_loss:.8f} val_bpb:{val_bpb:.8f} eval_time:{elapsed_ms:.0f}ms"
+    )
+    return val_loss, val_bpb
+def train_model(h, device, val_data):
+    base_model = GPT(h).to(device).bfloat16()
+    restore_fp32_params(base_model)
+    compiled_model = torch.compile(base_model, dynamic=False, fullgraph=True)
+    compiled_forward_logits = torch.compile(
+        base_model.forward_logits, dynamic=False, fullgraph=True
+    )
+    model = compiled_model
+    log(f"model_params:{sum(p.numel()for p in base_model.parameters())}")
+    optimizers = Optimizers(h, base_model)
+    train_loader = DocumentPackingLoader(h, device)
+    max_wallclock_ms = (
+        1e3 * h.max_wallclock_seconds if h.max_wallclock_seconds > 0 else None
+    )
+    if max_wallclock_ms is not None:
+        max_wallclock_ms -= h.gptq_reserve_seconds * 1e3
+        log(
+            f"gptq:reserving {h.gptq_reserve_seconds:.0f}s, effective={max_wallclock_ms:.0f}ms"
+        )
+    def training_frac(step, elapsed_ms):
+        if max_wallclock_ms is None:
+            return step / max(h.iterations, 1)
+        return elapsed_ms / max(max_wallclock_ms, 1e-09)
+    def lr_mul(frac):
+        if h.warmdown_frac <= 0:
+            return 1.0
+        if frac >= 1.0 - h.warmdown_frac:
+            return max((1.0 - frac) / h.warmdown_frac, h.min_lr)
+        return 1.0
+    def step_fn(step, lr_scale):
+        optimizers.zero_grad_all()
+        train_loss = torch.zeros((), device=device)
+        for micro_step in range(h.grad_accum_steps):
+            x, y, cu_seqlens, _max_seqlen = train_loader.next_batch(
+                h.train_batch_tokens, h.grad_accum_steps
+            )
+            with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True):
+                loss = model(x, y, cu_seqlens=cu_seqlens, max_seqlen=h.train_seq_len)
+            train_loss += loss.detach()
+            (loss / h.grad_accum_steps).backward()
+        train_loss /= h.grad_accum_steps
+        frac = (
+            min(step / h.muon_momentum_warmup_steps, 1.0)
+            if h.muon_momentum_warmup_steps > 0
+            else 1.0
+        )
+        muon_momentum = (
+            1 - frac
+        ) * h.muon_momentum_warmup_start + frac * h.muon_momentum
+        for group in optimizers.optimizer_muon.param_groups:
+            group["momentum"] = muon_momentum
+        for opt in optimizers:
+            for group in opt.param_groups:
+                group["lr"] = group["base_lr"] * lr_scale
+        if h.grad_clip_norm > 0:
+            torch.nn.utils.clip_grad_norm_(base_model.parameters(), h.grad_clip_norm)
+        optimizers.step(distributed=h.distributed)
+        return train_loss
+    if h.warmup_steps > 0:
+        initial_model_state = {
+            name: tensor.detach().cpu().clone()
+            for (name, tensor) in base_model.state_dict().items()
+        }
+        initial_optimizer_states = [
+            copy.deepcopy(opt.state_dict()) for opt in optimizers
+        ]
+        model.train()
+        num_tokens_local = h.train_batch_tokens // h.world_size
+        for blk in base_model.blocks:
+            blk.attn.rotary(num_tokens_local, device, torch.bfloat16)
+        cu_bucket_size = train_loader.cu_bucket_size
+        warmup_cu_buckets = tuple(cu_bucket_size * i for i in range(1, 5))
+        warmup_cu_iters = 3
+        x, y, cu_seqlens, _ = train_loader.next_batch(
+            h.train_batch_tokens, h.grad_accum_steps
+        )
+        log(f"warmup_cu_buckets:{','.join(str(b) for b in warmup_cu_buckets)} iters_each:{warmup_cu_iters}")
+        def _run_cu_bucket_warmup():
+            for bucket_len in warmup_cu_buckets:
+                boundaries = list(range(0, x.size(1), max(h.train_seq_len, 1)))
+                if boundaries[-1] != x.size(1):
+                    boundaries.append(x.size(1))
+                cu = torch.full((bucket_len,), x.size(1), dtype=torch.int32, device=device)
+                cu[: len(boundaries)] = torch.tensor(boundaries, dtype=torch.int32, device=device)
+                for _ in range(warmup_cu_iters):
+                    optimizers.zero_grad_all()
+                    with torch.autocast(device_type="cuda", dtype=torch.bfloat16, enabled=True):
+                        wloss = model(x, y, cu_seqlens=cu, max_seqlen=h.train_seq_len)
+                    (wloss / h.grad_accum_steps).backward()
+            optimizers.zero_grad_all()
+        _run_cu_bucket_warmup()
+        if h.num_loops > 0:
+            base_model.looping_active = True
+            _run_cu_bucket_warmup()
+            base_model.looping_active = False
+        for warmup_step in range(h.warmup_steps):
+            step_fn(warmup_step, 1.0)
+            if (
+                warmup_step <= 5
+                or (warmup_step + 1) % 10 == 0
+                or warmup_step + 1 == h.warmup_steps
+            ):
+                log(f"warmup_step: {warmup_step+1}/{h.warmup_steps}")
+        if h.num_loops > 0:
+            base_model.looping_active = True
+            log(
+                f"loop_warmup:enabled encoder:{base_model.encoder_indices} decoder:{base_model.decoder_indices}"
+            )
+            for warmup_step in range(h.warmup_steps):
+                step_fn(warmup_step, 1.0)
+                if (
+                    warmup_step <= 5
+                    or (warmup_step + 1) % 10 == 0
+                    or warmup_step + 1 == h.warmup_steps
+                ):
+                    log(f"loop_warmup_step: {warmup_step+1}/{h.warmup_steps}")
+            base_model.looping_active = False
+        base_model.load_state_dict(initial_model_state, strict=True)
+        for (opt, state) in zip(optimizers, initial_optimizer_states, strict=True):
+            opt.load_state_dict(state)
+        optimizers.zero_grad_all()
+        train_loader = DocumentPackingLoader(h, device)
+    ema_state = {
+        name: t.detach().float().clone()
+        for (name, t) in base_model.state_dict().items()
+    }
+    ema_decay = h.ema_decay
+    training_time_ms = 0.0
+    stop_after_step = None
+    torch.cuda.synchronize()
+    t0 = time.perf_counter()
+    step = 0
+    while True:
+        last_step = (
+            step == h.iterations
+            or stop_after_step is not None
+            and step >= stop_after_step
+        )
+        should_validate = (
+            last_step or h.val_loss_every > 0 and step % h.val_loss_every == 0
+        )
+        if should_validate:
+            torch.cuda.synchronize()
+            training_time_ms += 1e3 * (time.perf_counter() - t0)
+            val_loss, val_bpb = eval_val(
+                h, device, val_data, model, compiled_forward_logits
+            )
+            log(
+                f"{step}/{h.iterations} val_loss: {val_loss:.4f} val_bpb: {val_bpb:.4f}"
+            )
+            torch.cuda.synchronize()
+            t0 = time.perf_counter()
+        if last_step:
+            if stop_after_step is not None and step < h.iterations:
+                log(
+                    f"stopping_early: wallclock_cap train_time: {training_time_ms:.0f}ms step: {step}/{h.iterations}"
+                )
+            break
+        elapsed_ms = training_time_ms + 1e3 * (time.perf_counter() - t0)
+        frac = training_frac(step, elapsed_ms)
+        scale = lr_mul(frac)
+        if (
+            h.num_loops > 0
+            and not base_model.looping_active
+            and frac >= h.enable_looping_at
+        ):
+            base_model.looping_active = True
+            log(
+                f"layer_loop:enabled step:{step} frac:{frac:.3f} encoder:{base_model.encoder_indices} decoder:{base_model.decoder_indices}"
+            )
+        train_loss = step_fn(step, scale)
+        with torch.no_grad():
+            for (name, t) in base_model.state_dict().items():
+                ema_state[name].mul_(ema_decay).add_(
+                    t.detach().float(), alpha=1.0 - ema_decay
+                )
+        step += 1
+        approx_training_time_ms = training_time_ms + 1e3 * (time.perf_counter() - t0)
+        should_log_train = h.train_log_every > 0 and (
+            step <= 5 or step % h.train_log_every == 0 or stop_after_step is not None
+        )
+        if should_log_train:
+            tok_per_sec = step * h.train_batch_tokens / (approx_training_time_ms / 1e3)
+            log(
+                f"{step}/{h.iterations} train_loss: {train_loss.item():.4f} train_time: {approx_training_time_ms/60000:.1f}m tok/s: {tok_per_sec:.0f}"
+            )
+        reached_cap = (
+            max_wallclock_ms is not None and approx_training_time_ms >= max_wallclock_ms
+        )
+        if h.distributed and max_wallclock_ms is not None:
+            reached_cap_tensor = torch.tensor(int(reached_cap), device=device)
+            dist.all_reduce(reached_cap_tensor, op=dist.ReduceOp.MAX)
+            reached_cap = bool(reached_cap_tensor.item())
+        if stop_after_step is None and reached_cap:
+            stop_after_step = step
+    log(
+        f"peak memory allocated: {torch.cuda.max_memory_allocated()//1024//1024} MiB reserved: {torch.cuda.max_memory_reserved()//1024//1024} MiB"
+    )
+    log("ema:applying EMA weights")
+    current_state = base_model.state_dict()
+    avg_state = {
+        name: t.to(dtype=current_state[name].dtype) for (name, t) in ema_state.items()
+    }
+    base_model.load_state_dict(avg_state, strict=True)
+    return base_model, compiled_model, compiled_forward_logits
+def train_and_eval(h, device):
+    random.seed(h.seed)
+    np.random.seed(h.seed)
+    torch.manual_seed(h.seed)
+    torch.cuda.manual_seed_all(h.seed)
+    if h.artifact_dir and h.is_main_process:
+        os.makedirs(h.artifact_dir, exist_ok=True)
+    val_data = ValidationData(h, device)
+    log(
+        f"train_shards: {len(list(Path(h.datasets_dir).resolve().glob('fineweb_train_*.bin')))}"
+    )
+    log(f"val_tokens: {val_data.val_tokens.numel()-1}")
+    ttt_eval_only = os.environ.get("TTT_EVAL_ONLY", "0") == "1"
+    if ttt_eval_only:
+        log("TTT_EVAL_ONLY=1 — skipping training + GPTQ, loading saved artifact for TTT eval")
+        log(f"ttt_lora_alpha: {BatchedLinearLoRA._ALPHA}")
+        log(f"ttt_warm_start_a: {BatchedLinearLoRA._WARM_START_A}")
+        log(f"ttt_weight_decay: {h.ttt_weight_decay}")
+    else:
+        base_model, compiled_model, compiled_forward_logits = train_model(
+            h, device, val_data
+        )
+        torch._dynamo.reset()
+        timed_eval(
+            "diagnostic pre-quantization post-ema",
+            eval_val,
+            h,
+            device,
+            val_data,
+            compiled_model,
+            compiled_forward_logits,
+        )
+        if os.environ.get("PREQUANT_ONLY", "0") == "1":
+            log("PREQUANT_ONLY=1 — skipping serialize/GPTQ/post-quant eval/TTT")
+            return
+        serialize(h, base_model, Path(__file__).read_text(encoding="utf-8"))
+        if h.distributed:
+            dist.barrier()
+    eval_model = deserialize(h, device)
+    if h.num_loops > 0:
+        eval_model.looping_active = True
+    if not ttt_eval_only:
+        compiled_model = torch.compile(eval_model, dynamic=False, fullgraph=True)
+        compiled_forward_logits = torch.compile(
+            eval_model.forward_logits, dynamic=False, fullgraph=True
+        )
+        timed_eval(
+            "diagnostic quantized",
+            eval_val,
+            h,
+            device,
+            val_data,
+            compiled_model,
+            compiled_forward_logits,
+        )
+        del eval_model
+    if h.ttt_enabled:
+        if not ttt_eval_only:
+            del compiled_model
+        if ttt_eval_only:
+            del eval_model
+        torch._dynamo.reset()
+        torch.cuda.empty_cache()
+        ttt_model = deserialize(h, device)
+        if h.num_loops > 0:
+            ttt_model.looping_active = True
+        for p in ttt_model.parameters():
+            p.requires_grad_(False)
+        if h.rope_yarn:
+            _yarn_seqlen = h.train_batch_tokens // h.grad_accum_steps
+            for block in ttt_model.blocks:
+                block.attn.rotary(_yarn_seqlen, device, torch.bfloat16)
+        else:
+            for block in ttt_model.blocks:
+                block.attn.rotary._cos_cached = None
+                block.attn.rotary._sin_cached = None
+                block.attn.rotary._seq_len_cached = 0
+                block.attn.rotary(h.ttt_eval_seq_len, device, torch.bfloat16)
+        def _fwd_ttt_inner(input_ids, target_ids, lora):
+            return ttt_model.forward_ttt(input_ids, target_ids, lora=lora)
+        _fwd_ttt_compiled_inner = None
+        def _fwd_ttt(input_ids, target_ids, lora):
+            nonlocal _fwd_ttt_compiled_inner
+            if _fwd_ttt_compiled_inner is None:
+                _fwd_ttt_compiled_inner = torch.compile(_fwd_ttt_inner, dynamic=True)
+            return _fwd_ttt_compiled_inner(input_ids, target_ids, lora=lora)
+        fwd_ttt_compiled = _fwd_ttt
+        log(f"ttt_lora:warming up compile (random tokens, no val data)")
+        global BOS_ID
+        if BOS_ID is None:
+            BOS_ID = 1
+        t_warmup = time.perf_counter()
+        warmup_bszes = [h.ttt_batch_size]
+        for bsz in warmup_bszes:
+            wl = BatchedTTTLoRA(
+                bsz, ttt_model, h.ttt_lora_rank,
+                k_lora=h.ttt_k_lora, mlp_lora=h.ttt_mlp_lora, o_lora=h.ttt_o_lora,
+            ).to(device)
+            wo = torch.optim.AdamW(
+                wl.parameters(),
+                lr=h.ttt_lora_lr,
+                betas=(h.ttt_beta1, h.ttt_beta2),
+                eps=1e-10,
+                weight_decay=h.ttt_weight_decay,
+                fused=True,
+            )
+            for ctx_len in (h.ttt_chunk_size, h.ttt_eval_seq_len):
+                xw = torch.randint(0, h.vocab_size, (bsz, ctx_len), device=device, dtype=torch.int64)
+                yw = torch.randint(0, h.vocab_size, (bsz, ctx_len), device=device, dtype=torch.int64)
+                with torch.autocast(device_type="cuda", dtype=torch.bfloat16):
+                    ptl = fwd_ttt_compiled(xw, yw, lora=wl)
+                ptl[:, : min(h.ttt_chunk_size, ctx_len)].mean(dim=-1).sum().backward()
+                wo.step()
+                wo.zero_grad(set_to_none=True)
+            del wl, wo
+        torch.cuda.empty_cache()
+        compile_elapsed = time.perf_counter() - t_warmup
+        log(f"ttt_lora:compile warmup done ({compile_elapsed:.1f}s)")
+        log("\nbeginning TTT eval timer")
+        torch.cuda.synchronize()
+        t_ttt = time.perf_counter()
+        ttt_val_loss, ttt_val_bpb = eval_val_ttt_phased(
+            h, ttt_model, device, val_data, forward_ttt_train=fwd_ttt_compiled
+        )
+        torch.cuda.synchronize()
+        ttt_eval_elapsed = time.perf_counter() - t_ttt
+        log(
+            "quantized_ttt_phased "
+            f"val_loss:{ttt_val_loss:.8f} val_bpb:{ttt_val_bpb:.8f} "
+            f"eval_time:{1e3*ttt_eval_elapsed:.0f}ms"
+        )
+        log(f"total_eval_time:{ttt_eval_elapsed:.1f}s")
+        del ttt_model
+def main():
+    world_size = int(os.environ.get("WORLD_SIZE", "1"))
+    local_rank = int(os.environ.get("LOCAL_RANK", "0"))
+    distributed = "RANK" in os.environ and "WORLD_SIZE" in os.environ
+    if not torch.cuda.is_available():
+        raise RuntimeError("CUDA is required")
+    if world_size <= 0:
+        raise ValueError(f"WORLD_SIZE must be positive, got {world_size}")
+    if 8 % world_size != 0:
+        raise ValueError(
+            f"WORLD_SIZE={world_size} must divide 8 so grad_accum_steps stays integral"
+        )
+    device = torch.device("cuda", local_rank)
+    torch.cuda.set_device(device)
+    if distributed:
+        dist.init_process_group(backend="nccl", device_id=device)
+        dist.barrier()
+    torch.backends.cuda.matmul.allow_tf32 = True
+    torch.backends.cudnn.allow_tf32 = True
+    torch.set_float32_matmul_precision("high")
+    from torch.backends.cuda import (
+        enable_cudnn_sdp,
+        enable_flash_sdp,
+        enable_math_sdp,
+        enable_mem_efficient_sdp,
+    )
+    enable_cudnn_sdp(False)
+    enable_flash_sdp(True)
+    enable_mem_efficient_sdp(False)
+    enable_math_sdp(False)
+    torch._dynamo.config.optimize_ddp = False
+    torch._dynamo.config.cache_size_limit = 16
+    h = Hyperparameters()
+    set_logging_hparams(h)
+    if h.is_main_process:
+        os.makedirs(h.artifact_dir if h.artifact_dir else "logs", exist_ok=True)
+        os.makedirs(os.path.dirname(h.model_path) or "models", exist_ok=True)
+        os.makedirs(os.path.dirname(h.quantized_model_path) or "models", exist_ok=True)
+    train_and_eval(h, device)
+    if distributed:
+        dist.destroy_process_group()
+if __name__ == "__main__":
+    main()
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed1334.log b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed1334.log
new file mode 100644
index 0000000000..7f6821ad3c
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed1334.log
@@ -0,0 +1,1185 @@
+W0429 17:52:36.768000 100862 torch/distributed/run.py:803] 
+W0429 17:52:36.768000 100862 torch/distributed/run.py:803] *****************************************
+W0429 17:52:36.768000 100862 torch/distributed/run.py:803] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0429 17:52:36.768000 100862 torch/distributed/run.py:803] *****************************************
+train_shards: 80
+val_tokens: 47851520
+model_params:35945671
+gptq:reserving 16s, effective=584000ms
+warmup_cu_buckets:64,128,192,256 iters_each:3
+warmup_step: 1/20
+warmup_step: 2/20
+warmup_step: 3/20
+warmup_step: 4/20
+warmup_step: 5/20
+warmup_step: 6/20
+warmup_step: 10/20
+warmup_step: 20/20
+loop_warmup:enabled encoder:[0, 1, 2, 3, 4, 5, 3, 4] decoder:[5, 3, 4, 5, 6, 7, 8, 9, 10]
+loop_warmup_step: 1/20
+loop_warmup_step: 2/20
+loop_warmup_step: 3/20
+loop_warmup_step: 4/20
+loop_warmup_step: 5/20
+loop_warmup_step: 6/20
+loop_warmup_step: 10/20
+loop_warmup_step: 20/20
+0/20000 val_loss: 9.0007 val_bpb: 4.1127
+1/20000 train_loss: 9.0016 train_time: 0.0m tok/s: 12275171
+2/20000 train_loss: 12.8589 train_time: 0.0m tok/s: 11509645
+3/20000 train_loss: 10.1873 train_time: 0.0m tok/s: 10242533
+4/20000 train_loss: 8.6642 train_time: 0.0m tok/s: 9731774
+5/20000 train_loss: 7.9253 train_time: 0.0m tok/s: 9432809
+500/20000 train_loss: 2.5705 train_time: 0.8m tok/s: 8293624
+1000/20000 train_loss: 2.7960 train_time: 1.6m tok/s: 8266250
+1500/20000 train_loss: 2.6373 train_time: 2.4m tok/s: 8257371
+2000/20000 train_loss: 2.6638 train_time: 3.2m tok/s: 8255091
+layer_loop:enabled step:2146 frac:0.350 encoder:[0, 1, 2, 3, 4, 5, 3, 4] decoder:[5, 3, 4, 5, 6, 7, 8, 9, 10]
+2500/20000 train_loss: 2.5502 train_time: 4.2m tok/s: 7737102
+3000/20000 train_loss: 2.5623 train_time: 5.4m tok/s: 7276150
+3500/20000 train_loss: 2.5652 train_time: 6.6m tok/s: 6957307
+4000/20000 train_loss: 2.4035 train_time: 7.8m tok/s: 6754329
+4000/20000 val_loss: 2.4284 val_bpb: 1.1096
+4500/20000 train_loss: 2.2782 train_time: 9.0m tok/s: 6584803
+4833/20000 val_loss: 2.3584 val_bpb: 1.0776
+stopping_early: wallclock_cap train_time: 584153ms step: 4833/20000
+peak memory allocated: 41709 MiB reserved: 47026 MiB
+ema:applying EMA weights
+diagnostic pre-quantization post-ema val_loss:2.33536211 val_bpb:1.06710006 eval_time:6770ms
+Serialized model: 135419669 bytes
+Code size (uncompressed): 138859 bytes
+Code size (compressed): 31095 bytes
+GPTQ:collecting Hessians from calibration data...
+GPTQ:collected 67 Hessians in 3.5s
+Quantized weights:
+  gate_int8_row: blocks.attn.attn_gate_w
+  gptq (int6): blocks.attn.c_k.weight, blocks.attn.c_q.weight, blocks.attn.c_v.weight, blocks.attn.proj.weight, blocks.mlp.fc.weight, blocks.mlp.proj.weight
+  gptq (int7)+lqer_asym: tok_emb.weight
+  passthrough (float16): blocks.attn.q_gain, blocks.attn_scale, blocks.mlp_scale, blocks.resid_mix, parallel_post_lambdas, parallel_resid_lambdas, skip_gates, skip_weights, smear_gate.weight, smear_lambda
+GPTQ:quantized weights in 12.13s
+Serialized model quantized+brotli: 15916569 bytes
+Total submission size quantized+brotli: 15947664 bytes
+diagnostic quantized val_loss:2.35445830 val_bpb:1.07582570 eval_time:10246ms
+ttt_lora:warming up compile (random tokens, no val data)
+ttt_lora:compile warmup done (171.6s)
+
+beginning TTT eval timer
+ttt_phased: total_docs:50000 prefix_docs:2000 suffix_docs:48000 num_phases:3 boundaries:[666, 1333, 2000]
+ttp: b3121/3125 bl:2.2891 bb:1.1129 rl:2.2891 rb:1.1129 dl:17258-19362 gd:0
+ttp: b3114/3125 bl:2.2192 bb:1.0703 rl:2.2623 rb:1.0965 dl:11159-11603 gd:0
+ttp: b3108/3125 bl:2.3281 bb:1.0769 rl:2.2778 rb:1.0917 dl:9027-9229 gd:0
+ttp: b3100/3125 bl:2.3313 bb:1.0745 rl:2.2864 rb:1.0888 dl:7313-7524 gd:0
+ttp: b3093/3125 bl:2.3174 bb:1.0906 rl:2.2903 rb:1.0891 dl:6447-6522 gd:0
+ttp: b3088/3125 bl:2.3090 bb:1.1104 rl:2.2922 rb:1.0912 dl:6007-6095 gd:0
+ttp: b3081/3125 bl:2.3322 bb:1.0253 rl:2.2956 rb:1.0851 dl:5523-5586 gd:0
+ttpp: phase:1/3 pd:784 gd:666 t:205.8s
+tttg: c1/158 lr:0.001000 t:0.3s
+tttg: c2/158 lr:0.001000 t:0.3s
+tttg: c3/158 lr:0.001000 t:0.4s
+tttg: c4/158 lr:0.000999 t:0.5s
+tttg: c5/158 lr:0.000998 t:0.5s
+tttg: c6/158 lr:0.000997 t:0.6s
+tttg: c7/158 lr:0.000996 t:0.7s
+tttg: c8/158 lr:0.000995 t:0.7s
+tttg: c9/158 lr:0.000994 t:0.8s
+tttg: c10/158 lr:0.000992 t:0.9s
+tttg: c11/158 lr:0.000990 t:0.9s
+tttg: c12/158 lr:0.000988 t:1.0s
+tttg: c13/158 lr:0.000986 t:1.1s
+tttg: c14/158 lr:0.000983 t:1.1s
+tttg: c15/158 lr:0.000981 t:1.2s
+tttg: c16/158 lr:0.000978 t:1.3s
+tttg: c17/158 lr:0.000975 t:1.3s
+tttg: c18/158 lr:0.000971 t:1.4s
+tttg: c19/158 lr:0.000968 t:1.5s
+tttg: c20/158 lr:0.000964 t:1.5s
+tttg: c21/158 lr:0.000960 t:1.6s
+tttg: c22/158 lr:0.000957 t:1.7s
+tttg: c23/158 lr:0.000952 t:1.7s
+tttg: c24/158 lr:0.000948 t:1.8s
+tttg: c25/158 lr:0.000943 t:1.8s
+tttg: c26/158 lr:0.000939 t:1.9s
+tttg: c27/158 lr:0.000934 t:2.0s
+tttg: c28/158 lr:0.000929 t:2.0s
+tttg: c29/158 lr:0.000924 t:2.1s
+tttg: c30/158 lr:0.000918 t:2.2s
+tttg: c31/158 lr:0.000913 t:2.2s
+tttg: c32/158 lr:0.000907 t:2.3s
+tttg: c33/158 lr:0.000901 t:2.4s
+tttg: c34/158 lr:0.000895 t:2.4s
+tttg: c35/158 lr:0.000889 t:2.5s
+tttg: c36/158 lr:0.000882 t:2.6s
+tttg: c37/158 lr:0.000876 t:2.6s
+tttg: c38/158 lr:0.000869 t:2.7s
+tttg: c39/158 lr:0.000862 t:2.8s
+tttg: c40/158 lr:0.000855 t:2.8s
+tttg: c41/158 lr:0.000848 t:2.9s
+tttg: c42/158 lr:0.000841 t:3.0s
+tttg: c43/158 lr:0.000834 t:3.0s
+tttg: c44/158 lr:0.000826 t:3.1s
+tttg: c45/158 lr:0.000818 t:3.2s
+tttg: c46/158 lr:0.000811 t:3.2s
+tttg: c47/158 lr:0.000803 t:3.3s
+tttg: c48/158 lr:0.000795 t:3.3s
+tttg: c49/158 lr:0.000787 t:3.4s
+tttg: c50/158 lr:0.000778 t:3.5s
+tttg: c51/158 lr:0.000770 t:3.5s
+tttg: c52/158 lr:0.000761 t:3.6s
+tttg: c53/158 lr:0.000753 t:3.7s
+tttg: c54/158 lr:0.000744 t:3.7s
+tttg: c55/158 lr:0.000735 t:3.8s
+tttg: c56/158 lr:0.000727 t:3.9s
+tttg: c57/158 lr:0.000718 t:3.9s
+tttg: c58/158 lr:0.000709 t:4.0s
+tttg: c59/158 lr:0.000699 t:4.1s
+tttg: c60/158 lr:0.000690 t:4.1s
+tttg: c61/158 lr:0.000681 t:4.2s
+tttg: c62/158 lr:0.000672 t:4.3s
+tttg: c63/158 lr:0.000662 t:4.3s
+tttg: c64/158 lr:0.000653 t:4.4s
+tttg: c65/158 lr:0.000643 t:4.5s
+tttg: c66/158 lr:0.000633 t:4.5s
+tttg: c67/158 lr:0.000624 t:4.6s
+tttg: c68/158 lr:0.000614 t:4.7s
+tttg: c69/158 lr:0.000604 t:4.7s
+tttg: c70/158 lr:0.000594 t:4.8s
+tttg: c71/158 lr:0.000585 t:4.9s
+tttg: c72/158 lr:0.000575 t:4.9s
+tttg: c73/158 lr:0.000565 t:5.0s
+tttg: c74/158 lr:0.000555 t:5.1s
+tttg: c75/158 lr:0.000545 t:5.1s
+tttg: c76/158 lr:0.000535 t:5.2s
+tttg: c77/158 lr:0.000525 t:5.3s
+tttg: c78/158 lr:0.000515 t:5.3s
+tttg: c79/158 lr:0.000505 t:5.4s
+tttg: c80/158 lr:0.000495 t:5.5s
+tttg: c81/158 lr:0.000485 t:5.5s
+tttg: c82/158 lr:0.000475 t:5.6s
+tttg: c83/158 lr:0.000465 t:5.7s
+tttg: c84/158 lr:0.000455 t:5.7s
+tttg: c85/158 lr:0.000445 t:5.8s
+tttg: c86/158 lr:0.000435 t:5.9s
+tttg: c87/158 lr:0.000425 t:5.9s
+tttg: c88/158 lr:0.000415 t:6.0s
+tttg: c89/158 lr:0.000406 t:6.0s
+tttg: c90/158 lr:0.000396 t:6.1s
+tttg: c91/158 lr:0.000386 t:6.2s
+tttg: c92/158 lr:0.000376 t:6.2s
+tttg: c93/158 lr:0.000367 t:6.3s
+tttg: c94/158 lr:0.000357 t:6.4s
+tttg: c95/158 lr:0.000347 t:6.4s
+tttg: c96/158 lr:0.000338 t:6.5s
+tttg: c97/158 lr:0.000328 t:6.6s
+tttg: c98/158 lr:0.000319 t:6.6s
+tttg: c99/158 lr:0.000310 t:6.7s
+tttg: c100/158 lr:0.000301 t:6.8s
+tttg: c101/158 lr:0.000291 t:6.8s
+tttg: c102/158 lr:0.000282 t:6.9s
+tttg: c103/158 lr:0.000273 t:7.0s
+tttg: c104/158 lr:0.000265 t:7.0s
+tttg: c105/158 lr:0.000256 t:7.1s
+tttg: c106/158 lr:0.000247 t:7.2s
+tttg: c107/158 lr:0.000239 t:7.2s
+tttg: c108/158 lr:0.000230 t:7.3s
+tttg: c109/158 lr:0.000222 t:7.4s
+tttg: c110/158 lr:0.000213 t:7.4s
+tttg: c111/158 lr:0.000205 t:7.5s
+tttg: c112/158 lr:0.000197 t:7.6s
+tttg: c113/158 lr:0.000189 t:7.6s
+tttg: c114/158 lr:0.000182 t:7.7s
+tttg: c115/158 lr:0.000174 t:7.7s
+tttg: c116/158 lr:0.000166 t:7.8s
+tttg: c117/158 lr:0.000159 t:7.9s
+tttg: c118/158 lr:0.000152 t:7.9s
+tttg: c119/158 lr:0.000145 t:8.0s
+tttg: c120/158 lr:0.000138 t:8.1s
+tttg: c121/158 lr:0.000131 t:8.1s
+tttg: c122/158 lr:0.000124 t:8.2s
+tttg: c123/158 lr:0.000118 t:8.3s
+tttg: c124/158 lr:0.000111 t:8.3s
+tttg: c125/158 lr:0.000105 t:8.4s
+tttg: c126/158 lr:0.000099 t:8.5s
+tttg: c127/158 lr:0.000093 t:8.5s
+tttg: c128/158 lr:0.000087 t:8.6s
+tttg: c129/158 lr:0.000082 t:8.7s
+tttg: c130/158 lr:0.000076 t:8.7s
+tttg: c131/158 lr:0.000071 t:8.8s
+tttg: c132/158 lr:0.000066 t:8.9s
+tttg: c133/158 lr:0.000061 t:8.9s
+tttg: c134/158 lr:0.000057 t:9.0s
+tttg: c135/158 lr:0.000052 t:9.1s
+tttg: c136/158 lr:0.000048 t:9.1s
+tttg: c137/158 lr:0.000043 t:9.2s
+tttg: c138/158 lr:0.000040 t:9.3s
+tttg: c139/158 lr:0.000036 t:9.3s
+tttg: c140/158 lr:0.000032 t:9.4s
+tttg: c141/158 lr:0.000029 t:9.5s
+tttg: c142/158 lr:0.000025 t:9.5s
+tttg: c143/158 lr:0.000022 t:9.6s
+tttg: c144/158 lr:0.000019 t:9.7s
+tttg: c145/158 lr:0.000017 t:9.7s
+tttg: c146/158 lr:0.000014 t:9.8s
+tttg: c147/158 lr:0.000012 t:9.8s
+tttg: c148/158 lr:0.000010 t:9.9s
+tttg: c149/158 lr:0.000008 t:10.0s
+tttg: c150/158 lr:0.000006 t:10.0s
+tttg: c151/158 lr:0.000005 t:10.1s
+tttg: c152/158 lr:0.000004 t:10.2s
+tttg: c153/158 lr:0.000003 t:10.2s
+tttg: c154/158 lr:0.000002 t:10.3s
+tttg: c155/158 lr:0.000001 t:10.4s
+tttg: c156/158 lr:0.000000 t:10.4s
+tttg: c157/158 lr:0.000000 t:10.5s
+ttpr: phase:1/3 t:216.7s
+ttp: b3075/3125 bl:2.4111 bb:1.1553 rl:2.3043 rb:1.0903 dl:5190-5230 gd:0
+ttp: b3062/3125 bl:2.3011 bb:1.0738 rl:2.3041 rb:1.0893 dl:4558-4601 gd:0
+ttp: b3054/3125 bl:2.2352 bb:1.0817 rl:2.3003 rb:1.0889 dl:4309-4331 gd:0
+ttp: b3046/3125 bl:2.2456 bb:1.0413 rl:2.2976 rb:1.0865 dl:4058-4077 gd:0
+ttp: b3038/3125 bl:2.4709 bb:1.0720 rl:2.3053 rb:1.0858 dl:3843-3864 gd:0
+ttpp: phase:2/3 pd:1456 gd:1333 t:334.7s
+tttg: c1/247 lr:0.001000 t:0.1s
+tttg: c2/247 lr:0.001000 t:0.1s
+tttg: c3/247 lr:0.001000 t:0.2s
+tttg: c4/247 lr:0.001000 t:0.3s
+tttg: c5/247 lr:0.000999 t:0.3s
+tttg: c6/247 lr:0.000999 t:0.4s
+tttg: c7/247 lr:0.000999 t:0.5s
+tttg: c8/247 lr:0.000998 t:0.5s
+tttg: c9/247 lr:0.000997 t:0.6s
+tttg: c10/247 lr:0.000997 t:0.7s
+tttg: c11/247 lr:0.000996 t:0.7s
+tttg: c12/247 lr:0.000995 t:0.8s
+tttg: c13/247 lr:0.000994 t:0.8s
+tttg: c14/247 lr:0.000993 t:0.9s
+tttg: c15/247 lr:0.000992 t:1.0s
+tttg: c16/247 lr:0.000991 t:1.0s
+tttg: c17/247 lr:0.000990 t:1.1s
+tttg: c18/247 lr:0.000988 t:1.2s
+tttg: c19/247 lr:0.000987 t:1.2s
+tttg: c20/247 lr:0.000985 t:1.3s
+tttg: c21/247 lr:0.000984 t:1.4s
+tttg: c22/247 lr:0.000982 t:1.4s
+tttg: c23/247 lr:0.000980 t:1.5s
+tttg: c24/247 lr:0.000979 t:1.6s
+tttg: c25/247 lr:0.000977 t:1.6s
+tttg: c26/247 lr:0.000975 t:1.7s
+tttg: c27/247 lr:0.000973 t:1.8s
+tttg: c28/247 lr:0.000971 t:1.8s
+tttg: c29/247 lr:0.000968 t:1.9s
+tttg: c30/247 lr:0.000966 t:1.9s
+tttg: c31/247 lr:0.000964 t:2.0s
+tttg: c32/247 lr:0.000961 t:2.1s
+tttg: c33/247 lr:0.000959 t:2.1s
+tttg: c34/247 lr:0.000956 t:2.2s
+tttg: c35/247 lr:0.000954 t:2.3s
+tttg: c36/247 lr:0.000951 t:2.3s
+tttg: c37/247 lr:0.000948 t:2.4s
+tttg: c38/247 lr:0.000945 t:2.5s
+tttg: c39/247 lr:0.000942 t:2.5s
+tttg: c40/247 lr:0.000939 t:2.6s
+tttg: c41/247 lr:0.000936 t:2.7s
+tttg: c42/247 lr:0.000933 t:2.7s
+tttg: c43/247 lr:0.000930 t:2.8s
+tttg: c44/247 lr:0.000926 t:2.8s
+tttg: c45/247 lr:0.000923 t:2.9s
+tttg: c46/247 lr:0.000920 t:3.0s
+tttg: c47/247 lr:0.000916 t:3.0s
+tttg: c48/247 lr:0.000913 t:3.1s
+tttg: c49/247 lr:0.000909 t:3.2s
+tttg: c50/247 lr:0.000905 t:3.2s
+tttg: c51/247 lr:0.000901 t:3.3s
+tttg: c52/247 lr:0.000898 t:3.4s
+tttg: c53/247 lr:0.000894 t:3.4s
+tttg: c54/247 lr:0.000890 t:3.5s
+tttg: c55/247 lr:0.000886 t:3.6s
+tttg: c56/247 lr:0.000882 t:3.6s
+tttg: c57/247 lr:0.000877 t:3.7s
+tttg: c58/247 lr:0.000873 t:3.8s
+tttg: c59/247 lr:0.000869 t:3.8s
+tttg: c60/247 lr:0.000865 t:3.9s
+tttg: c61/247 lr:0.000860 t:3.9s
+tttg: c62/247 lr:0.000856 t:4.0s
+tttg: c63/247 lr:0.000851 t:4.1s
+tttg: c64/247 lr:0.000847 t:4.1s
+tttg: c65/247 lr:0.000842 t:4.2s
+tttg: c66/247 lr:0.000837 t:4.3s
+tttg: c67/247 lr:0.000833 t:4.3s
+tttg: c68/247 lr:0.000828 t:4.4s
+tttg: c69/247 lr:0.000823 t:4.5s
+tttg: c70/247 lr:0.000818 t:4.5s
+tttg: c71/247 lr:0.000813 t:4.6s
+tttg: c72/247 lr:0.000808 t:4.7s
+tttg: c73/247 lr:0.000803 t:4.7s
+tttg: c74/247 lr:0.000798 t:4.8s
+tttg: c75/247 lr:0.000793 t:4.8s
+tttg: c76/247 lr:0.000788 t:4.9s
+tttg: c77/247 lr:0.000782 t:5.0s
+tttg: c78/247 lr:0.000777 t:5.0s
+tttg: c79/247 lr:0.000772 t:5.1s
+tttg: c80/247 lr:0.000766 t:5.2s
+tttg: c81/247 lr:0.000761 t:5.2s
+tttg: c82/247 lr:0.000756 t:5.3s
+tttg: c83/247 lr:0.000750 t:5.4s
+tttg: c84/247 lr:0.000744 t:5.4s
+tttg: c85/247 lr:0.000739 t:5.5s
+tttg: c86/247 lr:0.000733 t:5.6s
+tttg: c87/247 lr:0.000728 t:5.6s
+tttg: c88/247 lr:0.000722 t:5.7s
+tttg: c89/247 lr:0.000716 t:5.7s
+tttg: c90/247 lr:0.000710 t:5.8s
+tttg: c91/247 lr:0.000705 t:5.9s
+tttg: c92/247 lr:0.000699 t:5.9s
+tttg: c93/247 lr:0.000693 t:6.0s
+tttg: c94/247 lr:0.000687 t:6.1s
+tttg: c95/247 lr:0.000681 t:6.1s
+tttg: c96/247 lr:0.000675 t:6.2s
+tttg: c97/247 lr:0.000669 t:6.3s
+tttg: c98/247 lr:0.000663 t:6.3s
+tttg: c99/247 lr:0.000657 t:6.4s
+tttg: c100/247 lr:0.000651 t:6.5s
+tttg: c101/247 lr:0.000645 t:6.5s
+tttg: c102/247 lr:0.000639 t:6.6s
+tttg: c103/247 lr:0.000632 t:6.7s
+tttg: c104/247 lr:0.000626 t:6.7s
+tttg: c105/247 lr:0.000620 t:6.8s
+tttg: c106/247 lr:0.000614 t:6.8s
+tttg: c107/247 lr:0.000608 t:6.9s
+tttg: c108/247 lr:0.000601 t:7.0s
+tttg: c109/247 lr:0.000595 t:7.0s
+tttg: c110/247 lr:0.000589 t:7.1s
+tttg: c111/247 lr:0.000583 t:7.2s
+tttg: c112/247 lr:0.000576 t:7.2s
+tttg: c113/247 lr:0.000570 t:7.3s
+tttg: c114/247 lr:0.000564 t:7.4s
+tttg: c115/247 lr:0.000557 t:7.4s
+tttg: c116/247 lr:0.000551 t:7.5s
+tttg: c117/247 lr:0.000545 t:7.6s
+tttg: c118/247 lr:0.000538 t:7.6s
+tttg: c119/247 lr:0.000532 t:7.7s
+tttg: c120/247 lr:0.000526 t:7.7s
+tttg: c121/247 lr:0.000519 t:7.8s
+tttg: c122/247 lr:0.000513 t:7.9s
+tttg: c123/247 lr:0.000506 t:7.9s
+tttg: c124/247 lr:0.000500 t:8.0s
+tttg: c125/247 lr:0.000494 t:8.1s
+tttg: c126/247 lr:0.000487 t:8.1s
+tttg: c127/247 lr:0.000481 t:8.2s
+tttg: c128/247 lr:0.000474 t:8.3s
+tttg: c129/247 lr:0.000468 t:8.3s
+tttg: c130/247 lr:0.000462 t:8.4s
+tttg: c131/247 lr:0.000455 t:8.5s
+tttg: c132/247 lr:0.000449 t:8.5s
+tttg: c133/247 lr:0.000443 t:8.6s
+tttg: c134/247 lr:0.000436 t:8.6s
+tttg: c135/247 lr:0.000430 t:8.7s
+tttg: c136/247 lr:0.000424 t:8.8s
+tttg: c137/247 lr:0.000417 t:8.8s
+tttg: c138/247 lr:0.000411 t:8.9s
+tttg: c139/247 lr:0.000405 t:9.0s
+tttg: c140/247 lr:0.000399 t:9.0s
+tttg: c141/247 lr:0.000392 t:9.1s
+tttg: c142/247 lr:0.000386 t:9.2s
+tttg: c143/247 lr:0.000380 t:9.2s
+tttg: c144/247 lr:0.000374 t:9.3s
+tttg: c145/247 lr:0.000368 t:9.4s
+tttg: c146/247 lr:0.000361 t:9.4s
+tttg: c147/247 lr:0.000355 t:9.5s
+tttg: c148/247 lr:0.000349 t:9.5s
+tttg: c149/247 lr:0.000343 t:9.6s
+tttg: c150/247 lr:0.000337 t:9.7s
+tttg: c151/247 lr:0.000331 t:9.7s
+tttg: c152/247 lr:0.000325 t:9.8s
+tttg: c153/247 lr:0.000319 t:9.9s
+tttg: c154/247 lr:0.000313 t:9.9s
+tttg: c155/247 lr:0.000307 t:10.0s
+tttg: c156/247 lr:0.000301 t:10.1s
+tttg: c157/247 lr:0.000295 t:10.1s
+tttg: c158/247 lr:0.000290 t:10.2s
+tttg: c159/247 lr:0.000284 t:10.3s
+tttg: c160/247 lr:0.000278 t:10.3s
+tttg: c161/247 lr:0.000272 t:10.4s
+tttg: c162/247 lr:0.000267 t:10.4s
+tttg: c163/247 lr:0.000261 t:10.5s
+tttg: c164/247 lr:0.000256 t:10.6s
+tttg: c165/247 lr:0.000250 t:10.6s
+tttg: c166/247 lr:0.000244 t:10.7s
+tttg: c167/247 lr:0.000239 t:10.8s
+tttg: c168/247 lr:0.000234 t:10.8s
+tttg: c169/247 lr:0.000228 t:10.9s
+tttg: c170/247 lr:0.000223 t:11.0s
+tttg: c171/247 lr:0.000218 t:11.0s
+tttg: c172/247 lr:0.000212 t:11.1s
+tttg: c173/247 lr:0.000207 t:11.2s
+tttg: c174/247 lr:0.000202 t:11.2s
+tttg: c175/247 lr:0.000197 t:11.3s
+tttg: c176/247 lr:0.000192 t:11.3s
+tttg: c177/247 lr:0.000187 t:11.4s
+tttg: c178/247 lr:0.000182 t:11.5s
+tttg: c179/247 lr:0.000177 t:11.5s
+tttg: c180/247 lr:0.000172 t:11.6s
+tttg: c181/247 lr:0.000167 t:11.7s
+tttg: c182/247 lr:0.000163 t:11.7s
+tttg: c183/247 lr:0.000158 t:11.8s
+tttg: c184/247 lr:0.000153 t:11.9s
+tttg: c185/247 lr:0.000149 t:11.9s
+tttg: c186/247 lr:0.000144 t:12.0s
+tttg: c187/247 lr:0.000140 t:12.1s
+tttg: c188/247 lr:0.000135 t:12.1s
+tttg: c189/247 lr:0.000131 t:12.2s
+tttg: c190/247 lr:0.000127 t:12.2s
+tttg: c191/247 lr:0.000123 t:12.3s
+tttg: c192/247 lr:0.000118 t:12.4s
+tttg: c193/247 lr:0.000114 t:12.4s
+tttg: c194/247 lr:0.000110 t:12.5s
+tttg: c195/247 lr:0.000106 t:12.6s
+tttg: c196/247 lr:0.000102 t:12.6s
+tttg: c197/247 lr:0.000099 t:12.7s
+tttg: c198/247 lr:0.000095 t:12.8s
+tttg: c199/247 lr:0.000091 t:12.8s
+tttg: c200/247 lr:0.000087 t:12.9s
+tttg: c201/247 lr:0.000084 t:13.0s
+tttg: c202/247 lr:0.000080 t:13.0s
+tttg: c203/247 lr:0.000077 t:13.1s
+tttg: c204/247 lr:0.000074 t:13.1s
+tttg: c205/247 lr:0.000070 t:13.2s
+tttg: c206/247 lr:0.000067 t:13.3s
+tttg: c207/247 lr:0.000064 t:13.3s
+tttg: c208/247 lr:0.000061 t:13.4s
+tttg: c209/247 lr:0.000058 t:13.5s
+tttg: c210/247 lr:0.000055 t:13.5s
+tttg: c211/247 lr:0.000052 t:13.6s
+tttg: c212/247 lr:0.000049 t:13.7s
+tttg: c213/247 lr:0.000046 t:13.7s
+tttg: c214/247 lr:0.000044 t:13.8s
+tttg: c215/247 lr:0.000041 t:13.9s
+tttg: c216/247 lr:0.000039 t:13.9s
+tttg: c217/247 lr:0.000036 t:14.0s
+tttg: c218/247 lr:0.000034 t:14.1s
+tttg: c219/247 lr:0.000032 t:14.1s
+tttg: c220/247 lr:0.000029 t:14.2s
+tttg: c221/247 lr:0.000027 t:14.2s
+tttg: c222/247 lr:0.000025 t:14.3s
+tttg: c223/247 lr:0.000023 t:14.4s
+tttg: c224/247 lr:0.000021 t:14.4s
+tttg: c225/247 lr:0.000020 t:14.5s
+tttg: c226/247 lr:0.000018 t:14.6s
+tttg: c227/247 lr:0.000016 t:14.6s
+tttg: c228/247 lr:0.000015 t:14.7s
+tttg: c229/247 lr:0.000013 t:14.8s
+tttg: c230/247 lr:0.000012 t:14.8s
+tttg: c231/247 lr:0.000010 t:14.9s
+tttg: c232/247 lr:0.000009 t:15.0s
+tttg: c233/247 lr:0.000008 t:15.0s
+tttg: c234/247 lr:0.000007 t:15.1s
+tttg: c235/247 lr:0.000006 t:15.1s
+tttg: c236/247 lr:0.000005 t:15.2s
+tttg: c237/247 lr:0.000004 t:15.3s
+tttg: c238/247 lr:0.000003 t:15.3s
+tttg: c239/247 lr:0.000003 t:15.4s
+tttg: c240/247 lr:0.000002 t:15.5s
+tttg: c241/247 lr:0.000001 t:15.5s
+tttg: c242/247 lr:0.000001 t:15.6s
+tttg: c243/247 lr:0.000001 t:15.7s
+tttg: c244/247 lr:0.000000 t:15.7s
+tttg: c245/247 lr:0.000000 t:15.8s
+tttg: c246/247 lr:0.000000 t:15.9s
+ttpr: phase:2/3 t:351.1s
+ttp: b3031/3125 bl:2.3761 bb:1.1262 rl:2.3082 rb:1.0874 dl:3682-3715 gd:0
+ttp: b3021/3125 bl:2.3235 bb:1.0397 rl:2.3088 rb:1.0856 dl:3466-3487 gd:0
+ttp: b3016/3125 bl:2.3276 bb:1.1052 rl:2.3095 rb:1.0862 dl:3387-3397 gd:0
+ttp: b3005/3125 bl:2.0798 bb:1.0030 rl:2.3021 rb:1.0836 dl:3222-3232 gd:0
+ttp: b3001/3125 bl:2.2894 bb:1.0168 rl:2.3017 rb:1.0815 dl:3150-3167 gd:0
+ttpp: phase:3/3 pd:2112 gd:2000 t:369.9s
+tttg: c1/319 lr:0.001000 t:0.1s
+tttg: c2/319 lr:0.001000 t:0.1s
+tttg: c3/319 lr:0.001000 t:0.2s
+tttg: c4/319 lr:0.001000 t:0.3s
+tttg: c5/319 lr:0.001000 t:0.3s
+tttg: c6/319 lr:0.000999 t:0.4s
+tttg: c7/319 lr:0.000999 t:0.5s
+tttg: c8/319 lr:0.000999 t:0.5s
+tttg: c9/319 lr:0.000998 t:0.6s
+tttg: c10/319 lr:0.000998 t:0.7s
+tttg: c11/319 lr:0.000998 t:0.7s
+tttg: c12/319 lr:0.000997 t:0.8s
+tttg: c13/319 lr:0.000996 t:0.9s
+tttg: c14/319 lr:0.000996 t:0.9s
+tttg: c15/319 lr:0.000995 t:1.0s
+tttg: c16/319 lr:0.000995 t:1.0s
+tttg: c17/319 lr:0.000994 t:1.1s
+tttg: c18/319 lr:0.000993 t:1.2s
+tttg: c19/319 lr:0.000992 t:1.2s
+tttg: c20/319 lr:0.000991 t:1.3s
+tttg: c21/319 lr:0.000990 t:1.4s
+tttg: c22/319 lr:0.000989 t:1.4s
+tttg: c23/319 lr:0.000988 t:1.5s
+tttg: c24/319 lr:0.000987 t:1.6s
+tttg: c25/319 lr:0.000986 t:1.6s
+tttg: c26/319 lr:0.000985 t:1.7s
+tttg: c27/319 lr:0.000984 t:1.8s
+tttg: c28/319 lr:0.000982 t:1.8s
+tttg: c29/319 lr:0.000981 t:1.9s
+tttg: c30/319 lr:0.000980 t:2.0s
+tttg: c31/319 lr:0.000978 t:2.0s
+tttg: c32/319 lr:0.000977 t:2.1s
+tttg: c33/319 lr:0.000975 t:2.1s
+tttg: c34/319 lr:0.000974 t:2.2s
+tttg: c35/319 lr:0.000972 t:2.3s
+tttg: c36/319 lr:0.000970 t:2.3s
+tttg: c37/319 lr:0.000969 t:2.4s
+tttg: c38/319 lr:0.000967 t:2.5s
+tttg: c39/319 lr:0.000965 t:2.5s
+tttg: c40/319 lr:0.000963 t:2.6s
+tttg: c41/319 lr:0.000961 t:2.7s
+tttg: c42/319 lr:0.000960 t:2.7s
+tttg: c43/319 lr:0.000958 t:2.8s
+tttg: c44/319 lr:0.000956 t:2.9s
+tttg: c45/319 lr:0.000954 t:2.9s
+tttg: c46/319 lr:0.000951 t:3.0s
+tttg: c47/319 lr:0.000949 t:3.0s
+tttg: c48/319 lr:0.000947 t:3.1s
+tttg: c49/319 lr:0.000945 t:3.2s
+tttg: c50/319 lr:0.000943 t:3.2s
+tttg: c51/319 lr:0.000940 t:3.3s
+tttg: c52/319 lr:0.000938 t:3.4s
+tttg: c53/319 lr:0.000935 t:3.4s
+tttg: c54/319 lr:0.000933 t:3.5s
+tttg: c55/319 lr:0.000931 t:3.6s
+tttg: c56/319 lr:0.000928 t:3.6s
+tttg: c57/319 lr:0.000925 t:3.7s
+tttg: c58/319 lr:0.000923 t:3.8s
+tttg: c59/319 lr:0.000920 t:3.8s
+tttg: c60/319 lr:0.000917 t:3.9s
+tttg: c61/319 lr:0.000915 t:4.0s
+tttg: c62/319 lr:0.000912 t:4.0s
+tttg: c63/319 lr:0.000909 t:4.1s
+tttg: c64/319 lr:0.000906 t:4.2s
+tttg: c65/319 lr:0.000903 t:4.2s
+tttg: c66/319 lr:0.000900 t:4.3s
+tttg: c67/319 lr:0.000897 t:4.4s
+tttg: c68/319 lr:0.000894 t:4.4s
+tttg: c69/319 lr:0.000891 t:4.5s
+tttg: c70/319 lr:0.000888 t:4.5s
+tttg: c71/319 lr:0.000885 t:4.6s
+tttg: c72/319 lr:0.000882 t:4.7s
+tttg: c73/319 lr:0.000879 t:4.7s
+tttg: c74/319 lr:0.000876 t:4.8s
+tttg: c75/319 lr:0.000872 t:4.9s
+tttg: c76/319 lr:0.000869 t:4.9s
+tttg: c77/319 lr:0.000866 t:5.0s
+tttg: c78/319 lr:0.000862 t:5.1s
+tttg: c79/319 lr:0.000859 t:5.1s
+tttg: c80/319 lr:0.000855 t:5.2s
+tttg: c81/319 lr:0.000852 t:5.3s
+tttg: c82/319 lr:0.000848 t:5.3s
+tttg: c83/319 lr:0.000845 t:5.4s
+tttg: c84/319 lr:0.000841 t:5.5s
+tttg: c85/319 lr:0.000837 t:5.5s
+tttg: c86/319 lr:0.000834 t:5.6s
+tttg: c87/319 lr:0.000830 t:5.6s
+tttg: c88/319 lr:0.000826 t:5.7s
+tttg: c89/319 lr:0.000823 t:5.8s
+tttg: c90/319 lr:0.000819 t:5.8s
+tttg: c91/319 lr:0.000815 t:5.9s
+tttg: c92/319 lr:0.000811 t:6.0s
+tttg: c93/319 lr:0.000807 t:6.0s
+tttg: c94/319 lr:0.000803 t:6.1s
+tttg: c95/319 lr:0.000799 t:6.2s
+tttg: c96/319 lr:0.000795 t:6.2s
+tttg: c97/319 lr:0.000791 t:6.3s
+tttg: c98/319 lr:0.000787 t:6.4s
+tttg: c99/319 lr:0.000783 t:6.4s
+tttg: c100/319 lr:0.000779 t:6.5s
+tttg: c101/319 lr:0.000775 t:6.6s
+tttg: c102/319 lr:0.000771 t:6.6s
+tttg: c103/319 lr:0.000767 t:6.7s
+tttg: c104/319 lr:0.000763 t:6.7s
+tttg: c105/319 lr:0.000759 t:6.8s
+tttg: c106/319 lr:0.000754 t:6.9s
+tttg: c107/319 lr:0.000750 t:6.9s
+tttg: c108/319 lr:0.000746 t:7.0s
+tttg: c109/319 lr:0.000741 t:7.1s
+tttg: c110/319 lr:0.000737 t:7.1s
+tttg: c111/319 lr:0.000733 t:7.2s
+tttg: c112/319 lr:0.000728 t:7.3s
+tttg: c113/319 lr:0.000724 t:7.3s
+tttg: c114/319 lr:0.000719 t:7.4s
+tttg: c115/319 lr:0.000715 t:7.5s
+tttg: c116/319 lr:0.000711 t:7.5s
+tttg: c117/319 lr:0.000706 t:7.6s
+tttg: c118/319 lr:0.000702 t:7.7s
+tttg: c119/319 lr:0.000697 t:7.7s
+tttg: c120/319 lr:0.000692 t:7.8s
+tttg: c121/319 lr:0.000688 t:7.8s
+tttg: c122/319 lr:0.000683 t:7.9s
+tttg: c123/319 lr:0.000679 t:8.0s
+tttg: c124/319 lr:0.000674 t:8.0s
+tttg: c125/319 lr:0.000669 t:8.1s
+tttg: c126/319 lr:0.000665 t:8.2s
+tttg: c127/319 lr:0.000660 t:8.2s
+tttg: c128/319 lr:0.000655 t:8.3s
+tttg: c129/319 lr:0.000651 t:8.4s
+tttg: c130/319 lr:0.000646 t:8.4s
+tttg: c131/319 lr:0.000641 t:8.5s
+tttg: c132/319 lr:0.000637 t:8.6s
+tttg: c133/319 lr:0.000632 t:8.6s
+tttg: c134/319 lr:0.000627 t:8.7s
+tttg: c135/319 lr:0.000622 t:8.8s
+tttg: c136/319 lr:0.000617 t:8.8s
+tttg: c137/319 lr:0.000613 t:8.9s
+tttg: c138/319 lr:0.000608 t:8.9s
+tttg: c139/319 lr:0.000603 t:9.0s
+tttg: c140/319 lr:0.000598 t:9.1s
+tttg: c141/319 lr:0.000593 t:9.1s
+tttg: c142/319 lr:0.000588 t:9.2s
+tttg: c143/319 lr:0.000584 t:9.3s
+tttg: c144/319 lr:0.000579 t:9.3s
+tttg: c145/319 lr:0.000574 t:9.4s
+tttg: c146/319 lr:0.000569 t:9.5s
+tttg: c147/319 lr:0.000564 t:9.5s
+tttg: c148/319 lr:0.000559 t:9.6s
+tttg: c149/319 lr:0.000554 t:9.7s
+tttg: c150/319 lr:0.000549 t:9.7s
+tttg: c151/319 lr:0.000544 t:9.8s
+tttg: c152/319 lr:0.000539 t:9.9s
+tttg: c153/319 lr:0.000535 t:9.9s
+tttg: c154/319 lr:0.000530 t:10.0s
+tttg: c155/319 lr:0.000525 t:10.0s
+tttg: c156/319 lr:0.000520 t:10.1s
+tttg: c157/319 lr:0.000515 t:10.2s
+tttg: c158/319 lr:0.000510 t:10.2s
+tttg: c159/319 lr:0.000505 t:10.3s
+tttg: c160/319 lr:0.000500 t:10.4s
+tttg: c161/319 lr:0.000495 t:10.4s
+tttg: c162/319 lr:0.000490 t:10.5s
+tttg: c163/319 lr:0.000485 t:10.6s
+tttg: c164/319 lr:0.000480 t:10.6s
+tttg: c165/319 lr:0.000475 t:10.7s
+tttg: c166/319 lr:0.000470 t:10.8s
+tttg: c167/319 lr:0.000465 t:10.8s
+tttg: c168/319 lr:0.000461 t:10.9s
+tttg: c169/319 lr:0.000456 t:11.0s
+tttg: c170/319 lr:0.000451 t:11.0s
+tttg: c171/319 lr:0.000446 t:11.1s
+tttg: c172/319 lr:0.000441 t:11.2s
+tttg: c173/319 lr:0.000436 t:11.2s
+tttg: c174/319 lr:0.000431 t:11.3s
+tttg: c175/319 lr:0.000426 t:11.3s
+tttg: c176/319 lr:0.000421 t:11.4s
+tttg: c177/319 lr:0.000416 t:11.5s
+tttg: c178/319 lr:0.000412 t:11.5s
+tttg: c179/319 lr:0.000407 t:11.6s
+tttg: c180/319 lr:0.000402 t:11.7s
+tttg: c181/319 lr:0.000397 t:11.7s
+tttg: c182/319 lr:0.000392 t:11.8s
+tttg: c183/319 lr:0.000387 t:11.9s
+tttg: c184/319 lr:0.000383 t:11.9s
+tttg: c185/319 lr:0.000378 t:12.0s
+tttg: c186/319 lr:0.000373 t:12.1s
+tttg: c187/319 lr:0.000368 t:12.1s
+tttg: c188/319 lr:0.000363 t:12.2s
+tttg: c189/319 lr:0.000359 t:12.3s
+tttg: c190/319 lr:0.000354 t:12.3s
+tttg: c191/319 lr:0.000349 t:12.4s
+tttg: c192/319 lr:0.000345 t:12.5s
+tttg: c193/319 lr:0.000340 t:12.5s
+tttg: c194/319 lr:0.000335 t:12.6s
+tttg: c195/319 lr:0.000331 t:12.6s
+tttg: c196/319 lr:0.000326 t:12.7s
+tttg: c197/319 lr:0.000321 t:12.8s
+tttg: c198/319 lr:0.000317 t:12.8s
+tttg: c199/319 lr:0.000312 t:12.9s
+tttg: c200/319 lr:0.000308 t:13.0s
+tttg: c201/319 lr:0.000303 t:13.0s
+tttg: c202/319 lr:0.000298 t:13.1s
+tttg: c203/319 lr:0.000294 t:13.2s
+tttg: c204/319 lr:0.000289 t:13.2s
+tttg: c205/319 lr:0.000285 t:13.3s
+tttg: c206/319 lr:0.000281 t:13.4s
+tttg: c207/319 lr:0.000276 t:13.4s
+tttg: c208/319 lr:0.000272 t:13.5s
+tttg: c209/319 lr:0.000267 t:13.6s
+tttg: c210/319 lr:0.000263 t:13.6s
+tttg: c211/319 lr:0.000259 t:13.7s
+tttg: c212/319 lr:0.000254 t:13.7s
+tttg: c213/319 lr:0.000250 t:13.8s
+tttg: c214/319 lr:0.000246 t:13.9s
+tttg: c215/319 lr:0.000241 t:13.9s
+tttg: c216/319 lr:0.000237 t:14.0s
+tttg: c217/319 lr:0.000233 t:14.1s
+tttg: c218/319 lr:0.000229 t:14.1s
+tttg: c219/319 lr:0.000225 t:14.2s
+tttg: c220/319 lr:0.000221 t:14.3s
+tttg: c221/319 lr:0.000217 t:14.3s
+tttg: c222/319 lr:0.000213 t:14.4s
+tttg: c223/319 lr:0.000209 t:14.5s
+tttg: c224/319 lr:0.000205 t:14.5s
+tttg: c225/319 lr:0.000201 t:14.6s
+tttg: c226/319 lr:0.000197 t:14.7s
+tttg: c227/319 lr:0.000193 t:14.7s
+tttg: c228/319 lr:0.000189 t:14.8s
+tttg: c229/319 lr:0.000185 t:14.8s
+tttg: c230/319 lr:0.000181 t:14.9s
+tttg: c231/319 lr:0.000177 t:15.0s
+tttg: c232/319 lr:0.000174 t:15.0s
+tttg: c233/319 lr:0.000170 t:15.1s
+tttg: c234/319 lr:0.000166 t:15.2s
+tttg: c235/319 lr:0.000163 t:15.2s
+tttg: c236/319 lr:0.000159 t:15.3s
+tttg: c237/319 lr:0.000155 t:15.4s
+tttg: c238/319 lr:0.000152 t:15.4s
+tttg: c239/319 lr:0.000148 t:15.5s
+tttg: c240/319 lr:0.000145 t:15.6s
+tttg: c241/319 lr:0.000141 t:15.6s
+tttg: c242/319 lr:0.000138 t:15.7s
+tttg: c243/319 lr:0.000134 t:15.7s
+tttg: c244/319 lr:0.000131 t:15.8s
+tttg: c245/319 lr:0.000128 t:15.9s
+tttg: c246/319 lr:0.000124 t:15.9s
+tttg: c247/319 lr:0.000121 t:16.0s
+tttg: c248/319 lr:0.000118 t:16.1s
+tttg: c249/319 lr:0.000115 t:16.1s
+tttg: c250/319 lr:0.000112 t:16.2s
+tttg: c251/319 lr:0.000109 t:16.3s
+tttg: c252/319 lr:0.000106 t:16.3s
+tttg: c253/319 lr:0.000103 t:16.4s
+tttg: c254/319 lr:0.000100 t:16.5s
+tttg: c255/319 lr:0.000097 t:16.5s
+tttg: c256/319 lr:0.000094 t:16.6s
+tttg: c257/319 lr:0.000091 t:16.7s
+tttg: c258/319 lr:0.000088 t:16.7s
+tttg: c259/319 lr:0.000085 t:16.8s
+tttg: c260/319 lr:0.000083 t:16.9s
+tttg: c261/319 lr:0.000080 t:16.9s
+tttg: c262/319 lr:0.000077 t:17.0s
+tttg: c263/319 lr:0.000075 t:17.0s
+tttg: c264/319 lr:0.000072 t:17.1s
+tttg: c265/319 lr:0.000069 t:17.2s
+tttg: c266/319 lr:0.000067 t:17.2s
+tttg: c267/319 lr:0.000065 t:17.3s
+tttg: c268/319 lr:0.000062 t:17.4s
+tttg: c269/319 lr:0.000060 t:17.4s
+tttg: c270/319 lr:0.000057 t:17.5s
+tttg: c271/319 lr:0.000055 t:17.6s
+tttg: c272/319 lr:0.000053 t:17.6s
+tttg: c273/319 lr:0.000051 t:17.7s
+tttg: c274/319 lr:0.000049 t:17.8s
+tttg: c275/319 lr:0.000046 t:17.8s
+tttg: c276/319 lr:0.000044 t:17.9s
+tttg: c277/319 lr:0.000042 t:18.0s
+tttg: c278/319 lr:0.000040 t:18.0s
+tttg: c279/319 lr:0.000039 t:18.1s
+tttg: c280/319 lr:0.000037 t:18.2s
+tttg: c281/319 lr:0.000035 t:18.2s
+tttg: c282/319 lr:0.000033 t:18.3s
+tttg: c283/319 lr:0.000031 t:18.4s
+tttg: c284/319 lr:0.000030 t:18.4s
+tttg: c285/319 lr:0.000028 t:18.5s
+tttg: c286/319 lr:0.000026 t:18.5s
+tttg: c287/319 lr:0.000025 t:18.6s
+tttg: c288/319 lr:0.000023 t:18.7s
+tttg: c289/319 lr:0.000022 t:18.7s
+tttg: c290/319 lr:0.000020 t:18.8s
+tttg: c291/319 lr:0.000019 t:18.9s
+tttg: c292/319 lr:0.000018 t:18.9s
+tttg: c293/319 lr:0.000016 t:19.0s
+tttg: c294/319 lr:0.000015 t:19.1s
+tttg: c295/319 lr:0.000014 t:19.1s
+tttg: c296/319 lr:0.000013 t:19.2s
+tttg: c297/319 lr:0.000012 t:19.3s
+tttg: c298/319 lr:0.000011 t:19.3s
+tttg: c299/319 lr:0.000010 t:19.4s
+tttg: c300/319 lr:0.000009 t:19.5s
+tttg: c301/319 lr:0.000008 t:19.5s
+tttg: c302/319 lr:0.000007 t:19.6s
+tttg: c303/319 lr:0.000006 t:19.6s
+tttg: c304/319 lr:0.000005 t:19.7s
+tttg: c305/319 lr:0.000005 t:19.8s
+tttg: c306/319 lr:0.000004 t:19.8s
+tttg: c307/319 lr:0.000004 t:19.9s
+tttg: c308/319 lr:0.000003 t:20.0s
+tttg: c309/319 lr:0.000002 t:20.0s
+tttg: c310/319 lr:0.000002 t:20.1s
+tttg: c311/319 lr:0.000002 t:20.2s
+tttg: c312/319 lr:0.000001 t:20.2s
+tttg: c313/319 lr:0.000001 t:20.3s
+tttg: c314/319 lr:0.000001 t:20.4s
+tttg: c315/319 lr:0.000000 t:20.4s
+tttg: c316/319 lr:0.000000 t:20.5s
+tttg: c317/319 lr:0.000000 t:20.5s
+tttg: c318/319 lr:0.000000 t:20.6s
+ttpr: phase:3/3 t:391.0s
+ttp: b2993/3125 bl:2.4104 bb:1.0373 rl:2.3048 rb:1.0801 dl:3039-3050 gd:1
+ttp: b2979/3125 bl:2.1271 bb:0.9821 rl:2.3001 rb:1.0775 dl:2861-2872 gd:1
+ttp: b2975/3125 bl:2.4479 bb:1.1364 rl:2.3039 rb:1.0790 dl:2819-2830 gd:1
+ttp: b2968/3125 bl:2.3282 bb:1.0619 rl:2.3044 rb:1.0786 dl:2753-2762 gd:1
+ttp: b2956/3125 bl:2.3096 bb:1.0529 rl:2.3046 rb:1.0780 dl:2644-2652 gd:1
+ttp: b2947/3125 bl:2.2187 bb:0.9902 rl:2.3027 rb:1.0760 dl:2565-2572 gd:1
+ttp: b2942/3125 bl:2.2875 bb:1.0683 rl:2.3024 rb:1.0758 dl:2534-2538 gd:1
+ttp: b2934/3125 bl:2.2519 bb:1.0560 rl:2.3014 rb:1.0754 dl:2475-2480 gd:1
+ttp: b2927/3125 bl:2.3775 bb:1.1454 rl:2.3029 rb:1.0767 dl:2430-2435 gd:1
+ttp: b2919/3125 bl:2.3453 bb:1.0313 rl:2.3036 rb:1.0759 dl:2363-2367 gd:1
+ttp: b2908/3125 bl:2.3733 bb:1.0859 rl:2.3048 rb:1.0760 dl:2301-2305 gd:1
+ttp: b2903/3125 bl:2.4509 bb:1.0857 rl:2.3073 rb:1.0762 dl:2267-2272 gd:1
+ttp: b2892/3125 bl:2.5682 bb:1.1172 rl:2.3115 rb:1.0769 dl:2199-2203 gd:1
+ttp: b2888/3125 bl:2.3718 bb:1.1171 rl:2.3125 rb:1.0775 dl:2181-2185 gd:1
+ttp: b2881/3125 bl:2.3135 bb:0.9806 rl:2.3125 rb:1.0759 dl:2144-2147 gd:1
+ttp: b2872/3125 bl:2.2122 bb:1.0104 rl:2.3110 rb:1.0749 dl:2101-2106 gd:1
+ttp: b2866/3125 bl:2.2573 bb:0.9773 rl:2.3102 rb:1.0734 dl:2072-2078 gd:1
+ttp: b2857/3125 bl:2.4723 bb:1.0405 rl:2.3125 rb:1.0729 dl:2036-2039 gd:1
+ttp: b2838/3125 bl:2.3727 bb:1.0490 rl:2.3133 rb:1.0726 dl:1956-1958 gd:1
+ttp: b2830/3125 bl:2.2246 bb:0.9987 rl:2.3121 rb:1.0716 dl:1927-1929 gd:1
+ttp: b2822/3125 bl:2.2856 bb:1.0279 rl:2.3118 rb:1.0711 dl:1901-1903 gd:1
+ttp: b2815/3125 bl:2.3021 bb:1.0667 rl:2.3117 rb:1.0710 dl:1881-1883 gd:1
+ttp: b2807/3125 bl:2.6291 bb:1.1359 rl:2.3154 rb:1.0718 dl:1853-1855 gd:1
+ttp: b2799/3125 bl:2.2416 bb:0.9866 rl:2.3146 rb:1.0708 dl:1830-1832 gd:1
+ttp: b2792/3125 bl:2.2562 bb:1.0088 rl:2.3139 rb:1.0701 dl:1812-1814 gd:1
+ttp: b2786/3125 bl:2.3699 bb:1.0415 rl:2.3146 rb:1.0698 dl:1792-1795 gd:1
+ttp: b2779/3125 bl:2.4459 bb:1.1173 rl:2.3160 rb:1.0703 dl:1774-1777 gd:1
+ttp: b2771/3125 bl:2.4224 bb:1.0131 rl:2.3171 rb:1.0696 dl:1751-1755 gd:1
+ttp: b2763/3125 bl:2.4630 bb:1.0988 rl:2.3186 rb:1.0699 dl:1731-1733 gd:1
+ttp: b2755/3125 bl:2.5274 bb:1.0696 rl:2.3207 rb:1.0699 dl:1710-1712 gd:1
+ttp: b2747/3125 bl:2.2627 bb:1.0205 rl:2.3201 rb:1.0694 dl:1690-1693 gd:1
+ttp: b2739/3125 bl:2.3682 bb:1.0467 rl:2.3206 rb:1.0692 dl:1670-1673 gd:1
+ttp: b2731/3125 bl:2.1984 bb:1.0397 rl:2.3195 rb:1.0689 dl:1651-1654 gd:1
+ttp: b2723/3125 bl:2.2892 bb:1.0485 rl:2.3192 rb:1.0687 dl:1632-1635 gd:1
+ttp: b2715/3125 bl:2.2350 bb:1.0384 rl:2.3184 rb:1.0685 dl:1614-1616 gd:1
+ttp: b2707/3125 bl:2.2028 bb:0.9956 rl:2.3174 rb:1.0678 dl:1598-1600 gd:1
+ttp: b2699/3125 bl:2.3491 bb:1.0553 rl:2.3177 rb:1.0677 dl:1583-1584 gd:1
+ttp: b2691/3125 bl:2.2546 bb:1.0631 rl:2.3171 rb:1.0676 dl:1566-1569 gd:1
+ttp: b2683/3125 bl:2.3761 bb:1.0401 rl:2.3176 rb:1.0674 dl:1549-1551 gd:1
+ttp: b2675/3125 bl:2.2812 bb:1.0214 rl:2.3173 rb:1.0670 dl:1533-1535 gd:1
+ttp: b2667/3125 bl:2.3483 bb:1.0478 rl:2.3176 rb:1.0668 dl:1518-1519 gd:1
+ttp: b2659/3125 bl:2.3463 bb:1.0234 rl:2.3178 rb:1.0665 dl:1503-1505 gd:1
+ttp: b2651/3125 bl:2.1927 bb:0.9941 rl:2.3168 rb:1.0659 dl:1489-1491 gd:1
+ttp: b2643/3125 bl:2.3864 bb:1.0747 rl:2.3173 rb:1.0660 dl:1477-1479 gd:1
+ttp: b2635/3125 bl:2.3693 bb:1.0045 rl:2.3177 rb:1.0655 dl:1462-1464 gd:1
+ttp: b2628/3125 bl:2.3126 bb:1.0788 rl:2.3177 rb:1.0656 dl:1451-1452 gd:1
+ttp: b2621/3125 bl:2.3568 bb:1.1274 rl:2.3180 rb:1.0660 dl:1439-1440 gd:1
+ttp: b2613/3125 bl:2.3254 bb:1.0575 rl:2.3180 rb:1.0659 dl:1425-1427 gd:1
+ttp: b2605/3125 bl:2.2527 bb:1.0584 rl:2.3176 rb:1.0659 dl:1411-1413 gd:1
+ttp: b2597/3125 bl:2.2807 bb:1.0213 rl:2.3173 rb:1.0656 dl:1398-1399 gd:1
+ttp: b2589/3125 bl:2.3439 bb:0.9903 rl:2.3175 rb:1.0650 dl:1387-1388 gd:1
+ttp: b2581/3125 bl:2.2590 bb:1.0197 rl:2.3171 rb:1.0647 dl:1375-1376 gd:1
+ttp: b2573/3125 bl:2.3147 bb:1.0396 rl:2.3171 rb:1.0645 dl:1362-1363 gd:1
+ttp: b2565/3125 bl:2.3503 bb:1.0858 rl:2.3173 rb:1.0647 dl:1349-1351 gd:1
+ttp: b2557/3125 bl:2.2172 bb:1.0777 rl:2.3167 rb:1.0647 dl:1337-1338 gd:1
+ttp: b2549/3125 bl:2.2864 bb:1.0542 rl:2.3165 rb:1.0647 dl:1325-1327 gd:1
+ttp: b2541/3125 bl:2.4271 bb:1.0481 rl:2.3172 rb:1.0646 dl:1314-1316 gd:1
+ttp: b2534/3125 bl:2.3803 bb:0.9925 rl:2.3176 rb:1.0641 dl:1304-1305 gd:1
+ttp: b2526/3125 bl:2.2977 bb:1.0181 rl:2.3174 rb:1.0638 dl:1292-1293 gd:1
+ttp: b2518/3125 bl:2.4016 bb:1.0127 rl:2.3179 rb:1.0634 dl:1281-1283 gd:1
+ttp: b2510/3125 bl:2.2795 bb:1.0089 rl:2.3177 rb:1.0631 dl:1272-1273 gd:1
+ttp: b2502/3125 bl:2.3481 bb:1.0595 rl:2.3179 rb:1.0631 dl:1261-1262 gd:1
+ttp: b2494/3125 bl:2.2171 bb:1.0688 rl:2.3173 rb:1.0631 dl:1250-1252 gd:1
+ttp: b2484/3125 bl:2.4155 bb:1.0432 rl:2.3179 rb:1.0630 dl:1236-1237 gd:1
+ttp: b2476/3125 bl:2.1755 bb:0.9966 rl:2.3171 rb:1.0626 dl:1225-1226 gd:1
+ttp: b2468/3125 bl:2.4389 bb:1.0265 rl:2.3177 rb:1.0624 dl:1215-1216 gd:1
+ttp: b2460/3125 bl:2.5189 bb:1.0912 rl:2.3188 rb:1.0626 dl:1204-1205 gd:1
+ttp: b2452/3125 bl:2.3025 bb:1.0228 rl:2.3187 rb:1.0624 dl:1194-1195 gd:1
+ttp: b2444/3125 bl:2.2775 bb:1.0025 rl:2.3185 rb:1.0620 dl:1185-1186 gd:1
+ttp: b2436/3125 bl:2.3055 bb:0.9823 rl:2.3185 rb:1.0616 dl:1176-1177 gd:1
+ttp: b2428/3125 bl:2.2522 bb:0.9981 rl:2.3181 rb:1.0612 dl:1167-1168 gd:1
+ttp: b2420/3125 bl:2.2672 bb:0.9935 rl:2.3179 rb:1.0609 dl:1158-1159 gd:1
+ttp: b2412/3125 bl:2.5948 bb:1.0597 rl:2.3192 rb:1.0609 dl:1149-1150 gd:1
+ttp: b2404/3125 bl:2.2097 bb:0.9885 rl:2.3187 rb:1.0605 dl:1140-1141 gd:1
+ttp: b2396/3125 bl:2.4441 bb:1.0875 rl:2.3193 rb:1.0606 dl:1132-1133 gd:1
+ttp: b2388/3125 bl:2.2622 bb:1.0054 rl:2.3190 rb:1.0604 dl:1124-1124 gd:1
+ttp: b2381/3125 bl:2.1805 bb:1.0813 rl:2.3184 rb:1.0605 dl:1115-1115 gd:1
+ttp: b2371/3125 bl:2.2654 bb:0.9932 rl:2.3181 rb:1.0601 dl:1105-1106 gd:1
+ttp: b2363/3125 bl:2.4996 bb:1.1389 rl:2.3190 rb:1.0605 dl:1096-1097 gd:1
+ttp: b2355/3125 bl:2.4610 bb:1.0918 rl:2.3196 rb:1.0606 dl:1088-1088 gd:1
+ttp: b2346/3125 bl:2.3536 bb:0.9998 rl:2.3198 rb:1.0603 dl:1077-1079 gd:1
+ttp: b2336/3125 bl:2.3644 bb:1.0423 rl:2.3200 rb:1.0603 dl:1067-1069 gd:1
+ttp: b2329/3125 bl:2.3581 bb:1.0283 rl:2.3201 rb:1.0601 dl:1060-1061 gd:1
+ttp: b2322/3125 bl:2.2953 bb:1.0167 rl:2.3200 rb:1.0599 dl:1053-1054 gd:1
+ttp: b2314/3125 bl:2.5350 bb:1.0786 rl:2.3210 rb:1.0600 dl:1045-1046 gd:1
+ttp: b2308/3125 bl:2.0679 bb:0.9679 rl:2.3199 rb:1.0596 dl:1040-1041 gd:1
+ttp: b2301/3125 bl:2.2953 bb:1.0480 rl:2.3198 rb:1.0596 dl:1033-1034 gd:1
+ttp: b2291/3125 bl:2.4000 bb:1.0679 rl:2.3201 rb:1.0596 dl:1023-1024 gd:1
+ttp: b2283/3125 bl:2.1999 bb:1.0216 rl:2.3196 rb:1.0595 dl:1015-1016 gd:1
+ttp: b2273/3125 bl:2.3072 bb:1.0267 rl:2.3196 rb:1.0593 dl:1007-1008 gd:1
+ttp: b2267/3125 bl:2.1293 bb:0.9581 rl:2.3188 rb:1.0589 dl:1003-1003 gd:1
+ttp: b2259/3125 bl:2.3709 bb:0.9876 rl:2.3190 rb:1.0586 dl:995-996 gd:1
+ttp: b2252/3125 bl:2.1793 bb:1.0138 rl:2.3185 rb:1.0584 dl:989-990 gd:1
+ttp: b2244/3125 bl:2.3188 bb:1.0583 rl:2.3185 rb:1.0584 dl:982-983 gd:1
+ttp: b2215/3125 bl:2.3507 bb:1.0802 rl:2.3186 rb:1.0585 dl:957-958 gd:1
+ttp: b2207/3125 bl:2.2527 bb:1.0162 rl:2.3183 rb:1.0583 dl:951-951 gd:1
+ttp: b2198/3125 bl:2.2216 bb:1.0369 rl:2.3180 rb:1.0583 dl:944-945 gd:1
+ttp: b2191/3125 bl:2.0702 bb:1.0042 rl:2.3171 rb:1.0581 dl:937-938 gd:1
+ttp: b2185/3125 bl:2.2855 bb:1.0174 rl:2.3169 rb:1.0579 dl:934-934 gd:1
+ttp: b2177/3125 bl:2.2162 bb:1.0072 rl:2.3166 rb:1.0577 dl:927-928 gd:1
+ttp: b2170/3125 bl:2.3431 bb:1.1243 rl:2.3167 rb:1.0580 dl:922-923 gd:1
+ttp: b2161/3125 bl:2.2350 bb:1.0508 rl:2.3164 rb:1.0579 dl:915-916 gd:1
+ttp: b2154/3125 bl:2.3626 bb:1.0197 rl:2.3165 rb:1.0578 dl:909-910 gd:1
+ttp: b2146/3125 bl:2.3057 bb:1.0864 rl:2.3165 rb:1.0579 dl:903-904 gd:1
+ttp: b2140/3125 bl:2.5930 bb:1.1609 rl:2.3175 rb:1.0583 dl:898-899 gd:1
+ttp: b2133/3125 bl:2.3745 bb:1.0426 rl:2.3177 rb:1.0582 dl:893-893 gd:1
+ttp: b2125/3125 bl:2.3694 bb:1.0772 rl:2.3178 rb:1.0583 dl:887-888 gd:1
+ttp: b2120/3125 bl:2.2735 bb:1.0760 rl:2.3177 rb:1.0583 dl:884-884 gd:1
+ttp: b2112/3125 bl:2.4356 bb:1.0566 rl:2.3181 rb:1.0583 dl:877-878 gd:1
+ttp: b2104/3125 bl:2.3075 bb:1.0867 rl:2.3180 rb:1.0584 dl:871-872 gd:1
+ttp: b2096/3125 bl:2.3021 bb:1.0298 rl:2.3180 rb:1.0583 dl:865-866 gd:1
+ttp: b2088/3125 bl:2.3182 bb:1.0746 rl:2.3180 rb:1.0584 dl:859-860 gd:1
+ttp: b2079/3125 bl:2.3190 bb:0.9343 rl:2.3180 rb:1.0579 dl:853-854 gd:1
+ttp: b2073/3125 bl:2.1910 bb:1.0328 rl:2.3176 rb:1.0578 dl:850-850 gd:1
+ttp: b2065/3125 bl:2.4303 bb:1.0682 rl:2.3179 rb:1.0579 dl:843-844 gd:1
+ttp: b2058/3125 bl:2.3633 bb:1.0342 rl:2.3181 rb:1.0578 dl:839-839 gd:1
+ttp: b2049/3125 bl:2.3006 bb:1.0391 rl:2.3180 rb:1.0577 dl:832-833 gd:1
+ttp: b2040/3125 bl:2.4078 bb:1.1077 rl:2.3183 rb:1.0579 dl:825-826 gd:1
+ttp: b2032/3125 bl:2.4484 bb:1.0602 rl:2.3187 rb:1.0579 dl:819-820 gd:1
+ttp: b2025/3125 bl:2.3684 bb:1.0344 rl:2.3188 rb:1.0578 dl:814-815 gd:1
+ttp: b2017/3125 bl:2.3323 bb:1.0391 rl:2.3189 rb:1.0578 dl:809-810 gd:1
+ttp: b2010/3125 bl:2.5207 bb:1.0954 rl:2.3195 rb:1.0579 dl:805-805 gd:1
+ttp: b2001/3125 bl:2.3562 bb:0.9914 rl:2.3196 rb:1.0577 dl:799-800 gd:1
+ttp: b1992/3125 bl:2.2890 bb:1.0360 rl:2.3195 rb:1.0576 dl:793-794 gd:1
+ttp: b1983/3125 bl:2.3055 bb:1.0459 rl:2.3194 rb:1.0576 dl:787-788 gd:1
+ttp: b1975/3125 bl:2.4487 bb:1.0736 rl:2.3198 rb:1.0576 dl:782-782 gd:1
+ttp: b1966/3125 bl:2.2429 bb:1.0221 rl:2.3196 rb:1.0575 dl:776-777 gd:1
+ttp: b1957/3125 bl:2.4190 bb:1.0817 rl:2.3199 rb:1.0576 dl:771-772 gd:1
+ttp: b1950/3125 bl:2.3258 bb:1.0470 rl:2.3199 rb:1.0576 dl:767-768 gd:1
+ttp: b1941/3125 bl:2.5040 bb:1.1045 rl:2.3204 rb:1.0577 dl:761-762 gd:1
+ttp: b1933/3125 bl:2.4539 bb:1.0424 rl:2.3207 rb:1.0576 dl:756-757 gd:1
+ttp: b1926/3125 bl:2.3136 bb:1.0665 rl:2.3207 rb:1.0577 dl:753-753 gd:1
+ttp: b1918/3125 bl:2.4641 bb:1.1217 rl:2.3211 rb:1.0578 dl:747-748 gd:1
+ttp: b1909/3125 bl:2.2464 bb:1.0684 rl:2.3209 rb:1.0579 dl:742-743 gd:1
+ttp: b1903/3125 bl:2.3353 bb:1.0988 rl:2.3209 rb:1.0580 dl:739-739 gd:1
+ttp: b1894/3125 bl:2.1971 bb:1.0108 rl:2.3206 rb:1.0579 dl:734-734 gd:1
+ttp: b1886/3125 bl:2.3846 bb:1.1064 rl:2.3208 rb:1.0580 dl:729-729 gd:1
+ttp: b1879/3125 bl:2.3815 bb:1.0549 rl:2.3209 rb:1.0580 dl:725-725 gd:1
+ttp: b1869/3125 bl:2.2905 bb:1.0332 rl:2.3209 rb:1.0579 dl:719-720 gd:1
+ttp: b1861/3125 bl:2.3051 bb:0.9902 rl:2.3208 rb:1.0577 dl:714-715 gd:1
+ttp: b1853/3125 bl:2.2466 bb:1.0700 rl:2.3206 rb:1.0578 dl:710-711 gd:1
+ttp: b1847/3125 bl:2.3455 bb:1.0503 rl:2.3207 rb:1.0577 dl:707-707 gd:1
+ttp: b1838/3125 bl:2.1767 bb:1.0419 rl:2.3204 rb:1.0577 dl:701-702 gd:1
+ttp: b1831/3125 bl:2.1088 bb:1.0229 rl:2.3198 rb:1.0576 dl:699-699 gd:1
+ttp: b1821/3125 bl:2.4165 bb:1.0486 rl:2.3201 rb:1.0576 dl:693-694 gd:1
+ttp: b1813/3125 bl:2.2268 bb:1.0627 rl:2.3199 rb:1.0576 dl:689-690 gd:1
+ttp: b1806/3125 bl:2.2508 bb:1.0239 rl:2.3197 rb:1.0575 dl:685-686 gd:1
+ttp: b1800/3125 bl:2.4417 bb:1.0296 rl:2.3200 rb:1.0575 dl:682-682 gd:1
+ttp: b1791/3125 bl:2.3867 bb:1.0271 rl:2.3201 rb:1.0574 dl:677-678 gd:1
+ttp: b1786/3125 bl:2.4325 bb:1.0925 rl:2.3204 rb:1.0575 dl:675-675 gd:1
+ttp: b1779/3125 bl:2.4877 bb:1.1475 rl:2.3208 rb:1.0577 dl:671-671 gd:1
+ttp: b1771/3125 bl:2.1648 bb:1.0106 rl:2.3204 rb:1.0576 dl:667-667 gd:1
+ttp: b1763/3125 bl:2.4460 bb:1.0911 rl:2.3207 rb:1.0576 dl:663-663 gd:1
+ttp: b1754/3125 bl:2.3014 bb:1.0042 rl:2.3207 rb:1.0575 dl:657-658 gd:1
+ttp: b1745/3125 bl:2.3478 bb:1.0024 rl:2.3207 rb:1.0574 dl:653-654 gd:1
+ttp: b1739/3125 bl:2.3068 bb:1.0695 rl:2.3207 rb:1.0574 dl:650-650 gd:1
+ttp: b1730/3125 bl:2.1541 bb:1.0674 rl:2.3203 rb:1.0574 dl:646-646 gd:1
+ttp: b1724/3125 bl:2.3396 bb:1.0102 rl:2.3204 rb:1.0573 dl:643-643 gd:1
+ttp: b1713/3125 bl:2.4381 bb:1.1532 rl:2.3206 rb:1.0575 dl:638-639 gd:1
+ttp: b1705/3125 bl:2.3204 bb:1.0828 rl:2.3206 rb:1.0576 dl:634-635 gd:1
+ttp: b1697/3125 bl:2.2660 bb:0.9980 rl:2.3205 rb:1.0574 dl:630-631 gd:1
+ttp: b1689/3125 bl:2.3697 bb:1.0532 rl:2.3206 rb:1.0574 dl:626-627 gd:1
+ttp: b1681/3125 bl:2.1614 bb:0.9934 rl:2.3203 rb:1.0573 dl:622-623 gd:1
+ttp: b1673/3125 bl:2.2929 bb:1.0637 rl:2.3202 rb:1.0573 dl:618-619 gd:1
+ttp: b1667/3125 bl:2.2612 bb:1.0172 rl:2.3201 rb:1.0572 dl:616-616 gd:1
+ttp: b1658/3125 bl:2.3273 bb:1.0894 rl:2.3201 rb:1.0573 dl:612-612 gd:1
+ttp: b1652/3125 bl:2.4563 bb:1.0859 rl:2.3204 rb:1.0574 dl:609-609 gd:1
+ttp: b1644/3125 bl:2.3630 bb:1.0592 rl:2.3205 rb:1.0574 dl:605-605 gd:1
+ttp: b1634/3125 bl:2.2856 bb:1.0154 rl:2.3204 rb:1.0573 dl:599-600 gd:1
+ttp: b1625/3125 bl:2.0984 bb:1.0021 rl:2.3200 rb:1.0572 dl:595-596 gd:1
+ttp: b1620/3125 bl:2.3876 bb:1.0713 rl:2.3201 rb:1.0572 dl:593-593 gd:1
+ttp: b1608/3125 bl:2.2636 bb:0.9424 rl:2.3200 rb:1.0570 dl:587-588 gd:1
+ttp: b1600/3125 bl:2.2902 bb:1.0408 rl:2.3199 rb:1.0569 dl:584-584 gd:1
+ttp: b1592/3125 bl:2.3277 bb:1.0138 rl:2.3200 rb:1.0569 dl:580-581 gd:1
+ttp: b1587/3125 bl:2.3067 bb:1.0476 rl:2.3199 rb:1.0568 dl:578-578 gd:1
+ttp: b1577/3125 bl:2.2335 bb:1.0307 rl:2.3198 rb:1.0568 dl:573-574 gd:1
+ttp: b1568/3125 bl:2.0075 bb:0.9841 rl:2.3192 rb:1.0567 dl:569-570 gd:1
+ttp: b1562/3125 bl:2.3189 bb:1.0208 rl:2.3192 rb:1.0566 dl:566-567 gd:1
+ttp: b1553/3125 bl:2.1039 bb:0.9695 rl:2.3188 rb:1.0564 dl:563-563 gd:1
+ttp: b1544/3125 bl:2.3930 bb:1.1010 rl:2.3190 rb:1.0565 dl:558-559 gd:1
+ttp: b1537/3125 bl:2.3495 bb:1.0556 rl:2.3190 rb:1.0565 dl:555-556 gd:1
+ttp: b1531/3125 bl:2.2843 bb:1.0802 rl:2.3189 rb:1.0566 dl:553-553 gd:1
+ttp: b1520/3125 bl:2.3130 bb:1.0500 rl:2.3189 rb:1.0566 dl:548-549 gd:1
+ttp: b1513/3125 bl:2.3644 bb:1.0481 rl:2.3190 rb:1.0565 dl:545-546 gd:1
+ttp: b1506/3125 bl:2.1761 bb:1.0447 rl:2.3188 rb:1.0565 dl:543-543 gd:1
+ttp: b1497/3125 bl:2.5798 bb:1.1528 rl:2.3192 rb:1.0567 dl:538-539 gd:1
+ttp: b1490/3125 bl:2.5028 bb:1.0892 rl:2.3195 rb:1.0567 dl:536-536 gd:1
+ttp: b1484/3125 bl:2.2227 bb:1.1088 rl:2.3194 rb:1.0568 dl:533-533 gd:1
+ttp: b1473/3125 bl:2.4021 bb:1.1381 rl:2.3195 rb:1.0569 dl:528-529 gd:1
+ttp: b1467/3125 bl:2.2015 bb:1.0561 rl:2.3193 rb:1.0569 dl:526-526 gd:1
+ttp: b1455/3125 bl:2.4520 bb:1.0869 rl:2.3195 rb:1.0570 dl:521-522 gd:1
+ttp: b1449/3125 bl:2.3814 bb:1.0597 rl:2.3196 rb:1.0570 dl:518-519 gd:1
+ttp: b1443/3125 bl:2.3398 bb:1.1529 rl:2.3196 rb:1.0571 dl:516-516 gd:1
+ttp: b1432/3125 bl:2.4195 bb:1.0585 rl:2.3198 rb:1.0571 dl:511-512 gd:1
+ttp: b1423/3125 bl:2.5237 bb:1.1403 rl:2.3201 rb:1.0573 dl:507-508 gd:1
+ttp: b1417/3125 bl:2.2998 bb:1.1026 rl:2.3201 rb:1.0573 dl:504-505 gd:1
+ttp: b1407/3125 bl:2.4601 bb:1.1267 rl:2.3203 rb:1.0575 dl:500-501 gd:1
+ttp: b1404/3125 bl:2.4204 bb:1.0943 rl:2.3205 rb:1.0575 dl:499-499 gd:1
+ttp: b1392/3125 bl:2.3811 bb:1.0362 rl:2.3206 rb:1.0575 dl:495-495 gd:1
+ttp: b1383/3125 bl:2.4178 bb:1.0367 rl:2.3207 rb:1.0574 dl:491-492 gd:1
+ttp: b1375/3125 bl:2.2892 bb:1.0502 rl:2.3207 rb:1.0574 dl:488-489 gd:1
+ttp: b1367/3125 bl:2.2358 bb:1.0080 rl:2.3205 rb:1.0574 dl:485-486 gd:1
+ttp: b1359/3125 bl:2.4269 bb:1.0692 rl:2.3207 rb:1.0574 dl:482-483 gd:1
+ttp: b1351/3125 bl:2.5089 bb:1.1546 rl:2.3210 rb:1.0575 dl:480-480 gd:1
+ttp: b1343/3125 bl:2.4741 bb:1.1418 rl:2.3212 rb:1.0576 dl:476-477 gd:1
+ttp: b1336/3125 bl:2.3698 bb:1.0758 rl:2.3213 rb:1.0577 dl:474-474 gd:1
+ttp: b1328/3125 bl:2.2915 bb:1.0667 rl:2.3212 rb:1.0577 dl:471-471 gd:1
+ttp: b1321/3125 bl:2.2183 bb:1.0491 rl:2.3211 rb:1.0577 dl:468-468 gd:1
+ttp: b1313/3125 bl:2.2856 bb:1.1270 rl:2.3210 rb:1.0578 dl:465-465 gd:1
+ttp: b1303/3125 bl:2.2117 bb:0.9764 rl:2.3209 rb:1.0576 dl:462-462 gd:1
+ttp: b1294/3125 bl:2.2041 bb:1.0424 rl:2.3207 rb:1.0576 dl:458-459 gd:1
+ttp: b1291/3125 bl:2.5059 bb:1.1453 rl:2.3210 rb:1.0577 dl:457-457 gd:1
+ttp: b1280/3125 bl:2.2684 bb:1.0710 rl:2.3209 rb:1.0578 dl:453-453 gd:1
+ttp: b1274/3125 bl:2.5185 bb:1.1709 rl:2.3212 rb:1.0579 dl:450-450 gd:1
+ttp: b1262/3125 bl:2.3681 bb:1.0928 rl:2.3212 rb:1.0580 dl:445-446 gd:1
+ttp: b1258/3125 bl:2.4033 bb:1.0229 rl:2.3213 rb:1.0579 dl:444-444 gd:1
+ttp: b1250/3125 bl:2.3318 bb:1.0786 rl:2.3213 rb:1.0579 dl:441-441 gd:1
+ttp: b1240/3125 bl:2.1224 bb:1.0719 rl:2.3211 rb:1.0579 dl:437-438 gd:1
+ttp: b1232/3125 bl:2.3347 bb:1.0956 rl:2.3211 rb:1.0580 dl:434-435 gd:1
+ttp: b1224/3125 bl:2.4349 bb:1.0634 rl:2.3212 rb:1.0580 dl:431-432 gd:1
+ttp: b1217/3125 bl:2.3407 bb:1.1392 rl:2.3213 rb:1.0581 dl:429-429 gd:1
+ttp: b1209/3125 bl:2.4324 bb:1.0752 rl:2.3214 rb:1.0581 dl:426-426 gd:1
+ttp: b1198/3125 bl:2.2279 bb:1.0199 rl:2.3213 rb:1.0581 dl:421-422 gd:1
+ttp: b1194/3125 bl:2.4252 bb:1.1974 rl:2.3214 rb:1.0582 dl:420-420 gd:1
+ttp: b1186/3125 bl:2.4433 bb:1.0870 rl:2.3216 rb:1.0583 dl:417-417 gd:1
+ttp: b1178/3125 bl:2.1840 bb:1.0573 rl:2.3214 rb:1.0583 dl:414-414 gd:1
+ttp: b1166/3125 bl:2.4184 bb:1.1192 rl:2.3215 rb:1.0583 dl:409-410 gd:1
+ttp: b1158/3125 bl:2.2721 bb:1.0593 rl:2.3215 rb:1.0583 dl:406-407 gd:1
+ttp: b1154/3125 bl:2.2492 bb:1.0109 rl:2.3214 rb:1.0583 dl:405-405 gd:1
+ttp: b1146/3125 bl:2.5026 bb:1.1759 rl:2.3216 rb:1.0584 dl:402-402 gd:1
+ttp: b1135/3125 bl:2.4201 bb:1.1126 rl:2.3217 rb:1.0585 dl:398-399 gd:1
+ttp: b1128/3125 bl:2.1811 bb:1.0899 rl:2.3215 rb:1.0585 dl:396-396 gd:1
+ttp: b1122/3125 bl:2.2924 bb:1.0743 rl:2.3215 rb:1.0585 dl:394-394 gd:1
+ttp: b1111/3125 bl:2.3261 bb:1.0165 rl:2.3215 rb:1.0585 dl:390-391 gd:1
+ttp: b1108/3125 bl:2.2548 bb:1.0825 rl:2.3214 rb:1.0585 dl:389-389 gd:1
+ttp: b1098/3125 bl:2.4838 bb:1.1020 rl:2.3216 rb:1.0586 dl:386-386 gd:1
+ttp: b1093/3125 bl:2.3541 bb:1.0697 rl:2.3217 rb:1.0586 dl:384-384 gd:1
+ttp: b1085/3125 bl:2.3862 bb:1.0595 rl:2.3217 rb:1.0586 dl:382-382 gd:1
+ttp: b1081/3125 bl:2.5418 bb:1.1463 rl:2.3220 rb:1.0587 dl:380-380 gd:1
+ttp: b1074/3125 bl:2.5215 bb:1.1158 rl:2.3222 rb:1.0587 dl:378-378 gd:1
+ttp: b1067/3125 bl:2.5386 bb:1.2029 rl:2.3224 rb:1.0589 dl:376-376 gd:1
+ttp: b1061/3125 bl:2.2736 bb:1.1180 rl:2.3224 rb:1.0589 dl:374-374 gd:1
+ttp: b1055/3125 bl:2.4026 bb:1.1183 rl:2.3225 rb:1.0590 dl:372-372 gd:1
+ttp: b1049/3125 bl:2.4526 bb:1.1100 rl:2.3226 rb:1.0591 dl:370-370 gd:1
+ttp: b1038/3125 bl:2.4157 bb:1.0806 rl:2.3227 rb:1.0591 dl:366-367 gd:1
+ttp: b1033/3125 bl:2.3333 bb:1.0821 rl:2.3227 rb:1.0591 dl:365-365 gd:1
+ttp: b1023/3125 bl:2.4524 bb:1.1028 rl:2.3229 rb:1.0592 dl:361-362 gd:1
+ttp: b1017/3125 bl:2.3651 bb:1.0780 rl:2.3229 rb:1.0592 dl:360-360 gd:1
+ttp: b1006/3125 bl:2.3007 bb:1.0354 rl:2.3229 rb:1.0592 dl:356-357 gd:1
+ttp: b999/3125 bl:2.1018 bb:1.0294 rl:2.3226 rb:1.0591 dl:354-355 gd:1
+ttp: b995/3125 bl:2.5869 bb:1.1179 rl:2.3229 rb:1.0592 dl:353-353 gd:1
+ttp: b984/3125 bl:2.4025 bb:1.0793 rl:2.3230 rb:1.0592 dl:350-350 gd:1
+ttp: b974/3125 bl:2.2635 bb:1.1194 rl:2.3229 rb:1.0593 dl:346-347 gd:1
+ttp: b969/3125 bl:2.4981 bb:1.1263 rl:2.3231 rb:1.0593 dl:345-345 gd:1
+ttp: b959/3125 bl:2.3096 bb:1.0738 rl:2.3231 rb:1.0593 dl:342-342 gd:1
+ttp: b954/3125 bl:2.5120 bb:1.1507 rl:2.3233 rb:1.0594 dl:340-340 gd:1
+ttp: b943/3125 bl:2.2057 bb:1.0290 rl:2.3232 rb:1.0594 dl:336-337 gd:1
+ttp: b942/3125 bl:2.3774 bb:1.0806 rl:2.3232 rb:1.0594 dl:336-336 gd:1
+ttp: b935/3125 bl:2.3304 bb:1.1092 rl:2.3232 rb:1.0595 dl:334-334 gd:1
+ttp: b927/3125 bl:2.3783 bb:1.0615 rl:2.3233 rb:1.0595 dl:332-332 gd:1
+ttp: b920/3125 bl:2.4715 bb:1.1683 rl:2.3234 rb:1.0596 dl:330-330 gd:1
+ttp: b908/3125 bl:2.5397 bb:1.2484 rl:2.3236 rb:1.0597 dl:326-327 gd:1
+ttp: b901/3125 bl:2.3381 bb:1.0187 rl:2.3236 rb:1.0597 dl:324-325 gd:1
+ttp: b894/3125 bl:2.5773 bb:1.1972 rl:2.3239 rb:1.0598 dl:322-323 gd:1
+ttp: b888/3125 bl:2.3814 bb:1.1035 rl:2.3239 rb:1.0599 dl:321-321 gd:1
+ttp: b883/3125 bl:2.2975 bb:1.0589 rl:2.3239 rb:1.0599 dl:319-319 gd:1
+ttp: b870/3125 bl:2.5789 bb:1.0972 rl:2.3241 rb:1.0599 dl:315-316 gd:1
+ttp: b864/3125 bl:2.6307 bb:1.2001 rl:2.3244 rb:1.0600 dl:314-314 gd:1
+ttp: b858/3125 bl:2.4209 bb:1.1513 rl:2.3245 rb:1.0601 dl:312-312 gd:1
+ttp: b845/3125 bl:2.3132 bb:0.9991 rl:2.3245 rb:1.0600 dl:308-309 gd:1
+ttp: b838/3125 bl:2.3051 bb:1.1071 rl:2.3245 rb:1.0601 dl:306-307 gd:1
+ttp: b830/3125 bl:2.5503 bb:1.1451 rl:2.3246 rb:1.0602 dl:304-305 gd:1
+ttp: b823/3125 bl:2.2760 bb:1.0400 rl:2.3246 rb:1.0601 dl:302-303 gd:1
+ttp: b751/3125 bl:2.2824 bb:1.0568 rl:2.3246 rb:1.0601 dl:283-283 gd:1
+ttp: b745/3125 bl:2.3684 bb:1.1096 rl:2.3246 rb:1.0602 dl:281-281 gd:1
+ttp: b737/3125 bl:2.4034 bb:1.0454 rl:2.3247 rb:1.0602 dl:279-279 gd:1
+ttp: b727/3125 bl:2.3324 bb:1.0940 rl:2.3247 rb:1.0602 dl:277-277 gd:1
+ttp: b721/3125 bl:2.3250 bb:1.1513 rl:2.3247 rb:1.0603 dl:275-275 gd:1
+ttp: b713/3125 bl:2.1843 bb:1.0484 rl:2.3246 rb:1.0602 dl:273-273 gd:1
+ttp: b704/3125 bl:2.4947 bb:1.2714 rl:2.3247 rb:1.0604 dl:271-271 gd:1
+ttp: b696/3125 bl:2.5160 bb:1.1952 rl:2.3248 rb:1.0605 dl:269-269 gd:1
+ttp: b687/3125 bl:2.5075 bb:1.1468 rl:2.3250 rb:1.0605 dl:267-267 gd:1
+ttp: b680/3125 bl:2.4904 bb:1.1746 rl:2.3251 rb:1.0606 dl:265-265 gd:1
+ttp: b670/3125 bl:2.4106 bb:1.1568 rl:2.3252 rb:1.0607 dl:263-263 gd:1
+ttp: b662/3125 bl:2.6596 bb:1.1645 rl:2.3254 rb:1.0608 dl:261-261 gd:1
+ttp: b653/3125 bl:2.6084 bb:1.2352 rl:2.3256 rb:1.0609 dl:259-259 gd:1
+ttp: b647/3125 bl:2.4113 bb:1.0619 rl:2.3257 rb:1.0609 dl:257-257 gd:1
+ttp: b639/3125 bl:2.4253 bb:1.1244 rl:2.3257 rb:1.0609 dl:255-255 gd:1
+ttp: b629/3125 bl:2.1482 bb:0.9755 rl:2.3256 rb:1.0609 dl:253-253 gd:1
+ttp: b621/3125 bl:2.4137 bb:1.2234 rl:2.3257 rb:1.0610 dl:251-251 gd:1
+ttp: b611/3125 bl:2.3286 bb:1.0597 rl:2.3257 rb:1.0610 dl:248-249 gd:1
+ttp: b610/3125 bl:2.4399 bb:1.1092 rl:2.3258 rb:1.0610 dl:248-248 gd:1
+ttp: b602/3125 bl:2.5042 bb:1.0901 rl:2.3259 rb:1.0610 dl:246-246 gd:1
+ttp: b592/3125 bl:2.1603 bb:1.0253 rl:2.3258 rb:1.0610 dl:244-244 gd:1
+ttp: b582/3125 bl:2.5587 bb:1.2231 rl:2.3259 rb:1.0611 dl:241-242 gd:1
+ttp: b580/3125 bl:2.6011 bb:1.2198 rl:2.3261 rb:1.0612 dl:241-241 gd:1
+ttp: b571/3125 bl:2.4181 bb:1.1682 rl:2.3262 rb:1.0613 dl:239-239 gd:1
+ttp: b569/3125 bl:2.4798 bb:1.1430 rl:2.3263 rb:1.0613 dl:238-238 gd:1
+ttp: b559/3125 bl:2.4095 bb:1.1065 rl:2.3263 rb:1.0614 dl:236-236 gd:1
+ttp: b556/3125 bl:2.4242 bb:1.1802 rl:2.3264 rb:1.0614 dl:235-235 gd:1
+ttp: b544/3125 bl:2.3532 bb:1.1376 rl:2.3264 rb:1.0615 dl:232-233 gd:1
+ttp: b541/3125 bl:2.3598 bb:1.1425 rl:2.3264 rb:1.0615 dl:232-232 gd:1
+ttp: b539/3125 bl:2.4337 bb:1.1930 rl:2.3265 rb:1.0616 dl:231-231 gd:1
+ttp: b528/3125 bl:2.4049 bb:1.1397 rl:2.3265 rb:1.0617 dl:229-229 gd:1
+ttp: b518/3125 bl:2.6316 bb:1.2040 rl:2.3267 rb:1.0618 dl:226-227 gd:1
+ttp: b517/3125 bl:2.3125 bb:1.0444 rl:2.3267 rb:1.0617 dl:226-226 gd:1
+ttp: b507/3125 bl:2.4896 bb:1.1283 rl:2.3268 rb:1.0618 dl:224-224 gd:1
+ttp: b496/3125 bl:2.2633 bb:1.1219 rl:2.3268 rb:1.0618 dl:221-222 gd:1
+ttp: b495/3125 bl:2.5895 bb:1.2862 rl:2.3270 rb:1.0619 dl:221-221 gd:1
+ttp: b486/3125 bl:2.4256 bb:1.0764 rl:2.3270 rb:1.0620 dl:219-219 gd:1
+ttp: b475/3125 bl:2.3540 bb:1.1388 rl:2.3270 rb:1.0620 dl:216-217 gd:1
+ttp: b466/3125 bl:2.5826 bb:1.2505 rl:2.3272 rb:1.0621 dl:214-215 gd:1
+ttp: b465/3125 bl:2.4364 bb:1.2379 rl:2.3272 rb:1.0622 dl:214-214 gd:1
+ttp: b456/3125 bl:2.4365 bb:1.1941 rl:2.3273 rb:1.0623 dl:212-212 gd:1
+ttp: b446/3125 bl:2.5426 bb:1.2351 rl:2.3274 rb:1.0624 dl:210-210 gd:1
+ttp: b437/3125 bl:2.3292 bb:1.0541 rl:2.3274 rb:1.0624 dl:208-208 gd:1
+ttp: b435/3125 bl:2.4625 bb:1.1759 rl:2.3275 rb:1.0624 dl:207-207 gd:1
+ttp: b425/3125 bl:2.3908 bb:1.1897 rl:2.3275 rb:1.0625 dl:205-205 gd:1
+ttp: b422/3125 bl:2.4456 bb:1.2136 rl:2.3276 rb:1.0626 dl:204-204 gd:1
+ttp: b409/3125 bl:2.3710 bb:1.2227 rl:2.3276 rb:1.0626 dl:202-202 gd:1
+ttp: b407/3125 bl:2.7307 bb:1.2643 rl:2.3279 rb:1.0627 dl:201-201 gd:1
+ttp: b397/3125 bl:2.5125 bb:1.2601 rl:2.3280 rb:1.0628 dl:199-199 gd:1
+ttp: b386/3125 bl:2.4983 bb:1.1981 rl:2.3280 rb:1.0629 dl:196-197 gd:1
+ttp: b384/3125 bl:2.3730 bb:1.1321 rl:2.3281 rb:1.0629 dl:196-196 gd:1
+ttp: b375/3125 bl:2.6272 bb:1.2261 rl:2.3282 rb:1.0630 dl:194-194 gd:1
+ttp: b367/3125 bl:2.4560 bb:1.1808 rl:2.3283 rb:1.0631 dl:192-192 gd:1
+ttp: b356/3125 bl:2.5497 bb:1.1714 rl:2.3284 rb:1.0631 dl:189-190 gd:1
+ttp: b354/3125 bl:2.3583 bb:1.0864 rl:2.3284 rb:1.0632 dl:189-189 gd:1
+ttp: b344/3125 bl:2.5413 bb:1.1545 rl:2.3285 rb:1.0632 dl:187-187 gd:1
+ttp: b342/3125 bl:2.4769 bb:1.1598 rl:2.3286 rb:1.0632 dl:186-186 gd:1
+ttp: b331/3125 bl:2.4787 bb:1.1264 rl:2.3287 rb:1.0633 dl:184-184 gd:1
+ttp: b321/3125 bl:2.4532 bb:1.1312 rl:2.3287 rb:1.0633 dl:182-182 gd:1
+ttp: b313/3125 bl:2.5533 bb:1.2366 rl:2.3288 rb:1.0634 dl:180-180 gd:1
+ttp: b303/3125 bl:2.3509 bb:1.1719 rl:2.3289 rb:1.0634 dl:178-178 gd:1
+ttp: b301/3125 bl:2.5332 bb:1.1833 rl:2.3290 rb:1.0635 dl:177-177 gd:1
+ttp: b293/3125 bl:2.3908 bb:1.1007 rl:2.3290 rb:1.0635 dl:175-175 gd:1
+ttp: b282/3125 bl:2.4407 bb:1.1544 rl:2.3290 rb:1.0636 dl:173-173 gd:1
+ttp: b270/3125 bl:2.5386 bb:1.2794 rl:2.3291 rb:1.0636 dl:170-171 gd:1
+ttp: b269/3125 bl:2.3845 bb:1.1095 rl:2.3292 rb:1.0637 dl:170-170 gd:1
+ttp: b259/3125 bl:2.4676 bb:1.1846 rl:2.3292 rb:1.0637 dl:168-168 gd:1
+ttp: b251/3125 bl:2.7448 bb:1.2903 rl:2.3294 rb:1.0638 dl:166-166 gd:1
+ttp: b242/3125 bl:2.3539 bb:1.2561 rl:2.3294 rb:1.0639 dl:164-164 gd:1
+ttp: b233/3125 bl:2.4537 bb:1.0542 rl:2.3295 rb:1.0639 dl:162-162 gd:1
+ttp: b225/3125 bl:2.4754 bb:1.1887 rl:2.3295 rb:1.0639 dl:160-160 gd:1
+ttp: b217/3125 bl:2.3019 bb:1.0532 rl:2.3295 rb:1.0639 dl:158-158 gd:1
+ttp: b206/3125 bl:2.7452 bb:1.2357 rl:2.3297 rb:1.0640 dl:155-156 gd:1
+ttp: b199/3125 bl:2.4747 bb:1.2712 rl:2.3298 rb:1.0641 dl:153-154 gd:1
+ttp: b191/3125 bl:2.4681 bb:1.2076 rl:2.3298 rb:1.0641 dl:151-152 gd:1
+ttp: b183/3125 bl:2.5375 bb:1.2484 rl:2.3299 rb:1.0642 dl:150-150 gd:1
+ttp: b182/3125 bl:2.6826 bb:1.2145 rl:2.3300 rb:1.0643 dl:149-149 gd:1
+ttp: b167/3125 bl:2.4455 bb:1.2259 rl:2.3301 rb:1.0643 dl:146-146 gd:1
+ttp: b166/3125 bl:2.4147 bb:1.1653 rl:2.3301 rb:1.0644 dl:145-145 gd:1
+ttp: b156/3125 bl:2.5026 bb:1.1769 rl:2.3302 rb:1.0644 dl:143-143 gd:1
+ttp: b142/3125 bl:2.5552 bb:1.1953 rl:2.3303 rb:1.0645 dl:139-140 gd:1
+ttp: b134/3125 bl:2.6231 bb:1.1573 rl:2.3304 rb:1.0645 dl:137-138 gd:1
+ttp: b126/3125 bl:2.5872 bb:1.2076 rl:2.3305 rb:1.0645 dl:135-136 gd:1
+ttp: b118/3125 bl:2.5778 bb:1.2853 rl:2.3306 rb:1.0646 dl:133-134 gd:1
+ttp: b117/3125 bl:2.6042 bb:1.2163 rl:2.3307 rb:1.0647 dl:133-133 gd:1
+ttp: b103/3125 bl:2.7223 bb:1.4525 rl:2.3308 rb:1.0648 dl:129-130 gd:1
+ttp: b96/3125 bl:2.5811 bb:1.1935 rl:2.3309 rb:1.0648 dl:128-128 gd:1
+ttp: b90/3125 bl:2.7589 bb:1.2690 rl:2.3310 rb:1.0649 dl:126-126 gd:1
+ttp: b84/3125 bl:2.7225 bb:1.1818 rl:2.3312 rb:1.0649 dl:124-124 gd:1
+ttp: b73/3125 bl:2.4937 bb:1.1617 rl:2.3312 rb:1.0650 dl:121-121 gd:1
+ttp: b63/3125 bl:2.6582 bb:1.3073 rl:2.3313 rb:1.0650 dl:117-118 gd:1
+ttp: b54/3125 bl:2.6367 bb:1.2221 rl:2.3314 rb:1.0651 dl:114-115 gd:1
+ttp: b51/3125 bl:2.6176 bb:1.1493 rl:2.3315 rb:1.0651 dl:113-113 gd:1
+ttp: b41/3125 bl:2.5702 bb:1.1419 rl:2.3316 rb:1.0651 dl:109-109 gd:1
+ttp: b31/3125 bl:2.7817 bb:1.2884 rl:2.3317 rb:1.0652 dl:104-105 gd:1
+ttp: b26/3125 bl:2.6482 bb:1.1438 rl:2.3318 rb:1.0652 dl:102-102 gd:1
+ttp: b17/3125 bl:2.6717 bb:1.2231 rl:2.3319 rb:1.0653 dl:96-97 gd:1
+ttp: b12/3125 bl:2.4581 bb:1.1265 rl:2.3319 rb:1.0653 dl:92-93 gd:1
+ttp: b3/3125 bl:2.8697 bb:1.1969 rl:2.3320 rb:1.0653 dl:79-81 gd:1
+quantized_ttt_phased val_loss:2.32529863 val_bpb:1.06257073 eval_time:587671ms
+total_eval_time:587.7s
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed42.log b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed42.log
new file mode 100644
index 0000000000..edce3a6db6
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed42.log
@@ -0,0 +1,209 @@
+ttp: b1662/3125 bl:2.3431 bb:1.0303 rl:2.3159 rb:1.0611 dl:613-614 gd:1
+ttp: b1656/3125 bl:2.2861 bb:1.0830 rl:2.3159 rb:1.0612 dl:611-611 gd:1
+ttp: b1649/3125 bl:2.5226 bb:1.0977 rl:2.3163 rb:1.0612 dl:607-607 gd:1
+ttp: b1641/3125 bl:2.3883 bb:1.0702 rl:2.3164 rb:1.0612 dl:603-603 gd:1
+ttp: b1631/3125 bl:2.4157 bb:1.1188 rl:2.3166 rb:1.0614 dl:598-598 gd:1
+ttp: b1624/3125 bl:2.4566 bb:1.1213 rl:2.3169 rb:1.0615 dl:595-595 gd:1
+ttp: b1614/3125 bl:2.4457 bb:1.1402 rl:2.3171 rb:1.0616 dl:590-591 gd:1
+ttp: b1605/3125 bl:2.2000 bb:1.0407 rl:2.3169 rb:1.0616 dl:586-587 gd:1
+ttp: b1597/3125 bl:2.3293 bb:1.0208 rl:2.3169 rb:1.0615 dl:582-583 gd:1
+ttp: b1591/3125 bl:2.2107 bb:0.9882 rl:2.3167 rb:1.0614 dl:579-580 gd:1
+ttp: b1581/3125 bl:2.2376 bb:1.0543 rl:2.3166 rb:1.0613 dl:575-576 gd:1
+ttp: b1576/3125 bl:2.1508 bb:0.9542 rl:2.3163 rb:1.0611 dl:573-573 gd:1
+ttp: b1566/3125 bl:2.3691 bb:1.0311 rl:2.3164 rb:1.0611 dl:568-569 gd:1
+ttp: b1557/3125 bl:2.3040 bb:1.0233 rl:2.3164 rb:1.0610 dl:564-565 gd:1
+ttp: b1548/3125 bl:2.3628 bb:1.0866 rl:2.3164 rb:1.0611 dl:560-561 gd:1
+ttp: b1543/3125 bl:2.3261 bb:1.0801 rl:2.3165 rb:1.0611 dl:558-558 gd:1
+ttp: b1532/3125 bl:2.3406 bb:1.0348 rl:2.3165 rb:1.0610 dl:553-554 gd:1
+ttp: b1527/3125 bl:2.3565 bb:1.1016 rl:2.3166 rb:1.0611 dl:551-551 gd:1
+ttp: b1515/3125 bl:2.5268 bb:1.1096 rl:2.3169 rb:1.0612 dl:546-547 gd:1
+ttp: b1508/3125 bl:2.1477 bb:1.0322 rl:2.3166 rb:1.0612 dl:543-544 gd:1
+ttp: b1499/3125 bl:2.3812 bb:1.0358 rl:2.3167 rb:1.0611 dl:539-540 gd:1
+ttp: b1493/3125 bl:2.2299 bb:1.0269 rl:2.3166 rb:1.0611 dl:537-537 gd:1
+ttp: b1486/3125 bl:2.2294 bb:0.9855 rl:2.3165 rb:1.0609 dl:534-534 gd:1
+ttp: b1475/3125 bl:2.3243 bb:1.0522 rl:2.3165 rb:1.0609 dl:529-530 gd:1
+ttp: b1469/3125 bl:2.5323 bb:1.1565 rl:2.3168 rb:1.0611 dl:527-527 gd:1
+ttp: b1460/3125 bl:2.2912 bb:1.0318 rl:2.3168 rb:1.0610 dl:523-524 gd:1
+ttp: b1453/3125 bl:2.3124 bb:1.0708 rl:2.3168 rb:1.0610 dl:521-521 gd:1
+ttp: b1445/3125 bl:2.5046 bb:1.1158 rl:2.3171 rb:1.0611 dl:517-517 gd:1
+ttp: b1438/3125 bl:2.4646 bb:1.1084 rl:2.3173 rb:1.0612 dl:514-514 gd:1
+ttp: b1428/3125 bl:2.2503 bb:1.0333 rl:2.3172 rb:1.0612 dl:509-510 gd:1
+ttp: b1422/3125 bl:2.2288 bb:1.0219 rl:2.3171 rb:1.0611 dl:507-507 gd:1
+ttp: b1414/3125 bl:2.0854 bb:0.9660 rl:2.3167 rb:1.0609 dl:503-503 gd:1
+ttp: b1401/3125 bl:2.3524 bb:1.0738 rl:2.3168 rb:1.0610 dl:498-499 gd:1
+ttp: b1394/3125 bl:2.1788 bb:1.0005 rl:2.3166 rb:1.0609 dl:495-496 gd:1
+ttp: b1390/3125 bl:2.2522 bb:1.0148 rl:2.3165 rb:1.0608 dl:494-494 gd:1
+ttp: b1382/3125 bl:2.5724 bb:1.1214 rl:2.3168 rb:1.0609 dl:491-491 gd:1
+ttp: b1373/3125 bl:2.4415 bb:1.0475 rl:2.3170 rb:1.0609 dl:488-488 gd:1
+ttp: b1364/3125 bl:2.3296 bb:1.0906 rl:2.3170 rb:1.0609 dl:484-485 gd:1
+ttp: b1360/3125 bl:2.5676 bb:1.1304 rl:2.3174 rb:1.0610 dl:483-483 gd:1
+ttp: b1351/3125 bl:2.5126 bb:1.1563 rl:2.3177 rb:1.0612 dl:480-480 gd:1
+ttp: b1343/3125 bl:2.4695 bb:1.1397 rl:2.3179 rb:1.0613 dl:476-477 gd:1
+ttp: b1335/3125 bl:2.3794 bb:1.0657 rl:2.3180 rb:1.0613 dl:473-474 gd:1
+ttp: b1332/3125 bl:2.4201 bb:1.0788 rl:2.3182 rb:1.0613 dl:472-472 gd:1
+ttp: b1323/3125 bl:2.4053 bb:1.1092 rl:2.3183 rb:1.0614 dl:469-469 gd:1
+ttp: b1316/3125 bl:2.3342 bb:1.0921 rl:2.3183 rb:1.0614 dl:466-466 gd:1
+ttp: b1306/3125 bl:2.1627 bb:1.0190 rl:2.3181 rb:1.0614 dl:462-463 gd:1
+ttp: b1301/3125 bl:2.2725 bb:1.0798 rl:2.3180 rb:1.0614 dl:461-461 gd:1
+ttp: b1293/3125 bl:2.4554 bb:1.1106 rl:2.3182 rb:1.0615 dl:458-458 gd:1
+ttp: b1285/3125 bl:2.3385 bb:1.0558 rl:2.3182 rb:1.0615 dl:455-455 gd:1
+ttp: b1275/3125 bl:2.1772 bb:0.9617 rl:2.3180 rb:1.0613 dl:450-451 gd:1
+ttp: b1268/3125 bl:2.2848 bb:1.0769 rl:2.3180 rb:1.0613 dl:448-448 gd:1
+ttp: b1260/3125 bl:2.5005 bb:1.1251 rl:2.3182 rb:1.0614 dl:445-445 gd:1
+ttp: b1253/3125 bl:2.2552 bb:1.0932 rl:2.3182 rb:1.0615 dl:442-442 gd:1
+ttp: b1244/3125 bl:2.4107 bb:1.0742 rl:2.3183 rb:1.0615 dl:439-439 gd:1
+ttp: b1234/3125 bl:2.3185 bb:1.1172 rl:2.3183 rb:1.0615 dl:435-436 gd:1
+ttp: b1227/3125 bl:2.3082 bb:1.2320 rl:2.3183 rb:1.0617 dl:433-433 gd:1
+ttp: b1220/3125 bl:2.3473 bb:1.0632 rl:2.3183 rb:1.0617 dl:430-430 gd:1
+ttp: b1212/3125 bl:2.3243 bb:1.1033 rl:2.3183 rb:1.0618 dl:427-427 gd:1
+ttp: b1204/3125 bl:2.3192 bb:1.1169 rl:2.3183 rb:1.0619 dl:424-424 gd:1
+ttp: b1196/3125 bl:2.3544 bb:1.0616 rl:2.3184 rb:1.0619 dl:421-421 gd:1
+ttp: b1187/3125 bl:2.4534 bb:1.1534 rl:2.3185 rb:1.0620 dl:417-418 gd:1
+ttp: b1179/3125 bl:2.2654 bb:1.0186 rl:2.3185 rb:1.0619 dl:414-415 gd:1
+ttp: b1173/3125 bl:2.1594 bb:1.0237 rl:2.3183 rb:1.0619 dl:412-412 gd:1
+ttp: b1164/3125 bl:2.3513 bb:1.0775 rl:2.3183 rb:1.0619 dl:408-409 gd:1
+ttp: b1160/3125 bl:2.5243 bb:1.0951 rl:2.3186 rb:1.0619 dl:407-407 gd:1
+ttp: b1154/3125 bl:2.2594 bb:1.0155 rl:2.3185 rb:1.0619 dl:405-405 gd:1
+ttp: b1144/3125 bl:2.3849 bb:1.1077 rl:2.3186 rb:1.0619 dl:401-402 gd:1
+ttp: b1138/3125 bl:2.3859 bb:1.0942 rl:2.3186 rb:1.0620 dl:400-400 gd:1
+ttp: b1134/3125 bl:2.4784 bb:1.1173 rl:2.3188 rb:1.0620 dl:398-398 gd:1
+ttp: b1124/3125 bl:2.3191 bb:1.1677 rl:2.3188 rb:1.0621 dl:395-395 gd:1
+ttp: b1119/3125 bl:2.2882 bb:1.0946 rl:2.3188 rb:1.0622 dl:393-393 gd:1
+ttp: b1109/3125 bl:2.1137 bb:1.0467 rl:2.3186 rb:1.0622 dl:389-390 gd:1
+ttp: b1105/3125 bl:2.2860 bb:1.1030 rl:2.3185 rb:1.0622 dl:388-388 gd:1
+ttp: b1095/3125 bl:2.2831 bb:1.0841 rl:2.3185 rb:1.0622 dl:385-385 gd:1
+ttp: b1088/3125 bl:2.2647 bb:1.0819 rl:2.3184 rb:1.0622 dl:383-383 gd:1
+ttp: b1079/3125 bl:2.2230 bb:1.0149 rl:2.3183 rb:1.0622 dl:379-380 gd:1
+ttp: b1073/3125 bl:2.2358 bb:1.0912 rl:2.3182 rb:1.0622 dl:378-378 gd:1
+ttp: b1067/3125 bl:2.5242 bb:1.1961 rl:2.3185 rb:1.0624 dl:376-376 gd:1
+ttp: b1062/3125 bl:2.4379 bb:1.1520 rl:2.3186 rb:1.0625 dl:374-374 gd:1
+ttp: b1056/3125 bl:2.4395 bb:1.0933 rl:2.3187 rb:1.0625 dl:372-372 gd:1
+ttp: b1043/3125 bl:2.4482 bb:1.1499 rl:2.3188 rb:1.0626 dl:368-369 gd:1
+ttp: b1040/3125 bl:2.2472 bb:1.0367 rl:2.3188 rb:1.0626 dl:367-367 gd:1
+ttp: b1028/3125 bl:2.4296 bb:1.0870 rl:2.3189 rb:1.0626 dl:363-364 gd:1
+ttp: b990/3125 bl:2.2782 bb:1.1744 rl:2.3188 rb:1.0627 dl:352-352 gd:1
+ttp: b985/3125 bl:2.4312 bb:1.0917 rl:2.3190 rb:1.0627 dl:350-350 gd:1
+ttp: b974/3125 bl:2.2495 bb:1.1124 rl:2.3189 rb:1.0628 dl:346-347 gd:1
+ttp: b969/3125 bl:2.4888 bb:1.1221 rl:2.3191 rb:1.0628 dl:345-345 gd:1
+ttp: b959/3125 bl:2.3213 bb:1.0792 rl:2.3191 rb:1.0628 dl:342-342 gd:1
+ttp: b954/3125 bl:2.5120 bb:1.1507 rl:2.3193 rb:1.0629 dl:340-340 gd:1
+ttp: b944/3125 bl:2.4403 bb:1.1338 rl:2.3194 rb:1.0630 dl:337-337 gd:1
+ttp: b938/3125 bl:2.3248 bb:1.1348 rl:2.3194 rb:1.0631 dl:335-335 gd:1
+ttp: b931/3125 bl:2.3943 bb:1.1416 rl:2.3194 rb:1.0631 dl:333-333 gd:1
+ttp: b918/3125 bl:2.5492 bb:1.1622 rl:2.3197 rb:1.0632 dl:329-330 gd:1
+ttp: b915/3125 bl:2.3237 bb:1.1212 rl:2.3197 rb:1.0633 dl:328-328 gd:1
+ttp: b901/3125 bl:2.3313 bb:1.0157 rl:2.3197 rb:1.0632 dl:324-325 gd:1
+ttp: b895/3125 bl:2.3854 bb:1.1577 rl:2.3197 rb:1.0633 dl:323-323 gd:1
+ttp: b889/3125 bl:2.2918 bb:1.1047 rl:2.3197 rb:1.0633 dl:321-321 gd:1
+ttp: b877/3125 bl:2.6204 bb:1.2088 rl:2.3200 rb:1.0635 dl:317-318 gd:1
+ttp: b871/3125 bl:2.2826 bb:1.0657 rl:2.3199 rb:1.0635 dl:316-316 gd:1
+ttp: b865/3125 bl:2.5567 bb:1.1390 rl:2.3202 rb:1.0635 dl:314-314 gd:1
+ttp: b859/3125 bl:2.3186 bb:1.0581 rl:2.3202 rb:1.0635 dl:312-312 gd:1
+ttp: b846/3125 bl:2.4703 bb:1.0791 rl:2.3203 rb:1.0636 dl:309-309 gd:1
+ttp: b839/3125 bl:2.3482 bb:1.0486 rl:2.3203 rb:1.0635 dl:307-307 gd:1
+ttp: b831/3125 bl:2.2835 bb:1.1285 rl:2.3203 rb:1.0636 dl:305-305 gd:1
+ttp: b824/3125 bl:2.4834 bb:1.1261 rl:2.3204 rb:1.0636 dl:303-303 gd:1
+ttp: b817/3125 bl:2.5059 bb:1.1847 rl:2.3206 rb:1.0637 dl:301-301 gd:1
+ttp: b803/3125 bl:2.4235 bb:1.0751 rl:2.3207 rb:1.0638 dl:297-298 gd:1
+ttp: b796/3125 bl:2.4641 bb:1.2265 rl:2.3208 rb:1.0639 dl:295-296 gd:1
+ttp: b789/3125 bl:2.4018 bb:1.0631 rl:2.3209 rb:1.0639 dl:294-294 gd:1
+ttp: b784/3125 bl:2.3745 bb:1.0554 rl:2.3209 rb:1.0639 dl:292-292 gd:1
+ttp: b772/3125 bl:2.2475 bb:1.0358 rl:2.3208 rb:1.0638 dl:288-289 gd:1
+ttp: b771/3125 bl:2.3079 bb:1.1478 rl:2.3208 rb:1.0639 dl:288-288 gd:1
+ttp: b758/3125 bl:2.4884 bb:1.2018 rl:2.3210 rb:1.0640 dl:285-285 gd:1
+ttp: b752/3125 bl:2.3916 bb:1.1762 rl:2.3210 rb:1.0641 dl:283-283 gd:1
+ttp: b745/3125 bl:2.3747 bb:1.1125 rl:2.3211 rb:1.0641 dl:281-281 gd:1
+ttp: b737/3125 bl:2.4062 bb:1.0466 rl:2.3211 rb:1.0641 dl:279-279 gd:1
+ttp: b727/3125 bl:2.3157 bb:1.0861 rl:2.3211 rb:1.0641 dl:277-277 gd:1
+ttp: b721/3125 bl:2.3372 bb:1.1574 rl:2.3211 rb:1.0642 dl:275-275 gd:1
+ttp: b713/3125 bl:2.1878 bb:1.0501 rl:2.3210 rb:1.0642 dl:273-273 gd:1
+ttp: b704/3125 bl:2.4847 bb:1.2663 rl:2.3212 rb:1.0643 dl:271-271 gd:1
+ttp: b697/3125 bl:2.4145 bb:1.1813 rl:2.3212 rb:1.0644 dl:269-269 gd:1
+ttp: b688/3125 bl:2.4633 bb:1.1026 rl:2.3213 rb:1.0644 dl:267-267 gd:1
+ttp: b673/3125 bl:2.2663 bb:1.0035 rl:2.3213 rb:1.0644 dl:263-264 gd:1
+ttp: b671/3125 bl:2.3505 bb:1.1534 rl:2.3213 rb:1.0645 dl:263-263 gd:1
+ttp: b663/3125 bl:2.5697 bb:1.1356 rl:2.3215 rb:1.0645 dl:261-261 gd:1
+ttp: b654/3125 bl:2.4284 bb:1.2494 rl:2.3216 rb:1.0646 dl:259-259 gd:1
+ttp: b648/3125 bl:2.2740 bb:1.0628 rl:2.3215 rb:1.0646 dl:257-257 gd:1
+ttp: b639/3125 bl:2.4391 bb:1.1309 rl:2.3216 rb:1.0647 dl:255-255 gd:1
+ttp: b629/3125 bl:2.1573 bb:0.9796 rl:2.3215 rb:1.0646 dl:253-253 gd:1
+ttp: b621/3125 bl:2.4101 bb:1.2216 rl:2.3216 rb:1.0647 dl:251-251 gd:1
+ttp: b612/3125 bl:2.1494 bb:1.0099 rl:2.3214 rb:1.0647 dl:249-249 gd:1
+ttp: b605/3125 bl:2.2390 bb:1.1376 rl:2.3214 rb:1.0647 dl:247-247 gd:1
+ttp: b597/3125 bl:2.1864 bb:1.0811 rl:2.3213 rb:1.0647 dl:245-245 gd:1
+ttp: b589/3125 bl:2.4903 bb:1.1740 rl:2.3214 rb:1.0648 dl:243-243 gd:1
+ttp: b579/3125 bl:2.4009 bb:1.0539 rl:2.3215 rb:1.0648 dl:241-241 gd:1
+ttp: b571/3125 bl:2.4242 bb:1.1711 rl:2.3215 rb:1.0649 dl:239-239 gd:1
+ttp: b563/3125 bl:2.5703 bb:1.2677 rl:2.3217 rb:1.0650 dl:237-237 gd:1
+ttp: b554/3125 bl:2.3968 bb:1.0552 rl:2.3217 rb:1.0650 dl:235-235 gd:1
+ttp: b544/3125 bl:2.3237 bb:1.1233 rl:2.3217 rb:1.0650 dl:232-233 gd:1
+ttp: b543/3125 bl:2.5079 bb:1.1281 rl:2.3219 rb:1.0651 dl:232-232 gd:1
+ttp: b533/3125 bl:2.5242 bb:1.2074 rl:2.3220 rb:1.0651 dl:230-230 gd:1
+ttp: b523/3125 bl:2.3085 bb:1.0670 rl:2.3220 rb:1.0651 dl:228-228 gd:1
+ttp: b515/3125 bl:2.4513 bb:1.1581 rl:2.3221 rb:1.0652 dl:226-226 gd:1
+ttp: b506/3125 bl:2.4327 bb:1.2245 rl:2.3221 rb:1.0653 dl:224-224 gd:1
+ttp: b496/3125 bl:2.2742 bb:1.1273 rl:2.3221 rb:1.0653 dl:221-222 gd:1
+ttp: b489/3125 bl:2.2420 bb:1.1033 rl:2.3221 rb:1.0653 dl:219-220 gd:1
+ttp: b488/3125 bl:2.4497 bb:1.1877 rl:2.3221 rb:1.0654 dl:219-219 gd:1
+ttp: b477/3125 bl:2.3048 bb:1.1590 rl:2.3221 rb:1.0655 dl:217-217 gd:1
+ttp: b468/3125 bl:2.4146 bb:1.1508 rl:2.3222 rb:1.0655 dl:215-215 gd:1
+ttp: b459/3125 bl:2.3898 bb:1.1225 rl:2.3222 rb:1.0655 dl:213-213 gd:1
+ttp: b449/3125 bl:2.6808 bb:1.1958 rl:2.3224 rb:1.0656 dl:210-211 gd:1
+ttp: b448/3125 bl:2.4382 bb:1.1972 rl:2.3225 rb:1.0657 dl:210-210 gd:1
+ttp: b440/3125 bl:2.3568 bb:1.1675 rl:2.3225 rb:1.0657 dl:208-208 gd:1
+ttp: b431/3125 bl:2.3986 bb:1.1778 rl:2.3225 rb:1.0658 dl:206-206 gd:1
+ttp: b420/3125 bl:2.2975 bb:1.1692 rl:2.3225 rb:1.0659 dl:204-204 gd:1
+ttp: b409/3125 bl:2.3750 bb:1.2248 rl:2.3226 rb:1.0659 dl:202-202 gd:1
+ttp: b399/3125 bl:2.1993 bb:1.0590 rl:2.3225 rb:1.0659 dl:199-200 gd:1
+ttp: b391/3125 bl:2.8510 bb:1.2927 rl:2.3228 rb:1.0661 dl:197-198 gd:1
+ttp: b389/3125 bl:2.5632 bb:1.1914 rl:2.3229 rb:1.0661 dl:197-197 gd:1
+ttp: b380/3125 bl:2.2377 bb:1.0549 rl:2.3229 rb:1.0661 dl:195-195 gd:1
+ttp: b372/3125 bl:2.4119 bb:1.1479 rl:2.3229 rb:1.0662 dl:193-193 gd:1
+ttp: b362/3125 bl:2.6049 bb:1.2181 rl:2.3231 rb:1.0662 dl:191-191 gd:1
+ttp: b352/3125 bl:2.5586 bb:1.1907 rl:2.3232 rb:1.0663 dl:188-189 gd:1
+ttp: b343/3125 bl:2.5025 bb:1.2124 rl:2.3233 rb:1.0664 dl:186-187 gd:1
+ttp: b341/3125 bl:2.3292 bb:1.0265 rl:2.3233 rb:1.0663 dl:186-186 gd:1
+ttp: b330/3125 bl:2.4169 bb:1.1451 rl:2.3233 rb:1.0664 dl:184-184 gd:1
+ttp: b320/3125 bl:2.3082 bb:1.1173 rl:2.3233 rb:1.0664 dl:181-182 gd:1
+ttp: b312/3125 bl:2.5593 bb:1.1864 rl:2.3234 rb:1.0665 dl:179-180 gd:1
+ttp: b311/3125 bl:2.7077 bb:1.2690 rl:2.3236 rb:1.0666 dl:179-179 gd:1
+ttp: b299/3125 bl:2.6807 bb:1.1698 rl:2.3238 rb:1.0666 dl:177-177 gd:1
+ttp: b291/3125 bl:2.5337 bb:1.2193 rl:2.3239 rb:1.0667 dl:175-175 gd:1
+ttp: b280/3125 bl:2.5293 bb:1.2524 rl:2.3240 rb:1.0668 dl:173-173 gd:1
+ttp: b277/3125 bl:2.4487 bb:1.2155 rl:2.3240 rb:1.0668 dl:172-172 gd:1
+ttp: b267/3125 bl:2.5752 bb:1.1436 rl:2.3242 rb:1.0669 dl:170-170 gd:1
+ttp: b257/3125 bl:2.4073 bb:1.1574 rl:2.3242 rb:1.0669 dl:167-168 gd:1
+ttp: b249/3125 bl:2.5115 bb:1.2392 rl:2.3243 rb:1.0670 dl:166-166 gd:1
+ttp: b240/3125 bl:2.3887 bb:1.1993 rl:2.3243 rb:1.0670 dl:163-164 gd:1
+ttp: b239/3125 bl:2.5651 bb:1.2031 rl:2.3244 rb:1.0671 dl:163-163 gd:1
+ttp: b224/3125 bl:2.4648 bb:1.2291 rl:2.3245 rb:1.0672 dl:159-160 gd:1
+ttp: b217/3125 bl:2.3094 bb:1.0566 rl:2.3245 rb:1.0671 dl:158-158 gd:1
+ttp: b207/3125 bl:2.8197 bb:1.2780 rl:2.3247 rb:1.0672 dl:156-156 gd:1
+ttp: b200/3125 bl:2.3880 bb:1.1586 rl:2.3247 rb:1.0673 dl:154-154 gd:1
+ttp: b192/3125 bl:2.4836 bb:1.2025 rl:2.3248 rb:1.0673 dl:152-152 gd:1
+ttp: b184/3125 bl:2.4547 bb:1.2111 rl:2.3248 rb:1.0674 dl:150-150 gd:1
+ttp: b174/3125 bl:2.5150 bb:1.1488 rl:2.3249 rb:1.0674 dl:147-148 gd:1
+ttp: b168/3125 bl:2.5988 bb:1.2656 rl:2.3250 rb:1.0675 dl:146-146 gd:1
+ttp: b159/3125 bl:2.7331 bb:1.1812 rl:2.3252 rb:1.0675 dl:144-144 gd:1
+ttp: b157/3125 bl:2.6037 bb:1.1770 rl:2.3253 rb:1.0676 dl:143-143 gd:1
+ttp: b142/3125 bl:2.5552 bb:1.1953 rl:2.3253 rb:1.0676 dl:139-140 gd:1
+ttp: b135/3125 bl:2.5705 bb:1.1720 rl:2.3254 rb:1.0677 dl:138-138 gd:1
+ttp: b127/3125 bl:2.6560 bb:1.2047 rl:2.3256 rb:1.0677 dl:136-136 gd:1
+ttp: b121/3125 bl:2.4368 bb:1.1636 rl:2.3256 rb:1.0677 dl:134-134 gd:1
+ttp: b113/3125 bl:2.6515 bb:1.2039 rl:2.3257 rb:1.0678 dl:132-132 gd:1
+ttp: b106/3125 bl:2.7688 bb:1.3074 rl:2.3259 rb:1.0679 dl:130-130 gd:1
+ttp: b98/3125 bl:2.6451 bb:1.1708 rl:2.3260 rb:1.0679 dl:128-128 gd:1
+ttp: b85/3125 bl:2.5506 bb:1.2863 rl:2.3260 rb:1.0680 dl:124-125 gd:1
+ttp: b79/3125 bl:2.6908 bb:1.2276 rl:2.3262 rb:1.0680 dl:123-123 gd:1
+ttp: b69/3125 bl:2.5022 bb:1.1188 rl:2.3262 rb:1.0680 dl:119-120 gd:1
+ttp: b65/3125 bl:2.6165 bb:1.2971 rl:2.3263 rb:1.0681 dl:118-118 gd:1
+ttp: b56/3125 bl:2.5371 bb:1.1769 rl:2.3264 rb:1.0681 dl:115-115 gd:1
+ttp: b48/3125 bl:2.5898 bb:1.1359 rl:2.3265 rb:1.0682 dl:112-112 gd:1
+ttp: b38/3125 bl:2.6268 bb:1.2013 rl:2.3265 rb:1.0682 dl:107-108 gd:1
+ttp: b33/3125 bl:2.8351 bb:1.2748 rl:2.3267 rb:1.0683 dl:105-105 gd:1
+ttp: b24/3125 bl:2.6617 bb:1.1626 rl:2.3268 rb:1.0683 dl:101-101 gd:1
+ttp: b14/3125 bl:2.7732 bb:1.2727 rl:2.3269 rb:1.0683 dl:94-95 gd:1
+ttp: b9/3125 bl:2.7567 bb:1.1938 rl:2.3270 rb:1.0684 dl:89-90 gd:1
+ttp: b1/3125 bl:2.8249 bb:1.1866 rl:2.3271 rb:1.0684 dl:27-75 gd:1
+quantized_ttt_phased val_loss:2.32475990 val_bpb:1.06232455 eval_time:485616ms
+total_eval_time:485.6s
diff --git a/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed999.log b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed999.log
new file mode 100644
index 0000000000..68eaae3486
--- /dev/null
+++ b/records/track_10min_16mb/2026-04-29_ReLU-Slope_Reverse-Cholesky-GPTQ_SP8192_CaseOps_PhasedTTT_PolarNS_FusedCE/train_seed999.log
@@ -0,0 +1,1188 @@
+W0429 18:43:07.243000 249914 torch/distributed/run.py:803] 
+W0429 18:43:07.243000 249914 torch/distributed/run.py:803] *****************************************
+W0429 18:43:07.243000 249914 torch/distributed/run.py:803] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
+W0429 18:43:07.243000 249914 torch/distributed/run.py:803] *****************************************
+train_shards: 80
+val_tokens: 47851520
+model_params:35945671
+gptq:reserving 16s, effective=584000ms
+warmup_cu_buckets:64,128,192,256 iters_each:3
+warmup_step: 1/20
+warmup_step: 2/20
+warmup_step: 3/20
+warmup_step: 4/20
+warmup_step: 5/20
+warmup_step: 6/20
+warmup_step: 10/20
+warmup_step: 20/20
+loop_warmup:enabled encoder:[0, 1, 2, 3, 4, 5, 3, 4] decoder:[5, 3, 4, 5, 6, 7, 8, 9, 10]
+loop_warmup_step: 1/20
+loop_warmup_step: 2/20
+loop_warmup_step: 3/20
+loop_warmup_step: 4/20
+loop_warmup_step: 5/20
+loop_warmup_step: 6/20
+loop_warmup_step: 10/20
+loop_warmup_step: 20/20
+0/20000 val_loss: 9.0149 val_bpb: 4.1192
+1/20000 train_loss: 9.0152 train_time: 0.0m tok/s: 12476399
+2/20000 train_loss: 12.8832 train_time: 0.0m tok/s: 11464028
+3/20000 train_loss: 10.1976 train_time: 0.0m tok/s: 10222934
+4/20000 train_loss: 8.6674 train_time: 0.0m tok/s: 9673623
+5/20000 train_loss: 7.9217 train_time: 0.0m tok/s: 9385556
+500/20000 train_loss: 2.5712 train_time: 0.8m tok/s: 8290006
+1000/20000 train_loss: 2.7985 train_time: 1.6m tok/s: 8272115
+1500/20000 train_loss: 2.6277 train_time: 2.4m tok/s: 8266041
+2000/20000 train_loss: 2.6622 train_time: 3.2m tok/s: 8261486
+layer_loop:enabled step:2147 frac:0.350 encoder:[0, 1, 2, 3, 4, 5, 3, 4] decoder:[5, 3, 4, 5, 6, 7, 8, 9, 10]
+2500/20000 train_loss: 2.5521 train_time: 4.2m tok/s: 7742735
+3000/20000 train_loss: 2.5630 train_time: 5.4m tok/s: 7279323
+3500/20000 train_loss: 2.5649 train_time: 6.6m tok/s: 6960793
+4000/20000 train_loss: 2.4087 train_time: 7.8m tok/s: 6757643
+4000/20000 val_loss: 2.4278 val_bpb: 1.1093
+4500/20000 train_loss: 2.2761 train_time: 8.9m tok/s: 6608036
+4846/20000 val_loss: 2.3578 val_bpb: 1.0774
+stopping_early: wallclock_cap train_time: 584084ms step: 4846/20000
+peak memory allocated: 41709 MiB reserved: 47026 MiB
+ema:applying EMA weights
+diagnostic pre-quantization post-ema val_loss:2.33492623 val_bpb:1.06690089 eval_time:6763ms
+Serialized model: 135419597 bytes
+Code size (uncompressed): 138859 bytes
+Code size (compressed): 31095 bytes
+GPTQ:collecting Hessians from calibration data...
+GPTQ:collected 67 Hessians in 3.5s
+Quantized weights:
+  gate_int8_row: blocks.attn.attn_gate_w
+  gptq (int6): blocks.attn.c_k.weight, blocks.attn.c_q.weight, blocks.attn.c_v.weight, blocks.attn.proj.weight, blocks.mlp.fc.weight, blocks.mlp.proj.weight
+  gptq (int7)+lqer_asym: tok_emb.weight
+  passthrough (float16): blocks.attn.q_gain, blocks.attn_scale, blocks.mlp_scale, blocks.resid_mix, parallel_post_lambdas, parallel_resid_lambdas, skip_gates, skip_weights, smear_gate.weight, smear_lambda
+GPTQ:quantized weights in 12.15s
+Serialized model quantized+brotli: 15915437 bytes
+Total submission size quantized+brotli: 15946532 bytes
+diagnostic quantized val_loss:2.35389768 val_bpb:1.07556954 eval_time:10311ms
+ttt_lora:warming up compile (random tokens, no val data)
+ttt_lora:compile warmup done (133.9s)
+
+beginning TTT eval timer
+ttt_phased: total_docs:50000 prefix_docs:2000 suffix_docs:48000 num_phases:3 boundaries:[666, 1333, 2000]
+ttp: b3124/3125 bl:2.1816 bb:1.0808 rl:2.1816 rb:1.0808 dl:24255-30330 gd:0
+ttp: b3107/3125 bl:2.3408 bb:1.1003 rl:2.2210 rb:1.0859 dl:8823-8998 gd:0
+ttp: b3101/3125 bl:2.1964 bb:1.0606 rl:2.2167 rb:1.0814 dl:7534-7751 gd:0
+ttp: b3094/3125 bl:2.3777 bb:1.1127 rl:2.2378 rb:1.0857 dl:6523-6670 gd:0
+ttp: b3087/3125 bl:2.2163 bb:1.0335 rl:2.2355 rb:1.0799 dl:5910-5996 gd:0
+ttp: b3078/3125 bl:2.0107 bb:0.9559 rl:2.2159 rb:1.0690 dl:5341-5395 gd:0
+ttpp: phase:1/3 pd:784 gd:666 t:156.9s
+tttg: c1/158 lr:0.001000 t:0.3s
+tttg: c2/158 lr:0.001000 t:0.3s
+tttg: c3/158 lr:0.001000 t:0.4s
+tttg: c4/158 lr:0.000999 t:0.5s
+tttg: c5/158 lr:0.000998 t:0.5s
+tttg: c6/158 lr:0.000997 t:0.6s
+tttg: c7/158 lr:0.000996 t:0.7s
+tttg: c8/158 lr:0.000995 t:0.7s
+tttg: c9/158 lr:0.000994 t:0.8s
+tttg: c10/158 lr:0.000992 t:0.9s
+tttg: c11/158 lr:0.000990 t:1.0s
+tttg: c12/158 lr:0.000988 t:1.0s
+tttg: c13/158 lr:0.000986 t:1.1s
+tttg: c14/158 lr:0.000983 t:1.2s
+tttg: c15/158 lr:0.000981 t:1.2s
+tttg: c16/158 lr:0.000978 t:1.3s
+tttg: c17/158 lr:0.000975 t:1.4s
+tttg: c18/158 lr:0.000971 t:1.4s
+tttg: c19/158 lr:0.000968 t:1.5s
+tttg: c20/158 lr:0.000964 t:1.6s
+tttg: c21/158 lr:0.000960 t:1.6s
+tttg: c22/158 lr:0.000957 t:1.7s
+tttg: c23/158 lr:0.000952 t:1.8s
+tttg: c24/158 lr:0.000948 t:1.8s
+tttg: c25/158 lr:0.000943 t:1.9s
+tttg: c26/158 lr:0.000939 t:2.0s
+tttg: c27/158 lr:0.000934 t:2.1s
+tttg: c28/158 lr:0.000929 t:2.1s
+tttg: c29/158 lr:0.000924 t:2.2s
+tttg: c30/158 lr:0.000918 t:2.3s
+tttg: c31/158 lr:0.000913 t:2.3s
+tttg: c32/158 lr:0.000907 t:2.4s
+tttg: c33/158 lr:0.000901 t:2.5s
+tttg: c34/158 lr:0.000895 t:2.5s
+tttg: c35/158 lr:0.000889 t:2.6s
+tttg: c36/158 lr:0.000882 t:2.7s
+tttg: c37/158 lr:0.000876 t:2.7s
+tttg: c38/158 lr:0.000869 t:2.8s
+tttg: c39/158 lr:0.000862 t:2.9s
+tttg: c40/158 lr:0.000855 t:3.0s
+tttg: c41/158 lr:0.000848 t:3.0s
+tttg: c42/158 lr:0.000841 t:3.1s
+tttg: c43/158 lr:0.000834 t:3.2s
+tttg: c44/158 lr:0.000826 t:3.2s
+tttg: c45/158 lr:0.000818 t:3.3s
+tttg: c46/158 lr:0.000811 t:3.4s
+tttg: c47/158 lr:0.000803 t:3.4s
+tttg: c48/158 lr:0.000795 t:3.5s
+tttg: c49/158 lr:0.000787 t:3.6s
+tttg: c50/158 lr:0.000778 t:3.6s
+tttg: c51/158 lr:0.000770 t:3.7s
+tttg: c52/158 lr:0.000761 t:3.8s
+tttg: c53/158 lr:0.000753 t:3.9s
+tttg: c54/158 lr:0.000744 t:3.9s
+tttg: c55/158 lr:0.000735 t:4.0s
+tttg: c56/158 lr:0.000727 t:4.1s
+tttg: c57/158 lr:0.000718 t:4.1s
+tttg: c58/158 lr:0.000709 t:4.2s
+tttg: c59/158 lr:0.000699 t:4.3s
+tttg: c60/158 lr:0.000690 t:4.3s
+tttg: c61/158 lr:0.000681 t:4.4s
+tttg: c62/158 lr:0.000672 t:4.5s
+tttg: c63/158 lr:0.000662 t:4.5s
+tttg: c64/158 lr:0.000653 t:4.6s
+tttg: c65/158 lr:0.000643 t:4.7s
+tttg: c66/158 lr:0.000633 t:4.8s
+tttg: c67/158 lr:0.000624 t:4.8s
+tttg: c68/158 lr:0.000614 t:4.9s
+tttg: c69/158 lr:0.000604 t:5.0s
+tttg: c70/158 lr:0.000594 t:5.0s
+tttg: c71/158 lr:0.000585 t:5.1s
+tttg: c72/158 lr:0.000575 t:5.2s
+tttg: c73/158 lr:0.000565 t:5.2s
+tttg: c74/158 lr:0.000555 t:5.3s
+tttg: c75/158 lr:0.000545 t:5.4s
+tttg: c76/158 lr:0.000535 t:5.4s
+tttg: c77/158 lr:0.000525 t:5.5s
+tttg: c78/158 lr:0.000515 t:5.6s
+tttg: c79/158 lr:0.000505 t:5.7s
+tttg: c80/158 lr:0.000495 t:5.7s
+tttg: c81/158 lr:0.000485 t:5.8s
+tttg: c82/158 lr:0.000475 t:5.9s
+tttg: c83/158 lr:0.000465 t:5.9s
+tttg: c84/158 lr:0.000455 t:6.0s
+tttg: c85/158 lr:0.000445 t:6.1s
+tttg: c86/158 lr:0.000435 t:6.1s
+tttg: c87/158 lr:0.000425 t:6.2s
+tttg: c88/158 lr:0.000415 t:6.3s
+tttg: c89/158 lr:0.000406 t:6.3s
+tttg: c90/158 lr:0.000396 t:6.4s
+tttg: c91/158 lr:0.000386 t:6.5s
+tttg: c92/158 lr:0.000376 t:6.5s
+tttg: c93/158 lr:0.000367 t:6.6s
+tttg: c94/158 lr:0.000357 t:6.7s
+tttg: c95/158 lr:0.000347 t:6.8s
+tttg: c96/158 lr:0.000338 t:6.8s
+tttg: c97/158 lr:0.000328 t:6.9s
+tttg: c98/158 lr:0.000319 t:7.0s
+tttg: c99/158 lr:0.000310 t:7.0s
+tttg: c100/158 lr:0.000301 t:7.1s
+tttg: c101/158 lr:0.000291 t:7.2s
+tttg: c102/158 lr:0.000282 t:7.2s
+tttg: c103/158 lr:0.000273 t:7.3s
+tttg: c104/158 lr:0.000265 t:7.4s
+tttg: c105/158 lr:0.000256 t:7.4s
+tttg: c106/158 lr:0.000247 t:7.5s
+tttg: c107/158 lr:0.000239 t:7.6s
+tttg: c108/158 lr:0.000230 t:7.7s
+tttg: c109/158 lr:0.000222 t:7.7s
+tttg: c110/158 lr:0.000213 t:7.8s
+tttg: c111/158 lr:0.000205 t:7.9s
+tttg: c112/158 lr:0.000197 t:7.9s
+tttg: c113/158 lr:0.000189 t:8.0s
+tttg: c114/158 lr:0.000182 t:8.1s
+tttg: c115/158 lr:0.000174 t:8.1s
+tttg: c116/158 lr:0.000166 t:8.2s
+tttg: c117/158 lr:0.000159 t:8.3s
+tttg: c118/158 lr:0.000152 t:8.3s
+tttg: c119/158 lr:0.000145 t:8.4s
+tttg: c120/158 lr:0.000138 t:8.5s
+tttg: c121/158 lr:0.000131 t:8.6s
+tttg: c122/158 lr:0.000124 t:8.6s
+tttg: c123/158 lr:0.000118 t:8.7s
+tttg: c124/158 lr:0.000111 t:8.8s
+tttg: c125/158 lr:0.000105 t:8.8s
+tttg: c126/158 lr:0.000099 t:8.9s
+tttg: c127/158 lr:0.000093 t:9.0s
+tttg: c128/158 lr:0.000087 t:9.0s
+tttg: c129/158 lr:0.000082 t:9.1s
+tttg: c130/158 lr:0.000076 t:9.2s
+tttg: c131/158 lr:0.000071 t:9.2s
+tttg: c132/158 lr:0.000066 t:9.3s
+tttg: c133/158 lr:0.000061 t:9.4s
+tttg: c134/158 lr:0.000057 t:9.5s
+tttg: c135/158 lr:0.000052 t:9.5s
+tttg: c136/158 lr:0.000048 t:9.6s
+tttg: c137/158 lr:0.000043 t:9.7s
+tttg: c138/158 lr:0.000040 t:9.7s
+tttg: c139/158 lr:0.000036 t:9.8s
+tttg: c140/158 lr:0.000032 t:9.9s
+tttg: c141/158 lr:0.000029 t:9.9s
+tttg: c142/158 lr:0.000025 t:10.0s
+tttg: c143/158 lr:0.000022 t:10.1s
+tttg: c144/158 lr:0.000019 t:10.1s
+tttg: c145/158 lr:0.000017 t:10.2s
+tttg: c146/158 lr:0.000014 t:10.3s
+tttg: c147/158 lr:0.000012 t:10.4s
+tttg: c148/158 lr:0.000010 t:10.4s
+tttg: c149/158 lr:0.000008 t:10.5s
+tttg: c150/158 lr:0.000006 t:10.6s
+tttg: c151/158 lr:0.000005 t:10.6s
+tttg: c152/158 lr:0.000004 t:10.7s
+tttg: c153/158 lr:0.000003 t:10.8s
+tttg: c154/158 lr:0.000002 t:10.8s
+tttg: c155/158 lr:0.000001 t:10.9s
+tttg: c156/158 lr:0.000000 t:11.0s
+tttg: c157/158 lr:0.000000 t:11.0s
+ttpr: phase:1/3 t:168.4s
+ttp: b3073/3125 bl:2.3394 bb:1.0608 rl:2.2254 rb:1.0683 dl:5097-5130 gd:0
+ttp: b3062/3125 bl:2.3048 bb:1.0755 rl:2.2305 rb:1.0688 dl:4558-4601 gd:0
+ttp: b3053/3125 bl:2.2821 bb:1.0531 rl:2.2334 rb:1.0678 dl:4284-4309 gd:0
+ttp: b3046/3125 bl:2.2467 bb:1.0418 rl:2.2341 rb:1.0665 dl:4058-4077 gd:0
+ttp: b3037/3125 bl:2.2410 bb:1.0223 rl:2.2344 rb:1.0644 dl:3817-3837 gd:0
+ttpp: phase:2/3 pd:1456 gd:1333 t:237.9s
+tttg: c1/247 lr:0.001000 t:0.1s
+tttg: c2/247 lr:0.001000 t:0.1s
+tttg: c3/247 lr:0.001000 t:0.2s
+tttg: c4/247 lr:0.001000 t:0.3s
+tttg: c5/247 lr:0.000999 t:0.3s
+tttg: c6/247 lr:0.000999 t:0.4s
+tttg: c7/247 lr:0.000999 t:0.5s
+tttg: c8/247 lr:0.000998 t:0.5s
+tttg: c9/247 lr:0.000997 t:0.6s
+tttg: c10/247 lr:0.000997 t:0.7s
+tttg: c11/247 lr:0.000996 t:0.7s
+tttg: c12/247 lr:0.000995 t:0.8s
+tttg: c13/247 lr:0.000994 t:0.9s
+tttg: c14/247 lr:0.000993 t:0.9s
+tttg: c15/247 lr:0.000992 t:1.0s
+tttg: c16/247 lr:0.000991 t:1.1s
+tttg: c17/247 lr:0.000990 t:1.1s
+tttg: c18/247 lr:0.000988 t:1.2s
+tttg: c19/247 lr:0.000987 t:1.3s
+tttg: c20/247 lr:0.000985 t:1.3s
+tttg: c21/247 lr:0.000984 t:1.4s
+tttg: c22/247 lr:0.000982 t:1.5s
+tttg: c23/247 lr:0.000980 t:1.5s
+tttg: c24/247 lr:0.000979 t:1.6s
+tttg: c25/247 lr:0.000977 t:1.7s
+tttg: c26/247 lr:0.000975 t:1.8s
+tttg: c27/247 lr:0.000973 t:1.8s
+tttg: c28/247 lr:0.000971 t:1.9s
+tttg: c29/247 lr:0.000968 t:2.0s
+tttg: c30/247 lr:0.000966 t:2.0s
+tttg: c31/247 lr:0.000964 t:2.1s
+tttg: c32/247 lr:0.000961 t:2.2s
+tttg: c33/247 lr:0.000959 t:2.2s
+tttg: c34/247 lr:0.000956 t:2.3s
+tttg: c35/247 lr:0.000954 t:2.4s
+tttg: c36/247 lr:0.000951 t:2.4s
+tttg: c37/247 lr:0.000948 t:2.5s
+tttg: c38/247 lr:0.000945 t:2.6s
+tttg: c39/247 lr:0.000942 t:2.6s
+tttg: c40/247 lr:0.000939 t:2.7s
+tttg: c41/247 lr:0.000936 t:2.8s
+tttg: c42/247 lr:0.000933 t:2.8s
+tttg: c43/247 lr:0.000930 t:2.9s
+tttg: c44/247 lr:0.000926 t:3.0s
+tttg: c45/247 lr:0.000923 t:3.0s
+tttg: c46/247 lr:0.000920 t:3.1s
+tttg: c47/247 lr:0.000916 t:3.2s
+tttg: c48/247 lr:0.000913 t:3.2s
+tttg: c49/247 lr:0.000909 t:3.3s
+tttg: c50/247 lr:0.000905 t:3.4s
+tttg: c51/247 lr:0.000901 t:3.5s
+tttg: c52/247 lr:0.000898 t:3.5s
+tttg: c53/247 lr:0.000894 t:3.6s
+tttg: c54/247 lr:0.000890 t:3.7s
+tttg: c55/247 lr:0.000886 t:3.7s
+tttg: c56/247 lr:0.000882 t:3.8s
+tttg: c57/247 lr:0.000877 t:3.9s
+tttg: c58/247 lr:0.000873 t:3.9s
+tttg: c59/247 lr:0.000869 t:4.0s
+tttg: c60/247 lr:0.000865 t:4.1s
+tttg: c61/247 lr:0.000860 t:4.1s
+tttg: c62/247 lr:0.000856 t:4.2s
+tttg: c63/247 lr:0.000851 t:4.3s
+tttg: c64/247 lr:0.000847 t:4.3s
+tttg: c65/247 lr:0.000842 t:4.4s
+tttg: c66/247 lr:0.000837 t:4.5s
+tttg: c67/247 lr:0.000833 t:4.5s
+tttg: c68/247 lr:0.000828 t:4.6s
+tttg: c69/247 lr:0.000823 t:4.7s
+tttg: c70/247 lr:0.000818 t:4.7s
+tttg: c71/247 lr:0.000813 t:4.8s
+tttg: c72/247 lr:0.000808 t:4.9s
+tttg: c73/247 lr:0.000803 t:4.9s
+tttg: c74/247 lr:0.000798 t:5.0s
+tttg: c75/247 lr:0.000793 t:5.1s
+tttg: c76/247 lr:0.000788 t:5.1s
+tttg: c77/247 lr:0.000782 t:5.2s
+tttg: c78/247 lr:0.000777 t:5.3s
+tttg: c79/247 lr:0.000772 t:5.4s
+tttg: c80/247 lr:0.000766 t:5.4s
+tttg: c81/247 lr:0.000761 t:5.5s
+tttg: c82/247 lr:0.000756 t:5.6s
+tttg: c83/247 lr:0.000750 t:5.6s
+tttg: c84/247 lr:0.000744 t:5.7s
+tttg: c85/247 lr:0.000739 t:5.8s
+tttg: c86/247 lr:0.000733 t:5.8s
+tttg: c87/247 lr:0.000728 t:5.9s
+tttg: c88/247 lr:0.000722 t:6.0s
+tttg: c89/247 lr:0.000716 t:6.0s
+tttg: c90/247 lr:0.000710 t:6.1s
+tttg: c91/247 lr:0.000705 t:6.2s
+tttg: c92/247 lr:0.000699 t:6.2s
+tttg: c93/247 lr:0.000693 t:6.3s
+tttg: c94/247 lr:0.000687 t:6.4s
+tttg: c95/247 lr:0.000681 t:6.4s
+tttg: c96/247 lr:0.000675 t:6.5s
+tttg: c97/247 lr:0.000669 t:6.6s
+tttg: c98/247 lr:0.000663 t:6.6s
+tttg: c99/247 lr:0.000657 t:6.7s
+tttg: c100/247 lr:0.000651 t:6.8s
+tttg: c101/247 lr:0.000645 t:6.8s
+tttg: c102/247 lr:0.000639 t:6.9s
+tttg: c103/247 lr:0.000632 t:7.0s
+tttg: c104/247 lr:0.000626 t:7.0s
+tttg: c105/247 lr:0.000620 t:7.1s
+tttg: c106/247 lr:0.000614 t:7.2s
+tttg: c107/247 lr:0.000608 t:7.2s
+tttg: c108/247 lr:0.000601 t:7.3s
+tttg: c109/247 lr:0.000595 t:7.4s
+tttg: c110/247 lr:0.000589 t:7.4s
+tttg: c111/247 lr:0.000583 t:7.5s
+tttg: c112/247 lr:0.000576 t:7.6s
+tttg: c113/247 lr:0.000570 t:7.7s
+tttg: c114/247 lr:0.000564 t:7.7s
+tttg: c115/247 lr:0.000557 t:7.8s
+tttg: c116/247 lr:0.000551 t:7.9s
+tttg: c117/247 lr:0.000545 t:7.9s
+tttg: c118/247 lr:0.000538 t:8.0s
+tttg: c119/247 lr:0.000532 t:8.1s
+tttg: c120/247 lr:0.000526 t:8.1s
+tttg: c121/247 lr:0.000519 t:8.2s
+tttg: c122/247 lr:0.000513 t:8.3s
+tttg: c123/247 lr:0.000506 t:8.3s
+tttg: c124/247 lr:0.000500 t:8.4s
+tttg: c125/247 lr:0.000494 t:8.5s
+tttg: c126/247 lr:0.000487 t:8.5s
+tttg: c127/247 lr:0.000481 t:8.6s
+tttg: c128/247 lr:0.000474 t:8.7s
+tttg: c129/247 lr:0.000468 t:8.7s
+tttg: c130/247 lr:0.000462 t:8.8s
+tttg: c131/247 lr:0.000455 t:8.9s
+tttg: c132/247 lr:0.000449 t:8.9s
+tttg: c133/247 lr:0.000443 t:9.0s
+tttg: c134/247 lr:0.000436 t:9.1s
+tttg: c135/247 lr:0.000430 t:9.1s
+tttg: c136/247 lr:0.000424 t:9.2s
+tttg: c137/247 lr:0.000417 t:9.3s
+tttg: c138/247 lr:0.000411 t:9.3s
+tttg: c139/247 lr:0.000405 t:9.4s
+tttg: c140/247 lr:0.000399 t:9.5s
+tttg: c141/247 lr:0.000392 t:9.6s
+tttg: c142/247 lr:0.000386 t:9.6s
+tttg: c143/247 lr:0.000380 t:9.7s
+tttg: c144/247 lr:0.000374 t:9.8s
+tttg: c145/247 lr:0.000368 t:9.8s
+tttg: c146/247 lr:0.000361 t:9.9s
+tttg: c147/247 lr:0.000355 t:10.0s
+tttg: c148/247 lr:0.000349 t:10.0s
+tttg: c149/247 lr:0.000343 t:10.1s
+tttg: c150/247 lr:0.000337 t:10.2s
+tttg: c151/247 lr:0.000331 t:10.2s
+tttg: c152/247 lr:0.000325 t:10.3s
+tttg: c153/247 lr:0.000319 t:10.4s
+tttg: c154/247 lr:0.000313 t:10.4s
+tttg: c155/247 lr:0.000307 t:10.5s
+tttg: c156/247 lr:0.000301 t:10.6s
+tttg: c157/247 lr:0.000295 t:10.6s
+tttg: c158/247 lr:0.000290 t:10.7s
+tttg: c159/247 lr:0.000284 t:10.8s
+tttg: c160/247 lr:0.000278 t:10.8s
+tttg: c161/247 lr:0.000272 t:10.9s
+tttg: c162/247 lr:0.000267 t:11.0s
+tttg: c163/247 lr:0.000261 t:11.0s
+tttg: c164/247 lr:0.000256 t:11.1s
+tttg: c165/247 lr:0.000250 t:11.2s
+tttg: c166/247 lr:0.000244 t:11.2s
+tttg: c167/247 lr:0.000239 t:11.3s
+tttg: c168/247 lr:0.000234 t:11.4s
+tttg: c169/247 lr:0.000228 t:11.5s
+tttg: c170/247 lr:0.000223 t:11.5s
+tttg: c171/247 lr:0.000218 t:11.6s
+tttg: c172/247 lr:0.000212 t:11.7s
+tttg: c173/247 lr:0.000207 t:11.7s
+tttg: c174/247 lr:0.000202 t:11.8s
+tttg: c175/247 lr:0.000197 t:11.9s
+tttg: c176/247 lr:0.000192 t:11.9s
+tttg: c177/247 lr:0.000187 t:12.0s
+tttg: c178/247 lr:0.000182 t:12.1s
+tttg: c179/247 lr:0.000177 t:12.1s
+tttg: c180/247 lr:0.000172 t:12.2s
+tttg: c181/247 lr:0.000167 t:12.3s
+tttg: c182/247 lr:0.000163 t:12.3s
+tttg: c183/247 lr:0.000158 t:12.4s
+tttg: c184/247 lr:0.000153 t:12.5s
+tttg: c185/247 lr:0.000149 t:12.5s
+tttg: c186/247 lr:0.000144 t:12.6s
+tttg: c187/247 lr:0.000140 t:12.7s
+tttg: c188/247 lr:0.000135 t:12.7s
+tttg: c189/247 lr:0.000131 t:12.8s
+tttg: c190/247 lr:0.000127 t:12.9s
+tttg: c191/247 lr:0.000123 t:12.9s
+tttg: c192/247 lr:0.000118 t:13.0s
+tttg: c193/247 lr:0.000114 t:13.1s
+tttg: c194/247 lr:0.000110 t:13.1s
+tttg: c195/247 lr:0.000106 t:13.2s
+tttg: c196/247 lr:0.000102 t:13.3s
+tttg: c197/247 lr:0.000099 t:13.4s
+tttg: c198/247 lr:0.000095 t:13.4s
+tttg: c199/247 lr:0.000091 t:13.5s
+tttg: c200/247 lr:0.000087 t:13.6s
+tttg: c201/247 lr:0.000084 t:13.6s
+tttg: c202/247 lr:0.000080 t:13.7s
+tttg: c203/247 lr:0.000077 t:13.8s
+tttg: c204/247 lr:0.000074 t:13.8s
+tttg: c205/247 lr:0.000070 t:13.9s
+tttg: c206/247 lr:0.000067 t:14.0s
+tttg: c207/247 lr:0.000064 t:14.0s
+tttg: c208/247 lr:0.000061 t:14.1s
+tttg: c209/247 lr:0.000058 t:14.2s
+tttg: c210/247 lr:0.000055 t:14.2s
+tttg: c211/247 lr:0.000052 t:14.3s
+tttg: c212/247 lr:0.000049 t:14.4s
+tttg: c213/247 lr:0.000046 t:14.4s
+tttg: c214/247 lr:0.000044 t:14.5s
+tttg: c215/247 lr:0.000041 t:14.6s
+tttg: c216/247 lr:0.000039 t:14.6s
+tttg: c217/247 lr:0.000036 t:14.7s
+tttg: c218/247 lr:0.000034 t:14.8s
+tttg: c219/247 lr:0.000032 t:14.8s
+tttg: c220/247 lr:0.000029 t:14.9s
+tttg: c221/247 lr:0.000027 t:15.0s
+tttg: c222/247 lr:0.000025 t:15.1s
+tttg: c223/247 lr:0.000023 t:15.1s
+tttg: c224/247 lr:0.000021 t:15.2s
+tttg: c225/247 lr:0.000020 t:15.3s
+tttg: c226/247 lr:0.000018 t:15.3s
+tttg: c227/247 lr:0.000016 t:15.4s
+tttg: c228/247 lr:0.000015 t:15.5s
+tttg: c229/247 lr:0.000013 t:15.5s
+tttg: c230/247 lr:0.000012 t:15.6s
+tttg: c231/247 lr:0.000010 t:15.7s
+tttg: c232/247 lr:0.000009 t:15.7s
+tttg: c233/247 lr:0.000008 t:15.8s
+tttg: c234/247 lr:0.000007 t:15.9s
+tttg: c235/247 lr:0.000006 t:15.9s
+tttg: c236/247 lr:0.000005 t:16.0s
+tttg: c237/247 lr:0.000004 t:16.1s
+tttg: c238/247 lr:0.000003 t:16.1s
+tttg: c239/247 lr:0.000003 t:16.2s
+tttg: c240/247 lr:0.000002 t:16.3s
+tttg: c241/247 lr:0.000001 t:16.3s
+tttg: c242/247 lr:0.000001 t:16.4s
+tttg: c243/247 lr:0.000001 t:16.5s
+tttg: c244/247 lr:0.000000 t:16.5s
+tttg: c245/247 lr:0.000000 t:16.6s
+tttg: c246/247 lr:0.000000 t:16.7s
+ttpr: phase:2/3 t:255.1s
+ttp: b3033/3125 bl:2.3738 bb:1.0907 rl:2.2404 rb:1.0655 dl:3741-3760 gd:0
+ttp: b3019/3125 bl:2.4930 bb:1.0991 rl:2.2500 rb:1.0669 dl:3436-3446 gd:0
+ttp: b3015/3125 bl:2.4088 bb:1.0622 rl:2.2557 rb:1.0667 dl:3366-3381 gd:0
+ttp: b3006/3125 bl:2.4826 bb:1.0952 rl:2.2633 rb:1.0677 dl:3233-3249 gd:0
+ttp: b3000/3125 bl:2.4149 bb:1.1294 rl:2.2680 rb:1.0697 dl:3135-3149 gd:0
+ttpp: phase:3/3 pd:2112 gd:2000 t:273.9s
+tttg: c1/319 lr:0.001000 t:0.1s
+tttg: c2/319 lr:0.001000 t:0.1s
+tttg: c3/319 lr:0.001000 t:0.2s
+tttg: c4/319 lr:0.001000 t:0.3s
+tttg: c5/319 lr:0.001000 t:0.3s
+tttg: c6/319 lr:0.000999 t:0.4s
+tttg: c7/319 lr:0.000999 t:0.5s
+tttg: c8/319 lr:0.000999 t:0.5s
+tttg: c9/319 lr:0.000998 t:0.6s
+tttg: c10/319 lr:0.000998 t:0.7s
+tttg: c11/319 lr:0.000998 t:0.7s
+tttg: c12/319 lr:0.000997 t:0.8s
+tttg: c13/319 lr:0.000996 t:0.9s
+tttg: c14/319 lr:0.000996 t:0.9s
+tttg: c15/319 lr:0.000995 t:1.0s
+tttg: c16/319 lr:0.000995 t:1.1s
+tttg: c17/319 lr:0.000994 t:1.1s
+tttg: c18/319 lr:0.000993 t:1.2s
+tttg: c19/319 lr:0.000992 t:1.3s
+tttg: c20/319 lr:0.000991 t:1.3s
+tttg: c21/319 lr:0.000990 t:1.4s
+tttg: c22/319 lr:0.000989 t:1.5s
+tttg: c23/319 lr:0.000988 t:1.5s
+tttg: c24/319 lr:0.000987 t:1.6s
+tttg: c25/319 lr:0.000986 t:1.7s
+tttg: c26/319 lr:0.000985 t:1.8s
+tttg: c27/319 lr:0.000984 t:1.8s
+tttg: c28/319 lr:0.000982 t:1.9s
+tttg: c29/319 lr:0.000981 t:2.0s
+tttg: c30/319 lr:0.000980 t:2.0s
+tttg: c31/319 lr:0.000978 t:2.1s
+tttg: c32/319 lr:0.000977 t:2.2s
+tttg: c33/319 lr:0.000975 t:2.2s
+tttg: c34/319 lr:0.000974 t:2.3s
+tttg: c35/319 lr:0.000972 t:2.4s
+tttg: c36/319 lr:0.000970 t:2.4s
+tttg: c37/319 lr:0.000969 t:2.5s
+tttg: c38/319 lr:0.000967 t:2.6s
+tttg: c39/319 lr:0.000965 t:2.6s
+tttg: c40/319 lr:0.000963 t:2.7s
+tttg: c41/319 lr:0.000961 t:2.8s
+tttg: c42/319 lr:0.000960 t:2.8s
+tttg: c43/319 lr:0.000958 t:2.9s
+tttg: c44/319 lr:0.000956 t:3.0s
+tttg: c45/319 lr:0.000954 t:3.0s
+tttg: c46/319 lr:0.000951 t:3.1s
+tttg: c47/319 lr:0.000949 t:3.2s
+tttg: c48/319 lr:0.000947 t:3.2s
+tttg: c49/319 lr:0.000945 t:3.3s
+tttg: c50/319 lr:0.000943 t:3.4s
+tttg: c51/319 lr:0.000940 t:3.4s
+tttg: c52/319 lr:0.000938 t:3.5s
+tttg: c53/319 lr:0.000935 t:3.6s
+tttg: c54/319 lr:0.000933 t:3.7s
+tttg: c55/319 lr:0.000931 t:3.7s
+tttg: c56/319 lr:0.000928 t:3.8s
+tttg: c57/319 lr:0.000925 t:3.9s
+tttg: c58/319 lr:0.000923 t:3.9s
+tttg: c59/319 lr:0.000920 t:4.0s
+tttg: c60/319 lr:0.000917 t:4.1s
+tttg: c61/319 lr:0.000915 t:4.1s
+tttg: c62/319 lr:0.000912 t:4.2s
+tttg: c63/319 lr:0.000909 t:4.3s
+tttg: c64/319 lr:0.000906 t:4.3s
+tttg: c65/319 lr:0.000903 t:4.4s
+tttg: c66/319 lr:0.000900 t:4.5s
+tttg: c67/319 lr:0.000897 t:4.5s
+tttg: c68/319 lr:0.000894 t:4.6s
+tttg: c69/319 lr:0.000891 t:4.7s
+tttg: c70/319 lr:0.000888 t:4.7s
+tttg: c71/319 lr:0.000885 t:4.8s
+tttg: c72/319 lr:0.000882 t:4.9s
+tttg: c73/319 lr:0.000879 t:4.9s
+tttg: c74/319 lr:0.000876 t:5.0s
+tttg: c75/319 lr:0.000872 t:5.1s
+tttg: c76/319 lr:0.000869 t:5.1s
+tttg: c77/319 lr:0.000866 t:5.2s
+tttg: c78/319 lr:0.000862 t:5.3s
+tttg: c79/319 lr:0.000859 t:5.3s
+tttg: c80/319 lr:0.000855 t:5.4s
+tttg: c81/319 lr:0.000852 t:5.5s
+tttg: c82/319 lr:0.000848 t:5.5s
+tttg: c83/319 lr:0.000845 t:5.6s
+tttg: c84/319 lr:0.000841 t:5.7s
+tttg: c85/319 lr:0.000837 t:5.7s
+tttg: c86/319 lr:0.000834 t:5.8s
+tttg: c87/319 lr:0.000830 t:5.9s
+tttg: c88/319 lr:0.000826 t:5.9s
+tttg: c89/319 lr:0.000823 t:6.0s
+tttg: c90/319 lr:0.000819 t:6.1s
+tttg: c91/319 lr:0.000815 t:6.2s
+tttg: c92/319 lr:0.000811 t:6.2s
+tttg: c93/319 lr:0.000807 t:6.3s
+tttg: c94/319 lr:0.000803 t:6.4s
+tttg: c95/319 lr:0.000799 t:6.4s
+tttg: c96/319 lr:0.000795 t:6.5s
+tttg: c97/319 lr:0.000791 t:6.6s
+tttg: c98/319 lr:0.000787 t:6.6s
+tttg: c99/319 lr:0.000783 t:6.7s
+tttg: c100/319 lr:0.000779 t:6.8s
+tttg: c101/319 lr:0.000775 t:6.8s
+tttg: c102/319 lr:0.000771 t:6.9s
+tttg: c103/319 lr:0.000767 t:7.0s
+tttg: c104/319 lr:0.000763 t:7.0s
+tttg: c105/319 lr:0.000759 t:7.1s
+tttg: c106/319 lr:0.000754 t:7.2s
+tttg: c107/319 lr:0.000750 t:7.2s
+tttg: c108/319 lr:0.000746 t:7.3s
+tttg: c109/319 lr:0.000741 t:7.4s
+tttg: c110/319 lr:0.000737 t:7.4s
+tttg: c111/319 lr:0.000733 t:7.5s
+tttg: c112/319 lr:0.000728 t:7.6s
+tttg: c113/319 lr:0.000724 t:7.6s
+tttg: c114/319 lr:0.000719 t:7.7s
+tttg: c115/319 lr:0.000715 t:7.8s
+tttg: c116/319 lr:0.000711 t:7.8s
+tttg: c117/319 lr:0.000706 t:7.9s
+tttg: c118/319 lr:0.000702 t:8.0s
+tttg: c119/319 lr:0.000697 t:8.0s
+tttg: c120/319 lr:0.000692 t:8.1s
+tttg: c121/319 lr:0.000688 t:8.2s
+tttg: c122/319 lr:0.000683 t:8.2s
+tttg: c123/319 lr:0.000679 t:8.3s
+tttg: c124/319 lr:0.000674 t:8.4s
+tttg: c125/319 lr:0.000669 t:8.5s
+tttg: c126/319 lr:0.000665 t:8.5s
+tttg: c127/319 lr:0.000660 t:8.6s
+tttg: c128/319 lr:0.000655 t:8.7s
+tttg: c129/319 lr:0.000651 t:8.7s
+tttg: c130/319 lr:0.000646 t:8.8s
+tttg: c131/319 lr:0.000641 t:8.9s
+tttg: c132/319 lr:0.000637 t:8.9s
+tttg: c133/319 lr:0.000632 t:9.0s
+tttg: c134/319 lr:0.000627 t:9.1s
+tttg: c135/319 lr:0.000622 t:9.1s
+tttg: c136/319 lr:0.000617 t:9.2s
+tttg: c137/319 lr:0.000613 t:9.3s
+tttg: c138/319 lr:0.000608 t:9.3s
+tttg: c139/319 lr:0.000603 t:9.4s
+tttg: c140/319 lr:0.000598 t:9.5s
+tttg: c141/319 lr:0.000593 t:9.5s
+tttg: c142/319 lr:0.000588 t:9.6s
+tttg: c143/319 lr:0.000584 t:9.7s
+tttg: c144/319 lr:0.000579 t:9.7s
+tttg: c145/319 lr:0.000574 t:9.8s
+tttg: c146/319 lr:0.000569 t:9.9s
+tttg: c147/319 lr:0.000564 t:9.9s
+tttg: c148/319 lr:0.000559 t:10.0s
+tttg: c149/319 lr:0.000554 t:10.1s
+tttg: c150/319 lr:0.000549 t:10.1s
+tttg: c151/319 lr:0.000544 t:10.2s
+tttg: c152/319 lr:0.000539 t:10.3s
+tttg: c153/319 lr:0.000535 t:10.4s
+tttg: c154/319 lr:0.000530 t:10.4s
+tttg: c155/319 lr:0.000525 t:10.5s
+tttg: c156/319 lr:0.000520 t:10.6s
+tttg: c157/319 lr:0.000515 t:10.6s
+tttg: c158/319 lr:0.000510 t:10.7s
+tttg: c159/319 lr:0.000505 t:10.8s
+tttg: c160/319 lr:0.000500 t:10.8s
+tttg: c161/319 lr:0.000495 t:10.9s
+tttg: c162/319 lr:0.000490 t:11.0s
+tttg: c163/319 lr:0.000485 t:11.0s
+tttg: c164/319 lr:0.000480 t:11.1s
+tttg: c165/319 lr:0.000475 t:11.2s
+tttg: c166/319 lr:0.000470 t:11.2s
+tttg: c167/319 lr:0.000465 t:11.3s
+tttg: c168/319 lr:0.000461 t:11.4s
+tttg: c169/319 lr:0.000456 t:11.4s
+tttg: c170/319 lr:0.000451 t:11.5s
+tttg: c171/319 lr:0.000446 t:11.6s
+tttg: c172/319 lr:0.000441 t:11.6s
+tttg: c173/319 lr:0.000436 t:11.7s
+tttg: c174/319 lr:0.000431 t:11.8s
+tttg: c175/319 lr:0.000426 t:11.8s
+tttg: c176/319 lr:0.000421 t:11.9s
+tttg: c177/319 lr:0.000416 t:12.0s
+tttg: c178/319 lr:0.000412 t:12.0s
+tttg: c179/319 lr:0.000407 t:12.1s
+tttg: c180/319 lr:0.000402 t:12.2s
+tttg: c181/319 lr:0.000397 t:12.2s
+tttg: c182/319 lr:0.000392 t:12.3s
+tttg: c183/319 lr:0.000387 t:12.4s
+tttg: c184/319 lr:0.000383 t:12.4s
+tttg: c185/319 lr:0.000378 t:12.5s
+tttg: c186/319 lr:0.000373 t:12.6s
+tttg: c187/319 lr:0.000368 t:12.7s
+tttg: c188/319 lr:0.000363 t:12.7s
+tttg: c189/319 lr:0.000359 t:12.8s
+tttg: c190/319 lr:0.000354 t:12.9s
+tttg: c191/319 lr:0.000349 t:12.9s
+tttg: c192/319 lr:0.000345 t:13.0s
+tttg: c193/319 lr:0.000340 t:13.1s
+tttg: c194/319 lr:0.000335 t:13.1s
+tttg: c195/319 lr:0.000331 t:13.2s
+tttg: c196/319 lr:0.000326 t:13.3s
+tttg: c197/319 lr:0.000321 t:13.3s
+tttg: c198/319 lr:0.000317 t:13.4s
+tttg: c199/319 lr:0.000312 t:13.5s
+tttg: c200/319 lr:0.000308 t:13.5s
+tttg: c201/319 lr:0.000303 t:13.6s
+tttg: c202/319 lr:0.000298 t:13.7s
+tttg: c203/319 lr:0.000294 t:13.7s
+tttg: c204/319 lr:0.000289 t:13.8s
+tttg: c205/319 lr:0.000285 t:13.9s
+tttg: c206/319 lr:0.000281 t:13.9s
+tttg: c207/319 lr:0.000276 t:14.0s
+tttg: c208/319 lr:0.000272 t:14.1s
+tttg: c209/319 lr:0.000267 t:14.1s
+tttg: c210/319 lr:0.000263 t:14.2s
+tttg: c211/319 lr:0.000259 t:14.3s
+tttg: c212/319 lr:0.000254 t:14.3s
+tttg: c213/319 lr:0.000250 t:14.4s
+tttg: c214/319 lr:0.000246 t:14.5s
+tttg: c215/319 lr:0.000241 t:14.6s
+tttg: c216/319 lr:0.000237 t:14.6s
+tttg: c217/319 lr:0.000233 t:14.7s
+tttg: c218/319 lr:0.000229 t:14.8s
+tttg: c219/319 lr:0.000225 t:14.8s
+tttg: c220/319 lr:0.000221 t:14.9s
+tttg: c221/319 lr:0.000217 t:15.0s
+tttg: c222/319 lr:0.000213 t:15.0s
+tttg: c223/319 lr:0.000209 t:15.1s
+tttg: c224/319 lr:0.000205 t:15.2s
+tttg: c225/319 lr:0.000201 t:15.2s
+tttg: c226/319 lr:0.000197 t:15.3s
+tttg: c227/319 lr:0.000193 t:15.4s
+tttg: c228/319 lr:0.000189 t:15.4s
+tttg: c229/319 lr:0.000185 t:15.5s
+tttg: c230/319 lr:0.000181 t:15.6s
+tttg: c231/319 lr:0.000177 t:15.6s
+tttg: c232/319 lr:0.000174 t:15.7s
+tttg: c233/319 lr:0.000170 t:15.8s
+tttg: c234/319 lr:0.000166 t:15.8s
+tttg: c235/319 lr:0.000163 t:15.9s
+tttg: c236/319 lr:0.000159 t:16.0s
+tttg: c237/319 lr:0.000155 t:16.0s
+tttg: c238/319 lr:0.000152 t:16.1s
+tttg: c239/319 lr:0.000148 t:16.2s
+tttg: c240/319 lr:0.000145 t:16.2s
+tttg: c241/319 lr:0.000141 t:16.3s
+tttg: c242/319 lr:0.000138 t:16.4s
+tttg: c243/319 lr:0.000134 t:16.4s
+tttg: c244/319 lr:0.000131 t:16.5s
+tttg: c245/319 lr:0.000128 t:16.6s
+tttg: c246/319 lr:0.000124 t:16.6s
+tttg: c247/319 lr:0.000121 t:16.7s
+tttg: c248/319 lr:0.000118 t:16.8s
+tttg: c249/319 lr:0.000115 t:16.8s
+tttg: c250/319 lr:0.000112 t:16.9s
+tttg: c251/319 lr:0.000109 t:17.0s
+tttg: c252/319 lr:0.000106 t:17.1s
+tttg: c253/319 lr:0.000103 t:17.1s
+tttg: c254/319 lr:0.000100 t:17.2s
+tttg: c255/319 lr:0.000097 t:17.3s
+tttg: c256/319 lr:0.000094 t:17.3s
+tttg: c257/319 lr:0.000091 t:17.4s
+tttg: c258/319 lr:0.000088 t:17.5s
+tttg: c259/319 lr:0.000085 t:17.5s
+tttg: c260/319 lr:0.000083 t:17.6s
+tttg: c261/319 lr:0.000080 t:17.7s
+tttg: c262/319 lr:0.000077 t:17.7s
+tttg: c263/319 lr:0.000075 t:17.8s
+tttg: c264/319 lr:0.000072 t:17.9s
+tttg: c265/319 lr:0.000069 t:17.9s
+tttg: c266/319 lr:0.000067 t:18.0s
+tttg: c267/319 lr:0.000065 t:18.1s
+tttg: c268/319 lr:0.000062 t:18.1s
+tttg: c269/319 lr:0.000060 t:18.2s
+tttg: c270/319 lr:0.000057 t:18.3s
+tttg: c271/319 lr:0.000055 t:18.3s
+tttg: c272/319 lr:0.000053 t:18.4s
+tttg: c273/319 lr:0.000051 t:18.5s
+tttg: c274/319 lr:0.000049 t:18.5s
+tttg: c275/319 lr:0.000046 t:18.6s
+tttg: c276/319 lr:0.000044 t:18.7s
+tttg: c277/319 lr:0.000042 t:18.7s
+tttg: c278/319 lr:0.000040 t:18.8s
+tttg: c279/319 lr:0.000039 t:18.9s
+tttg: c280/319 lr:0.000037 t:18.9s
+tttg: c281/319 lr:0.000035 t:19.0s
+tttg: c282/319 lr:0.000033 t:19.1s
+tttg: c283/319 lr:0.000031 t:19.1s
+tttg: c284/319 lr:0.000030 t:19.2s
+tttg: c285/319 lr:0.000028 t:19.3s
+tttg: c286/319 lr:0.000026 t:19.4s
+tttg: c287/319 lr:0.000025 t:19.4s
+tttg: c288/319 lr:0.000023 t:19.5s
+tttg: c289/319 lr:0.000022 t:19.6s
+tttg: c290/319 lr:0.000020 t:19.6s
+tttg: c291/319 lr:0.000019 t:19.7s
+tttg: c292/319 lr:0.000018 t:19.8s
+tttg: c293/319 lr:0.000016 t:19.8s
+tttg: c294/319 lr:0.000015 t:19.9s
+tttg: c295/319 lr:0.000014 t:20.0s
+tttg: c296/319 lr:0.000013 t:20.0s
+tttg: c297/319 lr:0.000012 t:20.1s
+tttg: c298/319 lr:0.000011 t:20.2s
+tttg: c299/319 lr:0.000010 t:20.2s
+tttg: c300/319 lr:0.000009 t:20.3s
+tttg: c301/319 lr:0.000008 t:20.4s
+tttg: c302/319 lr:0.000007 t:20.4s
+tttg: c303/319 lr:0.000006 t:20.5s
+tttg: c304/319 lr:0.000005 t:20.6s
+tttg: c305/319 lr:0.000005 t:20.6s
+tttg: c306/319 lr:0.000004 t:20.7s
+tttg: c307/319 lr:0.000004 t:20.8s
+tttg: c308/319 lr:0.000003 t:20.8s
+tttg: c309/319 lr:0.000002 t:20.9s
+tttg: c310/319 lr:0.000002 t:21.0s
+tttg: c311/319 lr:0.000002 t:21.0s
+tttg: c312/319 lr:0.000001 t:21.1s
+tttg: c313/319 lr:0.000001 t:21.2s
+tttg: c314/319 lr:0.000001 t:21.2s
+tttg: c315/319 lr:0.000000 t:21.3s
+tttg: c316/319 lr:0.000000 t:21.4s
+tttg: c317/319 lr:0.000000 t:21.4s
+tttg: c318/319 lr:0.000000 t:21.5s
+ttpr: phase:3/3 t:295.9s
+ttp: b2993/3125 bl:2.4095 bb:1.0369 rl:2.2722 rb:1.0686 dl:3039-3050 gd:1
+ttp: b2979/3125 bl:2.1303 bb:0.9836 rl:2.2683 rb:1.0663 dl:2861-2872 gd:1
+ttp: b2975/3125 bl:2.4461 bb:1.1355 rl:2.2729 rb:1.0681 dl:2819-2830 gd:1
+ttp: b2968/3125 bl:2.3311 bb:1.0632 rl:2.2744 rb:1.0680 dl:2753-2762 gd:1
+ttp: b2957/3125 bl:2.3796 bb:1.1083 rl:2.2768 rb:1.0689 dl:2653-2658 gd:1
+ttp: b2950/3125 bl:2.3084 bb:1.0334 rl:2.2775 rb:1.0681 dl:2592-2603 gd:1
+ttp: b2939/3125 bl:2.3816 bb:1.1057 rl:2.2797 rb:1.0689 dl:2508-2517 gd:1
+ttp: b2930/3125 bl:2.3217 bb:1.0100 rl:2.2806 rb:1.0676 dl:2451-2455 gd:1
+ttp: b2922/3125 bl:2.4410 bb:1.0938 rl:2.2836 rb:1.0681 dl:2387-2397 gd:1
+ttp: b2914/3125 bl:2.1440 bb:1.0800 rl:2.2811 rb:1.0684 dl:2335-2341 gd:1
+ttp: b2908/3125 bl:2.3732 bb:1.0859 rl:2.2827 rb:1.0687 dl:2301-2305 gd:1
+ttp: b2900/3125 bl:2.2847 bb:1.0122 rl:2.2827 rb:1.0676 dl:2249-2254 gd:1
+ttp: b2892/3125 bl:2.5626 bb:1.1147 rl:2.2874 rb:1.0685 dl:2199-2203 gd:1
+ttp: b2886/3125 bl:2.3466 bb:1.0094 rl:2.2883 rb:1.0675 dl:2167-2173 gd:1
+ttp: b2878/3125 bl:2.3613 bb:1.0282 rl:2.2894 rb:1.0668 dl:2129-2134 gd:1
+ttp: b2870/3125 bl:2.3257 bb:1.0658 rl:2.2900 rb:1.0668 dl:2093-2097 gd:1
+ttp: b2860/3125 bl:2.3270 bb:1.0445 rl:2.2905 rb:1.0665 dl:2049-2053 gd:1
+ttp: b2852/3125 bl:2.2276 bb:0.9661 rl:2.2896 rb:1.0650 dl:2013-2017 gd:1
+ttp: b2848/3125 bl:2.4369 bb:1.0523 rl:2.2916 rb:1.0648 dl:1997-2002 gd:1
+ttp: b2840/3125 bl:2.1437 bb:1.0136 rl:2.2897 rb:1.0641 dl:1962-1966 gd:1
+ttp: b2833/3125 bl:2.3131 bb:1.0195 rl:2.2900 rb:1.0635 dl:1937-1940 gd:1
+ttp: b2825/3125 bl:2.4228 bb:1.0404 rl:2.2917 rb:1.0632 dl:1910-1914 gd:1
+ttp: b2816/3125 bl:2.3147 bb:1.0610 rl:2.2919 rb:1.0632 dl:1883-1885 gd:1
+ttp: b2807/3125 bl:2.6276 bb:1.1352 rl:2.2959 rb:1.0641 dl:1853-1855 gd:1
+ttp: b2799/3125 bl:2.2373 bb:0.9847 rl:2.2953 rb:1.0631 dl:1830-1832 gd:1
+ttp: b2790/3125 bl:2.1337 bb:1.0164 rl:2.2934 rb:1.0626 dl:1805-1808 gd:1
+ttp: b2783/3125 bl:2.4102 bb:1.0673 rl:2.2947 rb:1.0626 dl:1785-1788 gd:1
+ttp: b2773/3125 bl:2.1972 bb:1.0034 rl:2.2937 rb:1.0620 dl:1758-1760 gd:1
+ttp: b2765/3125 bl:2.1942 bb:1.0238 rl:2.2926 rb:1.0616 dl:1737-1739 gd:1
+ttp: b2757/3125 bl:2.0867 bb:0.9656 rl:2.2905 rb:1.0606 dl:1715-1718 gd:1
+ttp: b2749/3125 bl:2.3338 bb:1.0077 rl:2.2909 rb:1.0600 dl:1696-1698 gd:1
+ttp: b2741/3125 bl:2.4190 bb:1.0730 rl:2.2922 rb:1.0602 dl:1675-1677 gd:1
+ttp: b2734/3125 bl:2.4387 bb:1.0544 rl:2.2936 rb:1.0601 dl:1659-1660 gd:1
+ttp: b2726/3125 bl:2.2493 bb:1.0720 rl:2.2932 rb:1.0602 dl:1640-1642 gd:1
+ttp: b2718/3125 bl:2.2343 bb:0.9933 rl:2.2926 rb:1.0596 dl:1620-1623 gd:1
+ttp: b2710/3125 bl:2.3724 bb:1.0470 rl:2.2934 rb:1.0594 dl:1604-1606 gd:1
+ttp: b2701/3125 bl:2.2334 bb:0.9816 rl:2.2928 rb:1.0587 dl:1586-1588 gd:1
+ttp: b2693/3125 bl:2.3723 bb:1.0822 rl:2.2935 rb:1.0589 dl:1571-1572 gd:1
+ttp: b2685/3125 bl:2.3205 bb:1.0645 rl:2.2938 rb:1.0590 dl:1553-1554 gd:1
+ttp: b2677/3125 bl:2.2614 bb:1.0861 rl:2.2935 rb:1.0592 dl:1537-1539 gd:1
+ttp: b2669/3125 bl:2.2983 bb:1.0622 rl:2.2935 rb:1.0592 dl:1521-1523 gd:1
+ttp: b2661/3125 bl:2.3552 bb:1.0140 rl:2.2940 rb:1.0588 dl:1507-1508 gd:1
+ttp: b2653/3125 bl:2.3240 bb:1.0158 rl:2.2943 rb:1.0585 dl:1493-1495 gd:1
+ttp: b2646/3125 bl:2.2843 bb:1.0298 rl:2.2942 rb:1.0582 dl:1481-1483 gd:1
+ttp: b2638/3125 bl:2.3852 bb:1.0936 rl:2.2949 rb:1.0585 dl:1467-1469 gd:1
+ttp: b2630/3125 bl:2.2383 bb:0.9897 rl:2.2945 rb:1.0580 dl:1453-1455 gd:1
+ttp: b2614/3125 bl:2.3241 bb:1.0231 rl:2.2947 rb:1.0577 dl:1427-1428 gd:1
+ttp: b2606/3125 bl:2.2081 bb:1.0243 rl:2.2941 rb:1.0574 dl:1413-1415 gd:1
+ttp: b2600/3125 bl:2.2374 bb:1.0690 rl:2.2936 rb:1.0575 dl:1404-1406 gd:1
+ttp: b2593/3125 bl:2.3805 bb:1.0703 rl:2.2943 rb:1.0576 dl:1392-1393 gd:1
+ttp: b2586/3125 bl:2.1108 bb:1.0238 rl:2.2930 rb:1.0574 dl:1383-1384 gd:1
+ttp: b2578/3125 bl:2.3545 bb:1.0447 rl:2.2934 rb:1.0573 dl:1368-1370 gd:1
+ttp: b2570/3125 bl:2.5391 bb:1.0772 rl:2.2951 rb:1.0575 dl:1357-1359 gd:1
+ttp: b2563/3125 bl:2.4436 bb:1.0597 rl:2.2960 rb:1.0575 dl:1346-1348 gd:1
+ttp: b2555/3125 bl:2.3849 bb:1.1026 rl:2.2966 rb:1.0578 dl:1334-1336 gd:1
+ttp: b2549/3125 bl:2.2842 bb:1.0532 rl:2.2966 rb:1.0577 dl:1325-1327 gd:1
+ttp: b2541/3125 bl:2.4330 bb:1.0507 rl:2.2974 rb:1.0577 dl:1314-1316 gd:1
+ttp: b2533/3125 bl:2.2755 bb:1.0583 rl:2.2973 rb:1.0577 dl:1302-1304 gd:1
+ttp: b2525/3125 bl:2.3341 bb:1.0277 rl:2.2975 rb:1.0575 dl:1290-1292 gd:1
+ttp: b2517/3125 bl:2.3720 bb:1.1129 rl:2.2980 rb:1.0578 dl:1280-1281 gd:1
+ttp: b2510/3125 bl:2.2767 bb:1.0077 rl:2.2978 rb:1.0575 dl:1272-1273 gd:1
+ttp: b2503/3125 bl:2.3122 bb:0.9982 rl:2.2979 rb:1.0571 dl:1262-1264 gd:1
+ttp: b2496/3125 bl:2.3182 bb:1.0436 rl:2.2980 rb:1.0571 dl:1253-1255 gd:1
+ttp: b2488/3125 bl:2.2886 bb:1.0249 rl:2.2980 rb:1.0569 dl:1242-1243 gd:1
+ttp: b2480/3125 bl:2.4044 bb:1.0680 rl:2.2986 rb:1.0569 dl:1230-1231 gd:1
+ttp: b2472/3125 bl:2.4936 bb:1.1049 rl:2.2997 rb:1.0572 dl:1220-1221 gd:1
+ttp: b2464/3125 bl:2.2946 bb:0.9850 rl:2.2997 rb:1.0568 dl:1209-1211 gd:1
+ttp: b2456/3125 bl:2.2801 bb:1.0518 rl:2.2995 rb:1.0568 dl:1198-1200 gd:1
+ttp: b2448/3125 bl:2.1750 bb:1.0078 rl:2.2989 rb:1.0565 dl:1189-1190 gd:1
+ttp: b2440/3125 bl:2.2663 bb:1.0304 rl:2.2987 rb:1.0564 dl:1181-1182 gd:1
+ttp: b2432/3125 bl:2.4078 bb:1.0919 rl:2.2993 rb:1.0565 dl:1171-1172 gd:1
+ttp: b2424/3125 bl:2.3158 bb:1.0320 rl:2.2994 rb:1.0564 dl:1162-1164 gd:1
+ttp: b2416/3125 bl:2.2932 bb:1.0082 rl:2.2993 rb:1.0562 dl:1153-1154 gd:1
+ttp: b2408/3125 bl:2.4202 bb:1.0108 rl:2.2999 rb:1.0559 dl:1145-1146 gd:1
+ttp: b2400/3125 bl:2.3466 bb:1.0546 rl:2.3002 rb:1.0559 dl:1136-1137 gd:1
+ttp: b2392/3125 bl:2.3697 bb:1.0825 rl:2.3005 rb:1.0560 dl:1128-1129 gd:1
+ttp: b2384/3125 bl:2.4054 bb:1.0505 rl:2.3010 rb:1.0560 dl:1118-1119 gd:1
+ttp: b2376/3125 bl:2.2865 bb:1.0446 rl:2.3010 rb:1.0559 dl:1109-1110 gd:1
+ttp: b2368/3125 bl:2.1948 bb:0.9894 rl:2.3004 rb:1.0556 dl:1101-1103 gd:1
+ttp: b2360/3125 bl:2.3182 bb:1.0455 rl:2.3005 rb:1.0556 dl:1093-1093 gd:1
+ttp: b2352/3125 bl:2.4180 bb:1.0205 rl:2.3011 rb:1.0554 dl:1084-1086 gd:1
+ttp: b2344/3125 bl:2.2934 bb:1.0346 rl:2.3010 rb:1.0553 dl:1075-1076 gd:1
+ttp: b2337/3125 bl:2.2176 bb:0.9923 rl:2.3007 rb:1.0550 dl:1069-1069 gd:1
+ttp: b2328/3125 bl:2.3358 bb:1.0693 rl:2.3008 rb:1.0551 dl:1059-1060 gd:1
+ttp: b2320/3125 bl:2.3524 bb:1.0733 rl:2.3010 rb:1.0552 dl:1051-1052 gd:1
+ttp: b2312/3125 bl:2.5177 bb:1.0192 rl:2.3020 rb:1.0550 dl:1043-1044 gd:1
+ttp: b2304/3125 bl:2.5108 bb:1.1245 rl:2.3029 rb:1.0553 dl:1036-1037 gd:1
+ttp: b2296/3125 bl:2.4701 bb:1.0724 rl:2.3036 rb:1.0554 dl:1028-1029 gd:1
+ttp: b2288/3125 bl:2.2129 bb:0.9902 rl:2.3032 rb:1.0551 dl:1021-1021 gd:1
+ttp: b2280/3125 bl:2.5569 bb:1.1056 rl:2.3043 rb:1.0553 dl:1012-1014 gd:1
+ttp: b2272/3125 bl:2.3104 bb:1.1314 rl:2.3043 rb:1.0556 dl:1006-1007 gd:1
+ttp: b2265/3125 bl:2.3830 bb:1.0625 rl:2.3046 rb:1.0556 dl:1001-1001 gd:1
+ttp: b2256/3125 bl:2.3763 bb:1.0446 rl:2.3049 rb:1.0556 dl:992-993 gd:1
+ttp: b2249/3125 bl:2.2812 bb:1.0130 rl:2.3048 rb:1.0554 dl:987-987 gd:1
+ttp: b2240/3125 bl:2.1884 bb:1.0188 rl:2.3043 rb:1.0553 dl:978-979 gd:1
+ttp: b2232/3125 bl:2.3512 bb:1.0139 rl:2.3045 rb:1.0551 dl:971-972 gd:1
+ttp: b2225/3125 bl:2.6684 bb:1.2046 rl:2.3059 rb:1.0557 dl:965-965 gd:1
+ttp: b2217/3125 bl:2.3500 bb:1.0341 rl:2.3061 rb:1.0556 dl:959-959 gd:1
+ttp: b2208/3125 bl:2.2807 bb:1.0018 rl:2.3060 rb:1.0554 dl:951-952 gd:1
+ttp: b2200/3125 bl:2.2928 bb:1.0242 rl:2.3059 rb:1.0553 dl:946-946 gd:1
+ttp: b2192/3125 bl:2.2859 bb:1.0782 rl:2.3059 rb:1.0553 dl:938-939 gd:1
+ttp: b2185/3125 bl:2.2876 bb:1.0183 rl:2.3058 rb:1.0552 dl:934-934 gd:1
+ttp: b2175/3125 bl:2.4160 bb:1.0921 rl:2.3062 rb:1.0553 dl:926-927 gd:1
+ttp: b2167/3125 bl:2.3687 bb:1.0610 rl:2.3064 rb:1.0554 dl:920-921 gd:1
+ttp: b2160/3125 bl:2.4208 bb:1.0885 rl:2.3068 rb:1.0555 dl:914-915 gd:1
+ttp: b2153/3125 bl:2.3532 bb:1.0233 rl:2.3070 rb:1.0554 dl:909-909 gd:1
+ttp: b2143/3125 bl:2.3544 bb:1.0950 rl:2.3072 rb:1.0555 dl:901-902 gd:1
+ttp: b2121/3125 bl:2.1697 bb:0.9709 rl:2.3067 rb:1.0552 dl:884-885 gd:1
+ttp: b2113/3125 bl:2.3277 bb:1.0436 rl:2.3068 rb:1.0552 dl:878-879 gd:1
+ttp: b2105/3125 bl:2.2620 bb:1.0482 rl:2.3066 rb:1.0551 dl:872-873 gd:1
+ttp: b2097/3125 bl:2.3941 bb:1.0447 rl:2.3069 rb:1.0551 dl:866-867 gd:1
+ttp: b2091/3125 bl:2.3798 bb:1.0417 rl:2.3071 rb:1.0551 dl:861-862 gd:1
+ttp: b2083/3125 bl:2.3991 bb:1.0833 rl:2.3074 rb:1.0552 dl:856-857 gd:1
+ttp: b2074/3125 bl:2.2589 bb:1.0724 rl:2.3073 rb:1.0552 dl:850-851 gd:1
+ttp: b2070/3125 bl:2.1366 bb:1.0067 rl:2.3067 rb:1.0551 dl:847-847 gd:1
+ttp: b2062/3125 bl:2.2129 bb:1.0398 rl:2.3064 rb:1.0550 dl:841-842 gd:1
+ttp: b2055/3125 bl:2.3306 bb:1.0977 rl:2.3065 rb:1.0551 dl:837-837 gd:1
+ttp: b2047/3125 bl:2.2348 bb:1.0381 rl:2.3063 rb:1.0551 dl:831-831 gd:1
+ttp: b2039/3125 bl:2.4475 bb:1.0853 rl:2.3067 rb:1.0552 dl:824-825 gd:1
+ttp: b2030/3125 bl:2.2935 bb:1.0366 rl:2.3067 rb:1.0551 dl:818-818 gd:1
+ttp: b2022/3125 bl:2.4123 bb:0.9805 rl:2.3070 rb:1.0549 dl:812-813 gd:1
+ttp: b2016/3125 bl:2.2172 bb:1.0362 rl:2.3067 rb:1.0548 dl:808-809 gd:1
+ttp: b2011/3125 bl:2.4056 bb:1.0938 rl:2.3070 rb:1.0549 dl:805-806 gd:1
+ttp: b2007/3125 bl:2.3268 bb:1.0265 rl:2.3071 rb:1.0548 dl:803-803 gd:1
+ttp: b1997/3125 bl:2.3530 bb:1.0901 rl:2.3072 rb:1.0549 dl:796-797 gd:1
+ttp: b1990/3125 bl:2.3859 bb:1.1248 rl:2.3074 rb:1.0551 dl:791-792 gd:1
+ttp: b1982/3125 bl:2.4041 bb:1.0536 rl:2.3077 rb:1.0551 dl:786-787 gd:1
+ttp: b1973/3125 bl:2.2802 bb:1.0685 rl:2.3076 rb:1.0552 dl:780-781 gd:1
+ttp: b1967/3125 bl:2.3247 bb:1.0736 rl:2.3077 rb:1.0552 dl:777-777 gd:1
+ttp: b1957/3125 bl:2.4104 bb:1.0778 rl:2.3080 rb:1.0553 dl:771-772 gd:1
+ttp: b1950/3125 bl:2.3221 bb:1.0454 rl:2.3080 rb:1.0553 dl:767-768 gd:1
+ttp: b1945/3125 bl:2.4809 bb:1.1007 rl:2.3085 rb:1.0554 dl:764-764 gd:1
+ttp: b1936/3125 bl:2.4945 bb:1.0450 rl:2.3090 rb:1.0554 dl:758-759 gd:1
+ttp: b1930/3125 bl:2.3936 bb:1.0718 rl:2.3092 rb:1.0554 dl:755-755 gd:1
+ttp: b1922/3125 bl:2.4875 bb:1.1270 rl:2.3097 rb:1.0556 dl:750-750 gd:1
+ttp: b1913/3125 bl:2.3079 bb:1.0605 rl:2.3097 rb:1.0556 dl:744-745 gd:1
+ttp: b1906/3125 bl:2.3378 bb:1.0499 rl:2.3097 rb:1.0556 dl:741-741 gd:1
+ttp: b1898/3125 bl:2.4203 bb:1.1112 rl:2.3100 rb:1.0557 dl:736-736 gd:1
+ttp: b1890/3125 bl:2.2887 bb:1.0132 rl:2.3100 rb:1.0556 dl:731-731 gd:1
+ttp: b1880/3125 bl:2.6001 bb:1.1412 rl:2.3107 rb:1.0558 dl:725-726 gd:1
+ttp: b1874/3125 bl:2.2177 bb:0.9756 rl:2.3105 rb:1.0556 dl:722-722 gd:1
+ttp: b1864/3125 bl:2.5162 bb:1.0475 rl:2.3110 rb:1.0556 dl:716-717 gd:1
+ttp: b1858/3125 bl:2.3943 bb:1.0738 rl:2.3112 rb:1.0557 dl:713-713 gd:1
+ttp: b1851/3125 bl:2.3492 bb:1.0256 rl:2.3113 rb:1.0556 dl:709-709 gd:1
+ttp: b1843/3125 bl:2.3044 bb:1.0112 rl:2.3113 rb:1.0555 dl:705-705 gd:1
+ttp: b1832/3125 bl:2.2885 bb:1.0985 rl:2.3112 rb:1.0556 dl:699-700 gd:1
+ttp: b1826/3125 bl:2.2773 bb:1.0535 rl:2.3111 rb:1.0556 dl:696-696 gd:1
+ttp: b1817/3125 bl:2.3090 bb:1.0496 rl:2.3111 rb:1.0556 dl:691-692 gd:1
+ttp: b1809/3125 bl:2.2725 bb:1.0276 rl:2.3110 rb:1.0555 dl:687-688 gd:1
+ttp: b1802/3125 bl:2.3156 bb:1.1201 rl:2.3110 rb:1.0556 dl:683-683 gd:1
+ttp: b1794/3125 bl:2.4483 bb:1.0818 rl:2.3114 rb:1.0557 dl:679-679 gd:1
+ttp: b1786/3125 bl:2.4283 bb:1.0907 rl:2.3116 rb:1.0558 dl:675-675 gd:1
+ttp: b1776/3125 bl:2.3377 bb:1.0995 rl:2.3117 rb:1.0559 dl:669-670 gd:1
+ttp: b1768/3125 bl:2.2911 bb:1.0896 rl:2.3116 rb:1.0559 dl:665-666 gd:1
+ttp: b1763/3125 bl:2.4333 bb:1.0855 rl:2.3119 rb:1.0560 dl:663-663 gd:1
+ttp: b1754/3125 bl:2.2858 bb:0.9974 rl:2.3119 rb:1.0559 dl:657-658 gd:1
+ttp: b1745/3125 bl:2.3406 bb:0.9993 rl:2.3119 rb:1.0557 dl:653-654 gd:1
+ttp: b1739/3125 bl:2.3049 bb:1.0686 rl:2.3119 rb:1.0558 dl:650-650 gd:1
+ttp: b1730/3125 bl:2.1638 bb:1.0722 rl:2.3116 rb:1.0558 dl:646-646 gd:1
+ttp: b1724/3125 bl:2.3399 bb:1.0104 rl:2.3116 rb:1.0557 dl:643-643 gd:1
+ttp: b1713/3125 bl:2.4344 bb:1.1515 rl:2.3119 rb:1.0559 dl:638-639 gd:1
+ttp: b1705/3125 bl:2.3202 bb:1.0826 rl:2.3119 rb:1.0559 dl:634-635 gd:1
+ttp: b1697/3125 bl:2.2672 bb:0.9986 rl:2.3118 rb:1.0558 dl:630-631 gd:1
+ttp: b1689/3125 bl:2.3609 bb:1.0493 rl:2.3119 rb:1.0558 dl:626-627 gd:1
+ttp: b1681/3125 bl:2.1588 bb:0.9922 rl:2.3116 rb:1.0557 dl:622-623 gd:1
+ttp: b1677/3125 bl:2.4592 bb:1.0799 rl:2.3119 rb:1.0557 dl:620-620 gd:1
+ttp: b1666/3125 bl:2.3145 bb:1.0684 rl:2.3119 rb:1.0558 dl:615-616 gd:1
+ttp: b1657/3125 bl:2.2359 bb:1.0654 rl:2.3118 rb:1.0558 dl:611-612 gd:1
+ttp: b1652/3125 bl:2.4562 bb:1.0859 rl:2.3121 rb:1.0558 dl:609-609 gd:1
+ttp: b1644/3125 bl:2.3590 bb:1.0574 rl:2.3122 rb:1.0558 dl:605-605 gd:1
+ttp: b1634/3125 bl:2.2829 bb:1.0142 rl:2.3121 rb:1.0558 dl:599-600 gd:1
+ttp: b1625/3125 bl:2.0938 bb:0.9999 rl:2.3117 rb:1.0557 dl:595-596 gd:1
+ttp: b1620/3125 bl:2.3943 bb:1.0743 rl:2.3118 rb:1.0557 dl:593-593 gd:1
+ttp: b1611/3125 bl:2.2607 bb:1.0467 rl:2.3117 rb:1.0557 dl:588-589 gd:1
+ttp: b1603/3125 bl:2.1899 bb:1.0242 rl:2.3115 rb:1.0556 dl:585-585 gd:1
+ttp: b1595/3125 bl:2.3644 bb:1.0758 rl:2.3116 rb:1.0557 dl:582-582 gd:1
+ttp: b1590/3125 bl:2.2349 bb:1.0220 rl:2.3115 rb:1.0556 dl:579-579 gd:1
+ttp: b1579/3125 bl:2.2167 bb:1.0171 rl:2.3113 rb:1.0555 dl:574-575 gd:1
+ttp: b1571/3125 bl:2.5190 bb:1.1814 rl:2.3117 rb:1.0557 dl:570-571 gd:1
+ttp: b1564/3125 bl:2.2763 bb:1.0853 rl:2.3116 rb:1.0558 dl:568-568 gd:1
+ttp: b1555/3125 bl:2.2926 bb:1.0866 rl:2.3116 rb:1.0558 dl:563-564 gd:1
+ttp: b1550/3125 bl:2.1962 bb:1.0008 rl:2.3114 rb:1.0558 dl:561-561 gd:1
+ttp: b1540/3125 bl:2.5194 bb:1.1030 rl:2.3117 rb:1.0558 dl:557-557 gd:1
+ttp: b1534/3125 bl:2.3329 bb:1.0510 rl:2.3118 rb:1.0558 dl:554-554 gd:1
+ttp: b1524/3125 bl:2.4004 bb:1.0831 rl:2.3119 rb:1.0559 dl:550-550 gd:1
+ttp: b1516/3125 bl:2.4670 bb:1.1177 rl:2.3122 rb:1.0560 dl:547-547 gd:1
+ttp: b1508/3125 bl:2.1623 bb:1.0392 rl:2.3119 rb:1.0560 dl:543-544 gd:1
+ttp: b1499/3125 bl:2.3899 bb:1.0396 rl:2.3121 rb:1.0559 dl:539-540 gd:1
+ttp: b1492/3125 bl:2.5078 bb:1.1097 rl:2.3124 rb:1.0560 dl:536-537 gd:1
+ttp: b1485/3125 bl:2.3928 bb:1.0445 rl:2.3125 rb:1.0560 dl:533-534 gd:1
+ttp: b1475/3125 bl:2.3222 bb:1.0512 rl:2.3125 rb:1.0560 dl:529-530 gd:1
+ttp: b1469/3125 bl:2.5409 bb:1.1604 rl:2.3129 rb:1.0562 dl:527-527 gd:1
+ttp: b1460/3125 bl:2.3016 bb:1.0365 rl:2.3129 rb:1.0561 dl:523-524 gd:1
+ttp: b1453/3125 bl:2.3122 bb:1.0707 rl:2.3129 rb:1.0562 dl:521-521 gd:1
+ttp: b1445/3125 bl:2.5041 bb:1.1156 rl:2.3132 rb:1.0563 dl:517-517 gd:1
+ttp: b1438/3125 bl:2.4534 bb:1.1033 rl:2.3134 rb:1.0563 dl:514-514 gd:1
+ttp: b1428/3125 bl:2.2541 bb:1.0351 rl:2.3133 rb:1.0563 dl:509-510 gd:1
+ttp: b1421/3125 bl:2.3376 bb:1.0145 rl:2.3134 rb:1.0562 dl:506-507 gd:1
+ttp: b1412/3125 bl:2.2543 bb:1.0594 rl:2.3133 rb:1.0562 dl:502-503 gd:1
+ttp: b1405/3125 bl:2.4636 bb:1.1594 rl:2.3135 rb:1.0564 dl:499-500 gd:1
+ttp: b1397/3125 bl:2.3130 bb:1.0702 rl:2.3135 rb:1.0564 dl:497-497 gd:1
+ttp: b1389/3125 bl:2.3749 bb:1.1154 rl:2.3136 rb:1.0565 dl:494-494 gd:1
+ttp: b1381/3125 bl:2.2571 bb:1.0968 rl:2.3135 rb:1.0566 dl:491-491 gd:1
+ttp: b1373/3125 bl:2.4389 bb:1.0464 rl:2.3137 rb:1.0565 dl:488-488 gd:1
+ttp: b1366/3125 bl:2.4084 bb:1.1444 rl:2.3139 rb:1.0567 dl:485-485 gd:1
+ttp: b1358/3125 bl:2.4158 bb:1.0467 rl:2.3140 rb:1.0566 dl:482-482 gd:1
+ttp: b1345/3125 bl:2.4044 bb:1.0726 rl:2.3141 rb:1.0567 dl:477-478 gd:1
+ttp: b1338/3125 bl:2.5031 bb:1.0962 rl:2.3144 rb:1.0567 dl:474-475 gd:1
+ttp: b1331/3125 bl:2.4463 bb:1.0729 rl:2.3146 rb:1.0568 dl:472-472 gd:1
+ttp: b1323/3125 bl:2.4027 bb:1.1080 rl:2.3147 rb:1.0568 dl:469-469 gd:1
+ttp: b1317/3125 bl:2.0931 bb:1.0454 rl:2.3144 rb:1.0568 dl:466-466 gd:1
+ttp: b1307/3125 bl:2.3754 bb:1.1391 rl:2.3145 rb:1.0569 dl:463-463 gd:1
+ttp: b1297/3125 bl:2.2877 bb:1.0868 rl:2.3145 rb:1.0570 dl:459-460 gd:1
+ttp: b1288/3125 bl:2.4814 bb:1.1187 rl:2.3147 rb:1.0570 dl:456-457 gd:1
+ttp: b1254/3125 bl:2.4871 bb:1.1632 rl:2.3149 rb:1.0572 dl:442-443 gd:1
+ttp: b1247/3125 bl:2.2843 bb:1.0591 rl:2.3149 rb:1.0572 dl:440-440 gd:1
+ttp: b1236/3125 bl:2.4924 bb:1.1042 rl:2.3151 rb:1.0573 dl:436-437 gd:1
+ttp: b1233/3125 bl:2.4828 bb:1.1068 rl:2.3153 rb:1.0573 dl:435-435 gd:1
+ttp: b1224/3125 bl:2.4224 bb:1.0579 rl:2.3155 rb:1.0573 dl:431-432 gd:1
+ttp: b1216/3125 bl:2.2790 bb:1.0642 rl:2.3154 rb:1.0573 dl:428-429 gd:1
+ttp: b1213/3125 bl:2.2458 bb:1.0403 rl:2.3153 rb:1.0573 dl:427-427 gd:1
+ttp: b1204/3125 bl:2.3296 bb:1.1219 rl:2.3154 rb:1.0574 dl:424-424 gd:1
+ttp: b1196/3125 bl:2.3512 bb:1.0601 rl:2.3154 rb:1.0574 dl:421-421 gd:1
+ttp: b1187/3125 bl:2.4551 bb:1.1542 rl:2.3156 rb:1.0575 dl:417-418 gd:1
+ttp: b1179/3125 bl:2.2704 bb:1.0208 rl:2.3155 rb:1.0575 dl:414-415 gd:1
+ttp: b1173/3125 bl:2.1614 bb:1.0246 rl:2.3153 rb:1.0574 dl:412-412 gd:1
+ttp: b1165/3125 bl:2.2180 bb:1.0844 rl:2.3152 rb:1.0575 dl:409-409 gd:1
+ttp: b1160/3125 bl:2.5137 bb:1.0905 rl:2.3155 rb:1.0575 dl:407-407 gd:1
+ttp: b1154/3125 bl:2.2399 bb:1.0067 rl:2.3154 rb:1.0574 dl:405-405 gd:1
+ttp: b1144/3125 bl:2.3812 bb:1.1060 rl:2.3154 rb:1.0575 dl:401-402 gd:1
+ttp: b1137/3125 bl:2.2193 bb:1.0266 rl:2.3153 rb:1.0575 dl:399-400 gd:1
+ttp: b1136/3125 bl:2.3867 bb:1.1286 rl:2.3154 rb:1.0575 dl:399-399 gd:1
+ttp: b1129/3125 bl:2.4247 bb:1.1108 rl:2.3155 rb:1.0576 dl:396-396 gd:1
+ttp: b1117/3125 bl:2.3493 bb:1.0537 rl:2.3156 rb:1.0576 dl:392-393 gd:1
+ttp: b1113/3125 bl:2.4615 bb:1.1014 rl:2.3157 rb:1.0576 dl:391-391 gd:1
+ttp: b1105/3125 bl:2.2852 bb:1.1026 rl:2.3157 rb:1.0577 dl:388-388 gd:1
+ttp: b1096/3125 bl:2.4058 bb:1.1128 rl:2.3158 rb:1.0578 dl:385-385 gd:1
+ttp: b1090/3125 bl:2.4421 bb:1.0364 rl:2.3160 rb:1.0577 dl:383-383 gd:1
+ttp: b1081/3125 bl:2.5487 bb:1.1494 rl:2.3162 rb:1.0578 dl:380-380 gd:1
+ttp: b1069/3125 bl:2.3720 bb:1.0878 rl:2.3163 rb:1.0579 dl:376-377 gd:1
+ttp: b1065/3125 bl:2.4245 bb:1.1884 rl:2.3164 rb:1.0580 dl:375-375 gd:1
+ttp: b1054/3125 bl:2.3705 bb:1.0925 rl:2.3165 rb:1.0580 dl:372-372 gd:1
+ttp: b1049/3125 bl:2.4469 bb:1.1074 rl:2.3166 rb:1.0581 dl:370-370 gd:1
+ttp: b1042/3125 bl:2.3076 bb:1.1287 rl:2.3166 rb:1.0582 dl:368-368 gd:1
+ttp: b1037/3125 bl:2.4547 bb:1.1307 rl:2.3167 rb:1.0582 dl:366-366 gd:1
+ttp: b1030/3125 bl:2.3730 bb:1.0650 rl:2.3168 rb:1.0582 dl:364-364 gd:1
+ttp: b1025/3125 bl:2.5481 bb:1.0990 rl:2.3170 rb:1.0583 dl:362-362 gd:1
+ttp: b1018/3125 bl:2.4800 bb:1.0747 rl:2.3172 rb:1.0583 dl:360-360 gd:1
+ttp: b1007/3125 bl:2.4263 bb:1.0376 rl:2.3173 rb:1.0583 dl:357-357 gd:1
+ttp: b1000/3125 bl:2.3249 bb:1.0562 rl:2.3173 rb:1.0583 dl:355-355 gd:1
+ttp: b990/3125 bl:2.2779 bb:1.1742 rl:2.3173 rb:1.0584 dl:352-352 gd:1
+ttp: b985/3125 bl:2.4306 bb:1.0915 rl:2.3174 rb:1.0584 dl:350-350 gd:1
+ttp: b974/3125 bl:2.2433 bb:1.1094 rl:2.3173 rb:1.0585 dl:346-347 gd:1
+ttp: b969/3125 bl:2.4928 bb:1.1239 rl:2.3175 rb:1.0585 dl:345-345 gd:1
+ttp: b959/3125 bl:2.3143 bb:1.0760 rl:2.3175 rb:1.0586 dl:342-342 gd:1
+ttp: b954/3125 bl:2.5106 bb:1.1501 rl:2.3177 rb:1.0586 dl:340-340 gd:1
+ttp: b943/3125 bl:2.2077 bb:1.0300 rl:2.3176 rb:1.0586 dl:336-337 gd:1
+ttp: b941/3125 bl:2.3072 bb:1.0509 rl:2.3176 rb:1.0586 dl:336-336 gd:1
+ttp: b935/3125 bl:2.3188 bb:1.1037 rl:2.3176 rb:1.0586 dl:334-334 gd:1
+ttp: b926/3125 bl:2.3796 bb:1.1288 rl:2.3176 rb:1.0587 dl:332-332 gd:1
+ttp: b919/3125 bl:2.4066 bb:1.1261 rl:2.3177 rb:1.0588 dl:330-330 gd:1
+ttp: b914/3125 bl:2.4338 bb:1.0161 rl:2.3178 rb:1.0587 dl:328-328 gd:1
+ttp: b906/3125 bl:2.5041 bb:1.1338 rl:2.3180 rb:1.0588 dl:326-326 gd:1
+ttp: b900/3125 bl:2.4840 bb:1.0993 rl:2.3182 rb:1.0588 dl:324-324 gd:1
+ttp: b893/3125 bl:2.2616 bb:1.0579 rl:2.3181 rb:1.0588 dl:322-322 gd:1
+ttp: b881/3125 bl:2.4961 bb:1.1215 rl:2.3183 rb:1.0589 dl:318-319 gd:1
+ttp: b875/3125 bl:2.2605 bb:1.1188 rl:2.3182 rb:1.0589 dl:317-317 gd:1
+ttp: b869/3125 bl:2.4693 bb:1.1726 rl:2.3183 rb:1.0590 dl:315-315 gd:1
+ttp: b857/3125 bl:2.4511 bb:1.1237 rl:2.3185 rb:1.0591 dl:312-312 gd:1
+ttp: b852/3125 bl:2.2662 bb:1.0129 rl:2.3184 rb:1.0591 dl:310-310 gd:1
+ttp: b844/3125 bl:2.6153 bb:1.1815 rl:2.3187 rb:1.0592 dl:308-308 gd:1
+ttp: b837/3125 bl:2.2003 bb:1.1212 rl:2.3186 rb:1.0592 dl:306-306 gd:1
+ttp: b829/3125 bl:2.3997 bb:1.1718 rl:2.3186 rb:1.0593 dl:304-304 gd:1
+ttp: b822/3125 bl:2.3732 bb:1.1080 rl:2.3187 rb:1.0593 dl:302-302 gd:1
+ttp: b807/3125 bl:2.3151 bb:1.1475 rl:2.3187 rb:1.0594 dl:298-299 gd:1
+ttp: b799/3125 bl:2.4725 bb:1.1891 rl:2.3188 rb:1.0595 dl:296-297 gd:1
+ttp: b793/3125 bl:2.4841 bb:1.0641 rl:2.3190 rb:1.0595 dl:295-295 gd:1
+ttp: b786/3125 bl:2.4158 bb:1.1384 rl:2.3190 rb:1.0596 dl:293-293 gd:1
+ttp: b775/3125 bl:2.5411 bb:1.1739 rl:2.3192 rb:1.0597 dl:289-290 gd:1
+ttp: b769/3125 bl:2.5086 bb:1.1748 rl:2.3194 rb:1.0598 dl:288-288 gd:1
+ttp: b763/3125 bl:2.5438 bb:1.1243 rl:2.3195 rb:1.0598 dl:286-286 gd:1
+ttp: b755/3125 bl:2.4004 bb:1.1212 rl:2.3196 rb:1.0599 dl:284-284 gd:1
+ttp: b749/3125 bl:2.4878 bb:1.1067 rl:2.3197 rb:1.0599 dl:282-282 gd:1
+ttp: b741/3125 bl:2.3868 bb:1.1105 rl:2.3198 rb:1.0600 dl:280-280 gd:1
+ttp: b732/3125 bl:2.4479 bb:1.1044 rl:2.3199 rb:1.0600 dl:278-278 gd:1
+ttp: b724/3125 bl:2.3787 bb:1.1264 rl:2.3199 rb:1.0600 dl:276-276 gd:1
+ttp: b717/3125 bl:2.4430 bb:1.0742 rl:2.3200 rb:1.0600 dl:274-274 gd:1
+ttp: b708/3125 bl:2.5705 bb:1.1048 rl:2.3202 rb:1.0601 dl:272-272 gd:1
+ttp: b700/3125 bl:2.4796 bb:1.1525 rl:2.3204 rb:1.0602 dl:270-270 gd:1
+ttp: b690/3125 bl:2.4017 bb:1.1408 rl:2.3204 rb:1.0602 dl:268-268 gd:1
+ttp: b682/3125 bl:2.3767 bb:1.1575 rl:2.3205 rb:1.0603 dl:266-266 gd:1
+ttp: b675/3125 bl:2.6280 bb:1.3645 rl:2.3207 rb:1.0605 dl:264-264 gd:1
+ttp: b664/3125 bl:2.3669 bb:1.0620 rl:2.3207 rb:1.0605 dl:261-262 gd:1
+ttp: b657/3125 bl:2.5350 bb:1.1664 rl:2.3209 rb:1.0606 dl:260-260 gd:1
+ttp: b649/3125 bl:2.2540 bb:1.0921 rl:2.3208 rb:1.0606 dl:257-258 gd:1
+ttp: b642/3125 bl:2.3038 bb:1.1546 rl:2.3208 rb:1.0606 dl:256-256 gd:1
+ttp: b632/3125 bl:2.4935 bb:1.1874 rl:2.3209 rb:1.0607 dl:253-254 gd:1
+ttp: b631/3125 bl:2.4456 bb:1.2326 rl:2.3210 rb:1.0608 dl:253-253 gd:1
+ttp: b615/3125 bl:2.5336 bb:1.1801 rl:2.3212 rb:1.0609 dl:249-250 gd:1
+ttp: b614/3125 bl:2.3937 bb:1.1444 rl:2.3212 rb:1.0610 dl:249-249 gd:1
+ttp: b599/3125 bl:2.3684 bb:1.1099 rl:2.3212 rb:1.0610 dl:245-246 gd:1
+ttp: b598/3125 bl:2.2606 bb:1.0438 rl:2.3212 rb:1.0610 dl:245-245 gd:1
+ttp: b590/3125 bl:2.4318 bb:1.1361 rl:2.3213 rb:1.0610 dl:243-243 gd:1
+ttp: b580/3125 bl:2.5870 bb:1.2132 rl:2.3215 rb:1.0611 dl:241-241 gd:1
+ttp: b572/3125 bl:2.4833 bb:1.1461 rl:2.3216 rb:1.0612 dl:239-239 gd:1
+ttp: b564/3125 bl:2.5478 bb:1.1408 rl:2.3217 rb:1.0612 dl:237-237 gd:1
+ttp: b555/3125 bl:2.4745 bb:1.1886 rl:2.3218 rb:1.0613 dl:235-235 gd:1
+ttp: b545/3125 bl:2.3104 bb:1.1048 rl:2.3218 rb:1.0613 dl:233-233 gd:1
+ttp: b535/3125 bl:2.3652 bb:1.0813 rl:2.3218 rb:1.0614 dl:230-231 gd:1
+ttp: b534/3125 bl:2.4562 bb:1.1627 rl:2.3219 rb:1.0614 dl:230-230 gd:1
+ttp: b524/3125 bl:2.4283 bb:1.2215 rl:2.3220 rb:1.0615 dl:228-228 gd:1
+ttp: b515/3125 bl:2.4340 bb:1.1499 rl:2.3221 rb:1.0616 dl:226-226 gd:1
+ttp: b506/3125 bl:2.4437 bb:1.2300 rl:2.3221 rb:1.0617 dl:224-224 gd:1
+ttp: b496/3125 bl:2.2704 bb:1.1254 rl:2.3221 rb:1.0617 dl:221-222 gd:1
+ttp: b489/3125 bl:2.2313 bb:1.0980 rl:2.3220 rb:1.0617 dl:219-220 gd:1
+ttp: b488/3125 bl:2.4616 bb:1.1935 rl:2.3221 rb:1.0618 dl:219-219 gd:1
+ttp: b477/3125 bl:2.2936 bb:1.1534 rl:2.3221 rb:1.0618 dl:217-217 gd:1
+ttp: b468/3125 bl:2.4338 bb:1.1599 rl:2.3222 rb:1.0619 dl:215-215 gd:1
+ttp: b459/3125 bl:2.4009 bb:1.1277 rl:2.3222 rb:1.0619 dl:213-213 gd:1
+ttp: b449/3125 bl:2.6859 bb:1.1980 rl:2.3224 rb:1.0620 dl:210-211 gd:1
+ttp: b448/3125 bl:2.4403 bb:1.1983 rl:2.3225 rb:1.0621 dl:210-210 gd:1
+ttp: b440/3125 bl:2.3448 bb:1.1615 rl:2.3225 rb:1.0621 dl:208-208 gd:1
+ttp: b431/3125 bl:2.3825 bb:1.1699 rl:2.3225 rb:1.0622 dl:206-206 gd:1
+ttp: b420/3125 bl:2.3029 bb:1.1719 rl:2.3225 rb:1.0623 dl:204-204 gd:1
+ttp: b408/3125 bl:2.4858 bb:1.1580 rl:2.3226 rb:1.0623 dl:201-202 gd:1
+ttp: b407/3125 bl:2.7312 bb:1.2646 rl:2.3228 rb:1.0624 dl:201-201 gd:1
+ttp: b398/3125 bl:2.4153 bb:1.1154 rl:2.3229 rb:1.0624 dl:199-199 gd:1
+ttp: b388/3125 bl:2.4716 bb:1.2710 rl:2.3230 rb:1.0625 dl:197-197 gd:1
+ttp: b379/3125 bl:2.3087 bb:1.1590 rl:2.3230 rb:1.0626 dl:195-195 gd:1
+ttp: b371/3125 bl:2.5584 bb:1.1959 rl:2.3231 rb:1.0627 dl:193-193 gd:1
+ttp: b361/3125 bl:2.2436 bb:1.1478 rl:2.3231 rb:1.0627 dl:190-191 gd:1
+ttp: b360/3125 bl:2.4170 bb:1.1824 rl:2.3231 rb:1.0628 dl:190-190 gd:1
+ttp: b343/3125 bl:2.5031 bb:1.2127 rl:2.3232 rb:1.0628 dl:186-187 gd:1
+ttp: b341/3125 bl:2.3121 bb:1.0190 rl:2.3232 rb:1.0628 dl:186-186 gd:1
+ttp: b330/3125 bl:2.3950 bb:1.1347 rl:2.3232 rb:1.0628 dl:184-184 gd:1
+ttp: b328/3125 bl:2.3986 bb:1.1845 rl:2.3233 rb:1.0629 dl:183-183 gd:1
+ttp: b319/3125 bl:2.3873 bb:1.1948 rl:2.3233 rb:1.0630 dl:181-181 gd:1
+ttp: b310/3125 bl:2.3304 bb:1.1036 rl:2.3233 rb:1.0630 dl:179-179 gd:1
+ttp: b298/3125 bl:2.5118 bb:1.2225 rl:2.3234 rb:1.0630 dl:177-177 gd:1
+ttp: b290/3125 bl:2.5829 bb:1.2940 rl:2.3235 rb:1.0631 dl:175-175 gd:1
+ttp: b279/3125 bl:2.6393 bb:1.2679 rl:2.3237 rb:1.0632 dl:172-173 gd:1
+ttp: b276/3125 bl:2.5383 bb:1.2789 rl:2.3238 rb:1.0633 dl:172-172 gd:1
+ttp: b266/3125 bl:2.7143 bb:1.2548 rl:2.3239 rb:1.0634 dl:170-170 gd:1
+ttp: b264/3125 bl:2.7506 bb:1.2706 rl:2.3241 rb:1.0635 dl:169-169 gd:1
+ttp: b248/3125 bl:2.2715 bb:1.2087 rl:2.3241 rb:1.0636 dl:165-166 gd:1
+ttp: b247/3125 bl:2.4654 bb:1.1023 rl:2.3242 rb:1.0636 dl:165-165 gd:1
+ttp: b238/3125 bl:2.3431 bb:1.1275 rl:2.3242 rb:1.0636 dl:163-163 gd:1
+ttp: b230/3125 bl:2.3649 bb:1.2375 rl:2.3242 rb:1.0637 dl:161-161 gd:1
+ttp: b223/3125 bl:2.8045 bb:1.2869 rl:2.3244 rb:1.0638 dl:159-159 gd:1
+ttp: b213/3125 bl:2.4025 bb:1.2749 rl:2.3244 rb:1.0638 dl:157-157 gd:1
+ttp: b204/3125 bl:2.4758 bb:1.2530 rl:2.3245 rb:1.0639 dl:155-155 gd:1
+ttp: b197/3125 bl:2.4123 bb:1.0725 rl:2.3245 rb:1.0639 dl:153-153 gd:1
+ttp: b189/3125 bl:2.6067 bb:1.2295 rl:2.3246 rb:1.0640 dl:151-151 gd:1
+ttp: b180/3125 bl:2.3632 bb:1.1307 rl:2.3247 rb:1.0640 dl:149-149 gd:1
+ttp: b172/3125 bl:2.5600 bb:1.1944 rl:2.3248 rb:1.0641 dl:147-147 gd:1
+ttp: b165/3125 bl:2.4570 bb:1.1731 rl:2.3248 rb:1.0641 dl:145-145 gd:1
+ttp: b153/3125 bl:2.5010 bb:1.2521 rl:2.3249 rb:1.0642 dl:142-143 gd:1
+ttp: b152/3125 bl:2.4932 bb:1.1609 rl:2.3249 rb:1.0642 dl:142-142 gd:1
+ttp: b138/3125 bl:2.6833 bb:1.2475 rl:2.3251 rb:1.0643 dl:138-139 gd:1
+ttp: b130/3125 bl:2.6433 bb:1.2474 rl:2.3252 rb:1.0643 dl:136-137 gd:1
+ttp: b123/3125 bl:2.6149 bb:1.2205 rl:2.3253 rb:1.0644 dl:134-135 gd:1
+ttp: b121/3125 bl:2.4374 bb:1.1639 rl:2.3253 rb:1.0644 dl:134-134 gd:1
+ttp: b113/3125 bl:2.6461 bb:1.2014 rl:2.3254 rb:1.0645 dl:132-132 gd:1
+ttp: b106/3125 bl:2.7513 bb:1.2992 rl:2.3256 rb:1.0646 dl:130-130 gd:1
+ttp: b98/3125 bl:2.6824 bb:1.1873 rl:2.3257 rb:1.0646 dl:128-128 gd:1
+ttp: b85/3125 bl:2.5309 bb:1.2763 rl:2.3258 rb:1.0647 dl:124-125 gd:1
+ttp: b79/3125 bl:2.7067 bb:1.2348 rl:2.3259 rb:1.0647 dl:123-123 gd:1
+ttp: b69/3125 bl:2.4728 bb:1.1057 rl:2.3260 rb:1.0647 dl:119-120 gd:1
+ttp: b65/3125 bl:2.5746 bb:1.2763 rl:2.3260 rb:1.0648 dl:118-118 gd:1
+ttp: b56/3125 bl:2.5772 bb:1.1954 rl:2.3261 rb:1.0648 dl:115-115 gd:1
+ttp: b48/3125 bl:2.5648 bb:1.1249 rl:2.3262 rb:1.0649 dl:112-112 gd:1
+ttp: b38/3125 bl:2.6098 bb:1.1935 rl:2.3263 rb:1.0649 dl:107-108 gd:1
+ttp: b33/3125 bl:2.7948 bb:1.2566 rl:2.3264 rb:1.0649 dl:105-105 gd:1
+ttp: b24/3125 bl:2.6547 bb:1.1595 rl:2.3265 rb:1.0650 dl:101-101 gd:1
+ttp: b14/3125 bl:2.7536 bb:1.2636 rl:2.3266 rb:1.0650 dl:94-95 gd:1
+ttp: b7/3125 bl:2.9315 bb:1.2624 rl:2.3267 rb:1.0651 dl:86-88 gd:1
+quantized_ttt_phased val_loss:2.32484879 val_bpb:1.06236517 eval_time:482574ms
+total_eval_time:482.6s