diff --git a/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/README.md b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/README.md new file mode 100644 index 0000000000..5834152c09 --- /dev/null +++ b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/README.md @@ -0,0 +1,159 @@ +# Record Candidate: Long-Context no-QV TTT with Rank-56 / Prefix-3000 Compute Reallocation + +**val_bpb = 1.05874877** (3-seed mean, population std 0.00091680) | **max artifact 15,980,110 bytes** | **8x H100 SXM** | **strict 600s train / 600s eval logs** + +This submission is a clean, score-first Track B refinement of the late-April CaseOps / LQER / SparseAttnGate / phased-TTT stack. It keeps the long-context `no_qv` TTT setup from PR #1953 and reallocates the TTT eval budget by reducing LoRA rank to 56 while increasing the phased global-TTT prefix to 3000 documents. + +Leaderboard context at submission time: + +- vs merged PR #1855 (`1.06108`): **-0.00233123 BPB**, approximately **-0.00510 nats**. +- vs open PR #1953 (`1.05855370`): **+0.00019507 BPB** worse. This PR is submitted as a clean record candidate against the merged leaderboard and as a reproducible ablation on the #1953 lineage, not as a claim to beat #1953 if that PR is accepted first. + +## 3-Seed Results + +| Seed | Stop step | Train ms | Pre-quant BPB | Quant no-TTT BPB | Final TTT BPB | Eval ms | Artifact bytes | +|---:|---:|---:|---:|---:|---:|---:|---:| +| 42 | 4900 | 596051 | 1.06108950 | 1.06949683 | **1.05780842** | 519467 | 15,975,989 | +| 0 | 4879 | 596086 | 1.06200352 | 1.07034149 | **1.05844590** | 425873 | 15,976,674 | +| 1234 | 4880 | 596146 | 1.06319088 | 1.07176584 | **1.05999198** | 400555 | 15,980,110 | +| **Mean** | **4886** | **596094** | **1.06209463** | **1.07053472** | **1.05874877** | **448632** | **15,977,591** | + +All three seeds satisfy: + +- artifact < 16,000,000 bytes +- training wallclock < 600,000 ms +- final TTT eval wallclock < 600,000 ms +- full validation set with CaseOps byte-sidecar BPB accounting + +## What Changed + +Relative to the PR #1953 long-context `no_qv` stack: + +```bash +TTT_LORA_RANK=56 # PR #1953 used rank 80 in its reproduction command +PHASED_TTT_PREFIX_DOCS=3000 # PR #1953 used 2500 +PHASED_TTT_NUM_PHASES=3 +EVAL_SEQ_LEN=2560 +TTT_EVAL_SEQ_LEN=2560 +TTT_MASK=no_qv +TTT_Q_LORA=0 +TTT_V_LORA=0 +TTT_LOCAL_LR_MULT=0.75 +QK_GAIN_INIT=5.25 +``` + +The intent is a simple eval-budget tradeoff: smaller per-document LoRA state with a longer score-first global-TTT prefix. The result is seed-sensitive: it improves seeds 42 and 0 relative to the public #1953 numbers, but seed 1234 is weaker, so the 3-seed mean lands behind #1953 while still clearing the merged #1855 record by the 0.005-nat threshold. + +## Key Inherited Components + +This submission inherits the core stack from the public #1953 / #1945 / #1908 / #1855 lineage: + +- CaseOps lossless tokenizer with validation byte sidecar +- BOS-fixed SmearGate and SparseAttnGate +- LQER asymmetric rank-4 quantization +- AWQ-lite mixed-precision GPTQ +- asymmetric logit rescale +- fused softcapped CE +- long-context score-first phased TTT with `no_qv` LoRA mask +- per-group lrzip artifact compression + +## Compliance Against Issue #1017 Conditions + +- **C1 strict causal dependence:** every token is scored from the submitted artifact plus strict prefix state only. CaseOps is a deterministic lossless transform; the byte sidecar supplies denominator accounting and does not affect predictions. +- **C2 full normalized distribution:** the neural model emits a standard full-vocabulary distribution over the 8192 CaseOps token ids. No byte-level scorer, entropy expert, PPM mixture, n-gram cache, or target-conditioned gate is used. +- **C3 score-before-update:** phased TTT scores chunks/documents before any LoRA or global update can use those scored tokens. Updated state is used only for later tokens/documents. +- **C4 single left-to-right pass:** every validation token is scored once. No rescoring, no multi-pass selection, and no oracle best-of-k over validation outcomes. + +Additional exclusions: + +- no SLOT +- no byte-PPM or token-PPM mixer +- no pre-quant TTT on validation data +- no eval-time logit bias selected from the observed target +- no validation-token leakage into training weights before scoring + +## Reproduction + +The CaseOps dataset/tokenizer are the same public data path used by the inherited stack (`romeerp/parameter-golf-caseops-v1`): + +```bash +pip install brotli sentencepiece huggingface-hub +pip install flash_attn_3 --no-deps --find-links https://windreamer.github.io/flash-attention3-wheels/cu128_torch291/ + +# Fetch / prepare the CaseOps SP8192 dataset used by the inherited stack. +# DATA_PATH should contain fineweb_train_*.bin, fineweb_val_*.bin, and fineweb_val_bytes_*.bin. + +export DATA_PATH=./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved +export TOKENIZER_PATH=./data/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model +export CASEOPS_ENABLED=1 +export VOCAB_SIZE=8192 +export MAX_WALLCLOCK_SECONDS=600 +export VAL_LOSS_EVERY=0 +export TTT_ENABLED=1 +export PHASED_TTT_ENABLED=1 +export PHASED_TTT_NUM_PHASES=3 +export PHASED_TTT_PREFIX_DOCS=3000 +export TTT_LORA_RANK=56 +export TTT_MASK=no_qv +export TTT_Q_LORA=0 +export TTT_V_LORA=0 +export TTT_LOCAL_LR_MULT=0.75 +export TTT_BETA2=0.99 +export TTT_WEIGHT_DECAY=0.5 +export TTT_CHUNK_SIZE=48 +export EVAL_SEQ_LEN=2560 +export TTT_EVAL_SEQ_LEN=2560 +export QK_GAIN_INIT=5.25 +export MATRIX_LR=0.026 +export MIN_LR=0.1 +export WARMDOWN_FRAC=0.85 +export BETA2=0.99 +export EMBED_BITS=7 +export MATRIX_CLIP_SIGMAS=12.85 +export ATTN_CLIP_SIGMAS=13.0 +export MLP_CLIP_SIGMAS=11.5 +export EMBED_CLIP_SIGMAS=14.0 +export GRAD_CLIP_NORM=0.3 +export FUSED_CE_ENABLED=1 +export SMEAR_GATE_ENABLED=1 +export GATE_WINDOW=12 +export SPARSE_ATTN_GATE_ENABLED=1 +export SPARSE_ATTN_GATE_SCALE=0.5 +export GATED_ATTN_QUANT_GATE=1 +export LQER_ENABLED=1 +export LQER_RANK=4 +export LQER_TOP_K=3 +export LQER_GROUP_SIZE=64 +export LQER_ASYM_ENABLED=1 +export LQER_ASYM_GROUP=64 +export AWQ_LITE_ENABLED=1 +export ASYM_LOGIT_RESCALE=1 +export GPTQ_RESERVE_SECONDS=4.0 +export GPTQ_CALIBRATION_BATCHES=16 +export COMPRESSOR=pergroup +export NCCL_NET=Socket + +SEED=42 RUN_ID=rank56_prefix3000_seed42 ARTIFACT_DIR=./artifacts/seed42 torchrun --standalone --nproc_per_node=8 train_gpt.py +SEED=0 RUN_ID=rank56_prefix3000_seed0 ARTIFACT_DIR=./artifacts/seed0 torchrun --standalone --nproc_per_node=8 train_gpt.py +SEED=1234 RUN_ID=rank56_prefix3000_seed1234 ARTIFACT_DIR=./artifacts/seed1234 torchrun --standalone --nproc_per_node=8 train_gpt.py +``` + +## Included Files + +- `README.md` +- `submission.json` +- `train_gpt.py` +- `train_seed42.log` +- `train_seed0.log` +- `train_seed1234.log` + +## Credits + +This is a small compute-allocation refinement on a broad community stack. Credit belongs primarily to the authors of the inherited lineage: + +- @codemath3000 for PR #1855 +- @alertcat, @romeerp, @jorge-asenjo, and @aquariouseworkman for PR #1945 / #1908 / #1923 components +- @nprime06, @dexhunter, @samacqua, and earlier VarLen/phased-TTT contributors +- @romeerp and @dexhunter for the CaseOps tokenizer and byte-sidecar precedent + +The contribution here is the rank-56 / prefix-3000 score-first TTT ablation and the fully logged 3-seed rerun under the standard constraints. diff --git a/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/submission.json b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/submission.json new file mode 100644 index 0000000000..bf48fe3a4f --- /dev/null +++ b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/submission.json @@ -0,0 +1,89 @@ +{ + "author": "Himanshu Dongre", + "github_id": "himanshudongre", + "name": "Long-context no-QV phased TTT with rank-56 / prefix-3000 compute reallocation", + "date": "2026-04-30", + "track": "10min_16mb", + "val_bpb": 1.05874877, + "val_bpb_std": 0.00091680, + "seeds": [ + 42, + 0, + 1234 + ], + "seed_results": { + "42": { + "stop_step": 4900, + "train_ms": 596051, + "prequant_bpb": 1.06108950, + "quant_bpb": 1.06949683, + "val_bpb": 1.05780842, + "eval_ms": 519467, + "artifact_bytes": 15975989 + }, + "0": { + "stop_step": 4879, + "train_ms": 596086, + "prequant_bpb": 1.06200352, + "quant_bpb": 1.07034149, + "val_bpb": 1.05844590, + "eval_ms": 425873, + "artifact_bytes": 15976674 + }, + "1234": { + "stop_step": 4880, + "train_ms": 596146, + "prequant_bpb": 1.06319088, + "quant_bpb": 1.07176584, + "val_bpb": 1.05999198, + "eval_ms": 400555, + "artifact_bytes": 15980110 + } + }, + "hardware": "8x H100 80GB SXM", + "pytorch_version": "2.9.1+cu128", + "technique_summary": "PR #1953 long-context no-QV CaseOps/LQER/AWQ-lite stack with TTT_LORA_RANK=56 and PHASED_TTT_PREFIX_DOCS=3000", + "comparison": { + "merged_pr_1855_bpb": 1.06108, + "delta_vs_merged_pr_1855_bpb": -0.00233123, + "delta_vs_merged_pr_1855_nats_approx": -0.00510, + "open_pr_1953_bpb": 1.05855370, + "delta_vs_open_pr_1953_bpb": 0.00019507 + }, + "run_overrides": { + "CASEOPS_ENABLED": 1, + "EVAL_SEQ_LEN": 2560, + "TTT_EVAL_SEQ_LEN": 2560, + "TTT_MASK": "no_qv", + "TTT_Q_LORA": 0, + "TTT_V_LORA": 0, + "TTT_LOCAL_LR_MULT": 0.75, + "TTT_LORA_RANK": 56, + "PHASED_TTT_PREFIX_DOCS": 3000, + "PHASED_TTT_NUM_PHASES": 3, + "QK_GAIN_INIT": 5.25, + "COMPRESSOR": "pergroup" + }, + "compliance": { + "train_under_600s": true, + "artifact_under_16mb": true, + "eval_under_600s": true, + "three_seeds": true, + "full_validation_set": true, + "caseops_byte_sidecar_accounting": true, + "score_first_ttt": true, + "no_slot": true, + "no_byte_ppm": true, + "no_token_ppm": true, + "no_ngram_cache": true, + "no_eval_time_logit_bias": true, + "no_pre_quant_ttt_on_val": true, + "single_pass_eval": true + }, + "attribution": { + "base_stack": "PR #1953 / #1945 / #1908 / #1855 lineage", + "caseops": "@romeerp and @dexhunter", + "phased_ttt": "@samacqua, @dexhunter, and VarLen/phased-TTT contributors", + "rank56_prefix3000_ablation": "@himanshudongre" + } +} diff --git a/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_gpt.py b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_gpt.py new file mode 100644 index 0000000000..7a146cc2bf --- /dev/null +++ b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_gpt.py @@ -0,0 +1,4064 @@ +import base64, collections, copy, fcntl, glob, io, lzma, math, os +from pathlib import Path +import random, re, subprocess, sys, 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 softcapped cross-entropy (Triton) — training-only path ===== +# Replaces the eager +# logits_softcap = softcap * tanh(logits / softcap) +# F.cross_entropy(logits_softcap.float(), targets, reduction="mean") +# sequence with a single fused kernel that reads logits_proj once, applies +# softcap in-register, and computes (LSE, loss) in one streaming pass. The +# backward kernel mirrors the forward so there's no stored softcapped logits. +# Numerically identical to the eager path up to fp32 accumulation differences. +_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", 1337)) + 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 softcapped CE (Triton). Training-only — forward_logits eval path still uses + # eager softcap+F.cross_entropy. Default ON since validated as at-worst neutral. + 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"))) + tie_embeddings = bool(int(os.environ.get("TIE_EMBEDDINGS", "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", 0.0)) + embed_lr = float(os.environ.get("EMBED_LR", 0.6)) + 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_local_lr_mult = float(os.environ.get("TTT_LOCAL_LR_MULT", 1.0)) + 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", 64)) + ttt_grad_steps = int(os.environ.get("TTT_GRAD_STEPS", 1)) + # V19: PR #1886 (renqianluo) + sunnypatneedi research log 2026-04-28 found that + # the Triton fused-CE kernel's fp32-accumulation interacts with warm-start LoRA-A + # to destabilize seeds 314/1337 at TTT_WEIGHT_DECAY=1.0. Raising the default to + # 2.0 prevents seed collapse without measurably moving stable seeds. + ttt_weight_decay = float(os.environ.get("TTT_WEIGHT_DECAY", 2.0)) + ttt_beta1 = float(os.environ.get("TTT_BETA1", 0)) + ttt_beta2 = float(os.environ.get("TTT_BETA2", 0.999)) + ttt_mask = os.environ.get("TTT_MASK", "").strip().lower() + _ttt_q_default = "1" + _ttt_v_default = "1" + if ttt_mask in ("", "all", "baseline_all"): + pass + elif ttt_mask == "no_q": + _ttt_q_default = "0" + elif ttt_mask == "no_v": + _ttt_v_default = "0" + elif ttt_mask == "no_qv": + _ttt_q_default = "0" + _ttt_v_default = "0" + else: + raise ValueError(f"Unsupported TTT_MASK={ttt_mask!r}") + ttt_q_lora = bool(int(os.environ.get("TTT_Q_LORA", _ttt_q_default))) + ttt_k_lora = bool(int(os.environ.get("TTT_K_LORA", "1"))) + ttt_v_lora = bool(int(os.environ.get("TTT_V_LORA", _ttt_v_default))) + 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", 4.0)) + 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", 1)) + 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", 8)) + matrix_clip_sigmas = float(os.environ.get("MATRIX_CLIP_SIGMAS", 12.85)) + embed_clip_sigmas = float(os.environ.get("EMBED_CLIP_SIGMAS", 2e1)) + mlp_clip_sigmas = float(os.environ.get("MLP_CLIP_SIGMAS", 10.0)) + attn_clip_sigmas = float(os.environ.get("ATTN_CLIP_SIGMAS", 13.0)) + # AttnOutGate (per-head multiplicative output gate, PR #1667 MarioPaerle). + # Zero-init weight: 2*sigmoid(0)=1 -> transparent at start. Source defaults to + # block input x ('proj'); 'q' uses raw Q projection output. + 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") + # SmearGate (input-dependent forward-1 token smear, modded-nanogpt @classiclarryd + # via PR #1667). x_t <- x_t + lam * sigmoid(W*x_t[:gate_window]) * x_{t-1}. + # lam=0 + W=0 -> transparent at init. + smear_gate_enabled = bool(int(os.environ.get("SMEAR_GATE_ENABLED", "0"))) + # Window: first GATE_WINDOW dims of the source feed the gate projection. + gate_window = int(os.environ.get("GATE_WINDOW", 12)) + # Gated Attention (Qwen, NeurIPS 2025 Best Paper, arXiv:2505.06708; + # qiuzh20/gated_attention). Per-head sigmoid gate on SDPA output, BEFORE + # out_proj. Gate input = full block input x (paper's headwise G1 variant + # driven from hidden_states). W_g shape (num_heads, dim), plain sigmoid. + # Near-zero init gives g~0.5 at step 0 (half attention output); per-block + # attn_scale (init 1.0) compensates during training. Name contains + # "attn_gate" so CONTROL_TENSOR_NAME_PATTERNS routes it to scalar AdamW. + 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)) + # Dedicated int8-per-row quantization for `attn_gate_w` tensors. These are + # small ((num_heads, dim) = (8, 512) = 4096 params) and bypass GPTQ via the + # numel<=65536 passthrough branch -> stored as fp16 (8 KB/layer, ~65 KB total + # compressed). int8-per-row cuts the raw tensor in half with negligible BPB + # impact: scales per head (8 values), symmetric quant over [-127, 127]. + # No Hessian needed (gate weights not in collect_hessians()). + gated_attn_quant_gate = bool(int(os.environ.get("GATED_ATTN_QUANT_GATE", "0"))) + # Sparse Attention Gate (modded-nanogpt-style). Keeps dense SDPA and only + # swaps the output-gate input to the first GATE_WINDOW residual dims. + # W_g: (num_heads, gate_window) = (8, 12) = 96 params/layer (~44K total), + # vs dense GatedAttn's (8, 512) = 4K/layer (~44K diff). Name "attn_gate_w" + # is shared so quant routing and int8 gate passthrough Just Work. Gate + # passthrough int8 still applies via GATED_ATTN_QUANT_GATE=1. + # Mutually exclusive with ATTN_OUT_GATE_ENABLED and GATED_ATTN_ENABLED. + sparse_attn_gate_enabled = bool(int(os.environ.get("SPARSE_ATTN_GATE_ENABLED", "0"))) + 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 asymmetric rank-k correction on top-K quant-error tensors (PR #1530 v2 port). + # Computes SVD of E = W_fp - W_quant, packs top-r A,B as INT2/INT4 (asym) or INTk (sym). + 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", 3)) + 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")) + lqer_scope = os.environ.get("LQER_SCOPE", "all") + lqer_gain_select = bool(int(os.environ.get("LQER_GAIN_SELECT", "0"))) + awq_lite_enabled = bool(int(os.environ.get("AWQ_LITE_ENABLED", "0"))) + awq_lite_bits = int(os.environ.get("AWQ_LITE_BITS", "8")) + awq_lite_group_top_k = int(os.environ.get("AWQ_LITE_GROUP_TOP_K", "1")) + awq_lite_group_size = int(os.environ.get("AWQ_LITE_GROUP_SIZE", "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 integration: optional override of dataset root + tokenizer path. + # When CASEOPS_ENABLED=1, the wrapper loads a per-token byte sidecar + # (fineweb_val_bytes_*.bin, identical shard layout to val_*.bin) and uses + # it as the canonical raw-byte budget for BPB accounting. The sidecar + # REPLACES the build_sentencepiece_luts byte-counting path entirely. + caseops_enabled = bool(int(os.environ.get("CASEOPS_ENABLED", "0"))) + _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.path.join(artifact_dir, "final_model.pt") + if artifact_dir + else "final_model.pt" + ) + quantized_model_path = ( + os.path.join(artifact_dir, "final_model.int6.ptz") + if artifact_dir + else "final_model.int6.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.caseops_enabled = bool(getattr(h, "caseops_enabled", False)) + if self.caseops_enabled: + self.base_bytes_lut = None + self.has_leading_space_lut = None + self.is_boundary_token_lut = None + else: + ( + self.base_bytes_lut, + self.has_leading_space_lut, + self.is_boundary_token_lut, + ) = build_sentencepiece_luts(self.sp, h.vocab_size, device) + 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): + # Filter out CaseOps byte sidecar shards which share the val_*.bin glob. + 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): + """Load CaseOps per-token byte sidecar(s). Same shard layout as token shards + (256 int32 header + uint16 array). Each entry = canonical raw-text byte + budget for that token in the corresponding val shard. Returns a CPU + int16 tensor sliced to match expected_len (i.e. val_tokens length).""" + 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] + # load_data_shard returns uint16 — that's exactly what the sidecar stores. + 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(" 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._prefetch_queue = [] + + 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 = torch.empty(per_rank_span - 1, dtype=torch.int64, pin_memory=True) + targets = torch.empty(per_rank_span - 1, dtype=torch.int64, pin_memory=True) + inputs.copy_(buf[:-1]) + targets.copy_(buf[1:]) + 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) + while len(self._prefetch_queue) < 2: + self._prefetch_queue.append( + self._batch_pool.submit(self._prepare_batch, num_tokens_local, self.max_seq_len)) + inputs, targets, cu_seqlens, max_seqlen = self._prefetch_queue.pop(0).result() + self._prefetch_queue.append( + 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, 0.5 * pre0) + c1 = c1 * tl.where(pre1 > 0, 2.0 * pre1, 0.5 * 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, 0.5 * c0) + aux1 = tl.where(c1 > 0, c1, 0.5 * 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 = 256, 128, 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 + # AttnOutGate (PR #1667 MarioPaerle): per-head multiplicative gate on attention + # output. CastedLinear so restore_fp32_params casts back to fp32 for GPTQ. + # _zero_init -> 2*sigmoid(0)=1 -> transparent at init. + 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 + # Gated Attention (arXiv:2505.06708, Qwen, NeurIPS 2025). Per-head sigmoid + # gate on SDPA output, BEFORE out_proj. Gate projection W_g: (num_heads, dim). + # Name "attn_gate_w" contains "attn_gate" substring so it matches + # CONTROL_TENSOR_NAME_PATTERNS and routes to the scalar AdamW group. + # fp32 Parameter -> restore_fp32_params path covers it via the ndim<2 OR + # name-pattern check (name matches "attn_gate"). Cast to x.dtype on use. + 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) + # Sparse attention head-output gate (modded-nanogpt style). Keeps dense SDPA + # and only narrows the gate input to the first gate_window residual dims. + # W_g: (num_heads, gate_window). y_{t,h} <- sigmoid(scale * W_g_h @ x_t[:gate_window]) * y_{t,h}. + # Shares attn_gate_w name with dense GatedAttn so the quant routing + # (CONTROL_TENSOR_NAME_PATTERNS / attn_gate_w int8 passthrough) is unchanged. + 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 kept around as a tap point for attn_out_gate_src='q' (post-projection, + # pre-reshape, pre-RoPE). + 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) + # AttnOutGate inlined (PR #1667). Inline + .contiguous() barrier so torch.compile + # fullgraph=True is happy (this avoids the @torch.compiler.disable trap that + # crashed gates v3). Per-head gate on (B,T,H,D) tensor: g shape [B,T,H], broadcast + # over D via [..., None]. zero-init weight -> 2*sigmoid(0)=1 -> transparent. + 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] + # Gated Attention (arXiv:2505.06708 G1). Inline + .contiguous() barrier so + # torch.compile fullgraph=True is happy. Per-head gate on (B,T,H,D): g shape + # [B,T,H], broadcast over D via [..., None]. Paper: g = sigmoid(x @ W_g.T) + # where W_g: (H, dim). .to(x.dtype) on fp32 param before broadcast with bf16. + 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] + # Sparse head-output gate: narrower (gate_window) input, same shape g as GatedAttn. + 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=0.5).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.tie_embeddings = h.tie_embeddings + 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() + self.lm_head = ( + None + if h.tie_embeddings + else CastedLinear(h.model_dim, h.vocab_size, bias=False) + ) + if self.lm_head is not None: + self.lm_head._zero_init = True + 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) + ) + # SmearGate (PR #1667 / modded-nanogpt @classiclarryd): + # x_t <- x_t + lam * sigmoid(W * x_t[:gate_window]) * x_{t-1}. + # Per-token forward-1 smear of the embedding lane. W zero-init + lam=0 -> + # transparent at init. Uses CastedLinear so restore_fp32_params handles dtype. + 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)) + # V19: Asymmetric Logit Rescale (PR #1923 jorge-asenjo). + # Two learnable softcap scales applied on the EVAL path (forward_logits + + # forward_ttt). Init to logit_softcap so the layer is identity at step 0. + # Train path keeps the single fused softcap to preserve PR #1855 numerics. + self.asym_logit_enabled = bool(int(os.environ.get("ASYM_LOGIT_RESCALE", "0"))) + if self.asym_logit_enabled: + self.softcap_pos = nn.Parameter(torch.tensor(float(h.logit_softcap), dtype=torch.float32)) + self.softcap_neg = nn.Parameter(torch.tensor(float(h.logit_softcap), dtype=torch.float32)) + self._init_weights() + + def _init_weights(self): + if self.tie_embeddings: + 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): + """Run the encoder/decoder stack to the final RMSNorm; returns pre-projection hidden. + Shared by eval (softcap+projection via forward_logits) and train (fused CE path).""" + x = self.tok_emb(input_ids) + # SmearGate (PR #1667). lam=0 + W=0 -> identity at init. + # Cross-doc leak fix: zero the prev-token smear at any position whose current token + # is BOS, so the BOS embedding starting doc N+1 in a packed stream is not + # contaminated by doc N's last token (audited issue on PR#1797 base). + 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)) + not_bos = (input_ids[:, 1:] != BOS_ID).to(x.dtype).unsqueeze(-1) + x = torch.cat([x[:, :1], x[:, 1:] + g * x[:, :-1] * not_bos], 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): + if self.tie_embeddings: + return F.linear(hidden, self.tok_emb.weight) + return self.lm_head(hidden) + + def _apply_asym_softcap(self, logits): + # V19: Asymmetric softcap (PR #1923). Splits the logit_softcap scalar into + # learnable positive/negative branches. Score-first preserved: still a + # bounded, normalized post-projection nonlinearity feeding a standard + # softmax over the full vocab. + sp = self.softcap_pos.to(logits.dtype) + sn = self.softcap_neg.to(logits.dtype) + return torch.where(logits > 0, sp * torch.tanh(logits / sp), sn * torch.tanh(logits / sn)) + + 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) + if self.asym_logit_enabled: + return self._apply_asym_softcap(logits_proj) + 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) + # Fused softcapped-CE kernel (training path only). Applies softcap inside the + # Triton kernel; takes pre-softcap logits_proj. Non-fused path matches stock + # PR-1736 numerics exactly (softcap in fp32, then F.cross_entropy on fp32). + 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) + # SmearGate on the TTT path — same inline compute as forward_logits. + # Cross-doc leak fix: see _forward_hidden comment. + 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)) + not_bos = (input_ids[:, 1:] != BOS_ID).to(x.dtype).unsqueeze(-1) + x = torch.cat([x[:, :1], x[:, 1:] + g * x[:, :-1] * not_bos], 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) + if self.tie_embeddings: + logits = F.linear(x, self.tok_emb.weight) + else: + logits = self.lm_head(x) + logits = logits + lora.lm_head_lora(x) + # V19: same asymmetric softcap on the TTT eval path. + if self.asym_logit_enabled: + logits = self._apply_asym_softcap(logits) + else: + 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 + # Keep raw Q for AttnOutGate src='q' (matches forward path semantics). + q_raw = F.linear(n, q_w.to(n.dtype)) + if lora.q_loras is not None: + q_raw = q_raw + 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)) + if lora.v_loras is not None: + v = v + lora.v_loras[slot](n) + v = v.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) + # AttnOutGate (TTT path) — inline + .contiguous() barrier, same as the eval path. + 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] + # Gated Attention (TTT path). Gate input is n (post-norm block input), same + # as eval path. .to(n.dtype) on fp32 param before bf16 broadcast. + 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] + # Sparse attention head-output gate (TTT path) — must match the eval path in + # forward() exactly, else training (which applied the gate) and TTT eval (which + # skipped it) produce mismatched representations and catastrophic BPB regression. + 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)) + if lora.q_loras is not None: + q_raw = q_raw + 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)) + if lora.v_loras is not None: + v = v + lora.v_loras[slot](n) + v = v.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) + # AttnOutGate (TTT parallel path) — inline + .contiguous() barrier. + 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] + # Gated Attention (TTT parallel path). Gate input is n (post-norm block input). + 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] + # Sparse attention head-output gate (TTT parallel path) — must match the + # eval path in forward() to keep train/eval semantics in sync. + 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): + # PR-1767: rank-scaled output (alpha/rank), like standard LoRA. Decouples + # effective magnitude from rank so changing rank does not change LR scale. + _ALPHA = float(os.environ.get("TTT_LORA_ALPHA", "144")) + # PR-1767: optionally keep A warm across per-doc resets (only B is zeroed). + # Accumulates useful feature directions across documents within a TTT phase. + _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, + q_lora=True, k_lora=True, v_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)] + ) + if q_lora + else None + ) + self.v_loras = ( + nn.ModuleList( + [BatchedLinearLoRA(bsz, dim, kv_dim, rank) for _ in range(num_slots)] + ) + if v_lora + else None + ) + 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() + + +# Polar Express per-iteration minimax Newton-Schulz coefficients (PR #1344). +# Replaces the fixed (3.4445, -4.775, 2.0315) coefficients of stock Muon. +# Applied at backend_steps=5 — taking more than 5 iterations from this list +# falls back to the final (converged) tuple via the slice guard below. +_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) + 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, 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, 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, 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) + # SmearGate params live on GPT root (not in .blocks), so add them by hand. + # Both are tiny (gate_window scalars + 1 lambda). Optimized via scalar Adam. + 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 if h.tie_embeddings else h.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) + self._aux_stream = torch.cuda.Stream() + + 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._aux_stream.wait_stream(torch.cuda.current_stream()) + with torch.cuda.stream(self._aux_stream): + self.optimizer_tok.step() + self.optimizer_scalar.step() + self.optimizer_muon.step() + torch.cuda.current_stream().wait_stream(self._aux_stream) + 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 = {} + act_sumsq = {} + act_counts = {} + 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]) + x_sq = x.square().sum(dim=0) + x_count = x.shape[0] + 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) + if name not in act_sumsq: + act_sumsq[name] = torch.zeros( + x.shape[1], dtype=torch.float32, device=device + ) + act_counts[name] = 0 + act_sumsq[name] += x_sq + act_counts[name] += x_count + 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) + if name not in act_sumsq: + act_sumsq[name] = torch.zeros( + y.shape[1], dtype=torch.float32, device=device + ) + act_counts[name] = 0 + act_sumsq[name] += y.square().sum(dim=0) + act_counts[name] += y.shape[0] + 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) + if name not in act_sumsq: + act_sumsq[name] = torch.zeros( + x.shape[1], dtype=torch.float32, device=device + ) + act_counts[name] = 0 + act_sumsq[name] += x.square().sum(dim=0) + act_counts[name] += x.shape[0] + 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) + if name not in act_sumsq: + act_sumsq[name] = torch.zeros( + h_act.shape[1], dtype=torch.float32, device=device + ) + act_counts[name] = 0 + act_sumsq[name] += h_act.square().sum(dim=0) + act_counts[name] += h_act.shape[0] + 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))) + + # Hessian hooks for embedding factorization projection layers + def make_linear_input_hook(weight_name): + def hook_fn(module, inp, out): + x = inp[0].detach().float() + if x.ndim == 3: + x = x.reshape(-1, x.shape[-1]) + if weight_name not in hessians: + hessians[weight_name] = torch.zeros( + x.shape[1], x.shape[1], dtype=torch.float32, device=device + ) + hessians[weight_name].addmm_(x.T, x) + return hook_fn + + if model.tie_embeddings: + hook_module = model.final_norm + + 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) + if name not in act_sumsq: + act_sumsq[name] = torch.zeros( + x.shape[1], dtype=torch.float32, device=device + ) + act_counts[name] = 0 + act_sumsq[name] += x.square().sum(dim=0) + act_counts[name] += x.shape[0] + return hook_fn + + hooks.append( + hook_module.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 + act_stats = {} + for name, sumsq in act_sumsq.items(): + count = max(act_counts.get(name, 0), 1) + act_stats[name] = (sumsq / count).sqrt().cpu() + return hessians, act_stats + + +def gptq_quantize_weight( + w, + H, + clip_sigmas=3.0, + clip_range=63, + block_size=128, + protect_groups=None, + group_size=None, + protect_clip_range=None, +): + 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] + Hinv = torch.cholesky_inverse(torch.linalg.cholesky(H)) + Hinv = torch.linalg.cholesky(Hinv, 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() + protect_meta = None + protect_mask_perm = None + s_hi = None + sf_hi = None + if ( + protect_groups + and group_size is not None + and protect_clip_range is not None + and protect_clip_range > clip_range + ): + protect_mask = torch.zeros(cols, dtype=torch.bool) + starts = [] + for (start, end) in protect_groups: + if start < 0 or end > cols or end <= start: + continue + protect_mask[start:end] = True + starts.append(start) + if starts: + protect_mask_perm = protect_mask[perm] + s_hi = (clip_sigmas * row_std / protect_clip_range).clamp_min(1e-10).to( + torch.float16 + ) + sf_hi = s_hi.float() + protect_meta = { + "starts": torch.tensor(starts, dtype=torch.int16), + "size": int(group_size), + "s_hi": s_hi, + } + 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] + if protect_mask_perm is not None and bool(protect_mask_perm[i1 + j]): + q_col = torch.clamp( + torch.round(w_col / sf_hi), + -protect_clip_range, + protect_clip_range, + ) + w_recon = q_col.float() * sf_hi + else: + q_col = torch.clamp(torch.round(w_col / sf), -clip_range, clip_range) + w_recon = q_col.float() * sf + Q[:, i1 + j] = q_col.to(torch.int8) + err = (w_col - w_recon) / 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, protect_meta + + +def _quantize_gate_int8_row(w): + # Symmetric int8-per-row quantization for small gate tensors. w shape + # (R, C) -> (R,) scales in fp16, int8 values in [-127, 127]. Single scale + # per row keeps accuracy high while halving storage vs fp16. + 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): + # A: INT2 per-matrix scalar (signed [-2,1], scale = |A|max/1.5). + 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) + # B: INT4 groupwise g over flattened B (signed [-8,7], per-group scale). + 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 _lqer_fit_quantized(E, h): + U, S, Vh = torch.linalg.svd(E, full_matrices=False) + r = min(h.lqer_rank, S.numel()) + if r <= 0: + return None + A = (U[:, :r] * S[:r]).contiguous() + B = Vh[:r, :].contiguous() + asym_on = bool(getattr(h, "lqer_asym_enabled", False)) + asym_g = int(getattr(h, "lqer_asym_group", 64)) + if asym_on and B.numel() % asym_g == 0: + qA, sA, qB, sB = _lqer_pack_asym(A, B, asym_g) + A_hat = qA.float() * float(sA) + g_sz = qB.numel() // sB.numel() + B_hat = (qB.reshape(-1, g_sz).float() * sB.float().view(-1, 1)).reshape( + qB.shape + ) + return { + "kind": "asym", + "qA": qA, + "sA": sA, + "qB": qB, + "sB": sB, + "delta": A_hat @ B_hat, + } + qA, sA, qB, sB = _lqer_pack(A, B, h.lqer_factor_bits) + A_hat = qA.float() * sA.float().view(-1, 1) + B_hat = qB.float() * sB.float().view(-1, 1) + return { + "kind": "sym", + "qA": qA, + "sA": sA, + "qB": qB, + "sB": sB, + "delta": A_hat @ B_hat, + } + + +def _awq_lite_group_candidates(w, act_rms, group_size): + cols = w.shape[1] + n_groups = cols // group_size + if n_groups <= 0: + return [] + weight_score = w.float().abs().mean(dim=0) + saliency = act_rms.float() * weight_score + cands = [] + for gi in range(n_groups): + start = gi * group_size + end = start + group_size + score = float(saliency[start:end].sum()) + cands.append((score, start, end)) + return cands + + +def gptq_mixed_quantize(state_dict, hessians, act_stats, h): + result = {} + meta = {} + quant_gate = bool(getattr(h, "gated_attn_quant_gate", False)) + lqer_on = bool(getattr(h, "lqer_enabled", False)) + awq_on = bool(getattr(h, "awq_lite_enabled", False)) + lqer_cands = {} + awq_selected = collections.defaultdict(list) + if awq_on: + awq_cands = [] + for (name, tensor) in state_dict.items(): + t = tensor.detach().cpu().contiguous() + if t.is_floating_point() and t.numel() > 65536 and name in act_stats: + bits = h.embed_bits if "tok_emb" in name else h.matrix_bits + if bits < h.awq_lite_bits: + for score, start, end in _awq_lite_group_candidates( + t, act_stats[name], h.awq_lite_group_size + ): + awq_cands.append((score, name, start, end)) + awq_cands.sort(key=lambda x: -x[0]) + for (_score, name, start, end) in awq_cands[: h.awq_lite_group_top_k]: + awq_selected[name].append((start, end)) + for (name, tensor) in state_dict.items(): + t = tensor.detach().cpu().contiguous() + # Dedicated int8-per-row path for attn_gate_w (bypasses both GPTQ and + # fp16 passthrough). Applied BEFORE the numel<=65536 passthrough check + # so the gate tensor is routed here instead of to fp16. + if ( + quant_gate + and t.is_floating_point() + and t.ndim == 2 + and name.endswith(".attn_gate_w") + # Dense GatedAttn: (num_heads, dim) = (8, 512) = 4096. + # Sparse gate: (num_heads, gate_window) = (8, 12) = 96. + # Both need int8-per-row routing; the 1024 lower bound in stock + # PR-1736 presumed dense-only. Widen to catch both. + 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 + q, s, protect_meta = gptq_quantize_weight( + t, + hessians[name], + clip_sigmas=cs, + clip_range=clip_range, + protect_groups=awq_selected.get(name), + group_size=h.awq_lite_group_size if name in awq_selected else None, + protect_clip_range=(2 ** (h.awq_lite_bits - 1) - 1) + if name in awq_selected + else None, + ) + result[name + ".q"] = q + result[name + ".scale"] = s + meta[name] = f"gptq (int{bits})" + W_q = q.float() * s.float().view(-1, 1) + if protect_meta is not None: + result[name + ".awqg_start"] = protect_meta["starts"] + result[name + ".awqg_s_hi"] = protect_meta["s_hi"] + result[name + ".awqg_size"] = torch.tensor( + protect_meta["size"], dtype=torch.int16 + ) + meta[name] = meta[name] + f"+awqgrpint{h.awq_lite_bits}" + gsz = protect_meta["size"] + for start in protect_meta["starts"].tolist(): + W_q[:, start : start + gsz] = ( + q[:, start : start + gsz].float() + * protect_meta["s_hi"].float().view(-1, 1) + ) + if lqer_on: + # LQER is fit on top of the fully realized GPTQ base, which already + # includes any higher-precision AWQ-protected groups. + scope = str(getattr(h, "lqer_scope", "all")).lower() + scope_ok = ( + scope == "all" + or (scope == "mlp" and ".mlp." in name) + or (scope == "attn" and ".attn." in name) + or (scope == "embed" and "tok_emb" in name) + ) + if scope_ok: + E = t.float() - W_q + err_norm = float(E.norm()) + if err_norm > 0: + lqer_cands[name] = (E, err_norm) + if lqer_on and lqer_cands: + if bool(getattr(h, "lqer_gain_select", False)): + scored = [] + for (name, (E, base_err)) in lqer_cands.items(): + fit = _lqer_fit_quantized(E, h) + if fit is None: + continue + new_err = float((E - fit["delta"]).norm()) + gain = base_err - new_err + if gain > 0: + scored.append((gain, name, fit)) + scored.sort(key=lambda x: -x[0]) + for (_gain, name, fit) in scored[: h.lqer_top_k]: + if fit["kind"] == "asym": + result[name + ".lqA_a"] = fit["qA"] + result[name + ".lqAs_a"] = fit["sA"] + result[name + ".lqB_a"] = fit["qB"] + result[name + ".lqBs_a"] = fit["sB"] + meta[name] = meta[name] + "+lqer_asym" + else: + result[name + ".lqA"] = fit["qA"] + result[name + ".lqAs"] = fit["sA"] + result[name + ".lqB"] = fit["qB"] + result[name + ".lqBs"] = fit["sB"] + meta[name] = meta[name] + "+lqer" + else: + 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 "awqgrpint" in info: + starts = result[name + ".awqg_start"].tolist() + s_hi = result[name + ".awqg_s_hi"].float() + gsz = int(result[name + ".awqg_size"].item()) + for start in starts: + W[:, start : start + gsz] = ( + q[:, start : start + gsz].float() * s_hi.view(-1, 1) + ) + 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" + + +# ── Per-group lrzip compression (ported from PR#1586 via PR#1667/1729) ──────── + +_GROUP_ORDER = [ + "_tok_emb.weight.q", + "attn.c_k.weight.q", "attn.c_q.weight.q", + "attn.c_v.weight.q", "attn.proj.weight.q", + "mlp.fc.weight.q", "mlp.proj.weight.q", +] +_SIMSORT_KEYS = {"_tok_emb.weight.q", "attn.c_q.weight.q", "mlp.fc.weight.q"} +_PACK_MAGIC = b"PGRP" + + +def _similarity_sort_l1(matrix): + import numpy as _np + n = matrix.shape[0] + used = _np.zeros(n, dtype=bool) + order = [0] + used[0] = True + cur = matrix[0].astype(_np.float32) + for _ in range(n - 1): + dists = _np.sum(_np.abs(matrix[~used].astype(_np.float32) - cur), axis=1) + unused = _np.where(~used)[0] + best = unused[_np.argmin(dists)] + order.append(best) + used[best] = True + cur = matrix[best].astype(_np.float32) + return _np.array(order, dtype=_np.uint16) + + +def _lrzip_compress(data, tmpdir, label): + inp = os.path.join(tmpdir, f"{label}.bin") + out = f"{inp}.lrz" + with open(inp, "wb") as f: + f.write(data) + subprocess.run(["lrzip", "-z", "-L", "9", "-o", out, inp], capture_output=True, check=True) + with open(out, "rb") as f: + result = f.read() + os.remove(inp); os.remove(out) + return result + + +def _lrzip_decompress(data, tmpdir, label): + inp = os.path.join(tmpdir, f"{label}.lrz") + out = os.path.join(tmpdir, f"{label}.bin") + with open(inp, "wb") as f: + f.write(data) + subprocess.run(["lrzip", "-d", "-f", "-o", out, inp], capture_output=True, check=True) + with open(out, "rb") as f: + result = f.read() + os.remove(inp); os.remove(out) + return result + + +def _pack_streams(streams): + import struct + n = len(streams) + hdr = _PACK_MAGIC + struct.pack("= 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 _loss_bpb_from_sums(loss_sum, token_count, byte_sum): + val_loss = (loss_sum / token_count).item() + val_bpb = val_loss / math.log(2.0) * (token_count.item() / byte_sum.item()) + return val_loss, val_bpb + + +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, + q_lora=h.ttt_q_lora, k_lora=h.ttt_k_lora, v_lora=h.ttt_v_lora, + mlp_lora=h.ttt_mlp_lora, o_lora=h.ttt_o_lora, + ).to(device) + + def _build_opt(lora): + local_lr = h.ttt_lora_lr * h.ttt_local_lr_mult + if h.ttt_optimizer == "sgd": + return torch.optim.SGD( + lora.parameters(), lr=local_lr, + momentum=h.ttt_beta1, weight_decay=h.ttt_weight_decay, + ) + return torch.optim.AdamW( + lora.parameters(), lr=local_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 + try: + 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, + q_lora=h.ttt_q_lora, k_lora=h.ttt_k_lora, v_lora=h.ttt_v_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) + # CaseOps sidecar-driven byte budget. Mirror the index pattern + # used to build y from all_tokens: y[b, j] corresponds to the + # token at global position tok_starts[b] + 1 + j (when valid). + 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 + ) + # Mirror the `valid` masking used for y so out-of-range tokens + # contribute zero bytes (matches y=0 substitution above). + 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, + q_lora=h.ttt_q_lora, k_lora=h.ttt_k_lora, v_lora=h.ttt_v_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 + finally: + pass + 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_from_sums(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 + + _clip_params = [p for p in base_model.parameters() if p.requires_grad] + def step_fn(step, lr_scale): + 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 + if step <= h.muon_momentum_warmup_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_(_clip_params, 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) + _live_state = base_model.state_dict(keep_vars=True) + ema_state = { + name: t.detach().float().clone() + for (name, t) in _live_state.items() + } + _ema_pairs = [(ema_state[name], t) for (name, t) in _live_state.items()] + ema_decay = h.ema_decay + training_time_ms = 0.0 + forced_stop_step = int(os.environ.get("FORCE_STOP_STEP", "0")) + stop_after_step = forced_stop_step if forced_stop_step > 0 else 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 ema_t, t in _ema_pairs: + ema_t.mul_(ema_decay).add_(t.detach(), 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 = ( + forced_stop_step <= 0 + and max_wallclock_ms is not None + and approx_training_time_ms >= max_wallclock_ms + ) + if h.distributed and forced_stop_step <= 0 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): + global BOS_ID + 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: skip training + GPTQ, jump straight to TTT eval on a + # pre-existing quantized artifact. Used to test TTT-only improvements + # (e.g., PR-1767's alpha/warm-start/WD) without retraining. + ttt_eval_only = os.environ.get("TTT_EVAL_ONLY", "0") == "1" + quantize_only = os.environ.get("QUANTIZE_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}") + elif quantize_only: + log("QUANTIZE_ONLY=1 — skipping training, loading saved full-precision checkpoint") + log(f"quantize_only checkpoint: {h.model_path}") + if BOS_ID is None: + BOS_ID = 1 + base_model = GPT(h).to(device).bfloat16() + state = torch.load(h.model_path, map_location="cpu") + base_model.load_state_dict(state, strict=True) + del state + serialize(h, base_model, Path(__file__).read_text(encoding="utf-8")) + if h.distributed: + dist.barrier() + 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)") + 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, + q_lora=h.ttt_q_lora, k_lora=h.ttt_k_lora, v_lora=h.ttt_v_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 * h.ttt_local_lr_mult, + 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 = 64 + 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) + log(100 * "=", console=False) + log("Hyperparameters:", console=True) + for (k, v) in sorted(vars(type(h)).items()): + if not k.startswith("_"): + log(f" {k}: {v}", console=True) + log("=" * 100, console=False) + log("Source code:", console=False) + log("=" * 100, console=False) + with open(__file__, "r", encoding="utf-8") as _src: + log(_src.read(), console=False) + log("=" * 100, console=False) + log(f"Running Python {sys.version}", console=False) + log(f"Running PyTorch {torch.__version__}", console=False) + log("=" * 100, console=False) + train_and_eval(h, device) + if distributed: + dist.destroy_process_group() + + +if __name__ == "__main__": + main() diff --git a/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed0.log b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed0.log new file mode 100644 index 0000000000..f0d07ecc65 --- /dev/null +++ b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed0.log @@ -0,0 +1,1040 @@ +W0430 04:51:22.050000 224156 torch/distributed/run.py:803] +W0430 04:51:22.050000 224156 torch/distributed/run.py:803] ***************************************** +W0430 04:51:22.050000 224156 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. +W0430 04:51:22.050000 224156 torch/distributed/run.py:803] ***************************************** +Hyperparameters: + adam_eps: 1e-08 + adam_wd: 0.02 + artifact_dir: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed0 + attn_clip_sigmas: 13.0 + attn_out_gate_enabled: False + attn_out_gate_src: proj + awq_lite_bits: 8 + awq_lite_enabled: True + awq_lite_group_size: 64 + awq_lite_group_top_k: 1 + beta1: 0.9 + beta2: 0.99 + caseops_enabled: True + compressor: pergroup + data_dir: ./data/ + datasets_dir: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved + distributed: True + ema_decay: 0.9965 + embed_bits: 7 + embed_clip_sigmas: 14.0 + embed_lr: 0.6 + embed_wd: 0.085 + enable_looping_at: 0.35 + eval_seq_len: 2560 + eval_stride: 64 + fused_ce_enabled: True + gate_window: 12 + gated_attn_enabled: False + gated_attn_init_std: 0.01 + gated_attn_quant_gate: True + global_ttt_batch_seqs: 32 + global_ttt_chunk_tokens: 32768 + global_ttt_epochs: 1 + global_ttt_grad_clip: 1.0 + global_ttt_lr: 0.001 + global_ttt_momentum: 0.9 + global_ttt_respect_doc_boundaries: True + global_ttt_warmup_chunks: 0 + global_ttt_warmup_start_lr: 0.0 + gptq_calibration_batches: 16 + gptq_reserve_seconds: 4.0 + grad_accum_steps: 1 + grad_clip_norm: 0.3 + is_main_process: True + iterations: 20000 + ln_scale: True + local_rank: 0 + logfile: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed0/rank56_prefix3000_seed0.txt + logit_softcap: 30.0 + loop_end: 5 + loop_start: 3 + lqer_asym_enabled: True + lqer_asym_group: 64 + lqer_enabled: True + lqer_factor_bits: 4 + lqer_gain_select: False + lqer_rank: 4 + lqer_scope: all + lqer_top_k: 3 + matrix_bits: 6 + matrix_clip_sigmas: 12.85 + matrix_lr: 0.026 + max_wallclock_seconds: 600.0 + min_lr: 0.1 + mlp_clip_sigmas: 11.5 + mlp_mult: 4.0 + model_dim: 512 + model_path: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed0/final_model.pt + muon_backend_steps: 5 + muon_momentum: 0.97 + muon_momentum_warmup_start: 0.92 + muon_momentum_warmup_steps: 1500 + muon_row_normalize: True + muon_wd: 0.095 + num_heads: 8 + num_kv_heads: 4 + num_layers: 11 + num_loops: 2 + parallel_final_lane: mean + parallel_start_layer: 8 + phased_ttt_num_phases: 3 + phased_ttt_prefix_docs: 3000 + qk_gain_init: 5.25 + quantized_model_path: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed0/final_model.int6.ptz + rank: 0 + rope_base: 10000.0 + rope_dims: 16 + rope_train_seq_len: 2048 + rope_yarn: False + run_id: rank56_prefix3000_seed0 + scalar_lr: 0.02 + seed: 0 + skip_gates_enabled: True + smear_gate_enabled: True + sparse_attn_gate_enabled: True + sparse_attn_gate_init_std: 0.0 + sparse_attn_gate_scale: 0.5 + tie_embeddings: True + tied_embed_init_std: 0.005 + tied_embed_lr: 0.03 + tokenizer_path: ./data/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model + train_batch_tokens: 786432 + train_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_train_*.bin + train_log_every: 500 + train_seq_len: 2048 + ttt_batch_size: 64 + ttt_beta1: 0.0 + ttt_beta2: 0.99 + ttt_chunk_size: 48 + ttt_enabled: True + ttt_eval_batches: + ttt_eval_seq_len: 2560 + ttt_grad_steps: 1 + ttt_k_lora: True + ttt_local_lr_mult: 0.75 + ttt_lora_lr: 0.0001 + ttt_lora_rank: 56 + ttt_mask: no_qv + ttt_mlp_lora: True + ttt_o_lora: True + ttt_optimizer: adam + ttt_q_lora: False + ttt_v_lora: False + ttt_weight_decay: 0.5 + val_batch_tokens: 524288 + val_bytes_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_val_bytes_*.bin + val_doc_fraction: 1.0 + val_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_val_*.bin + val_loss_every: 0 + vocab_size: 8192 + warmdown_frac: 0.85 + warmup_steps: 20 + world_size: 8 + xsa_last_n: 11 +train_shards: 80 +val_tokens: 47851520 +model_params:35945673 +gptq:reserving 4s, effective=596000ms +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 +1/20000 train_loss: 9.0105 train_time: 0.0m tok/s: 17751399 +2/20000 train_loss: 12.9464 train_time: 0.0m tok/s: 10251125 +3/20000 train_loss: 10.2623 train_time: 0.0m tok/s: 9563323 +4/20000 train_loss: 8.7687 train_time: 0.0m tok/s: 9280364 +5/20000 train_loss: 8.0043 train_time: 0.0m tok/s: 9101298 +500/20000 train_loss: 2.5748 train_time: 0.8m tok/s: 8195114 +1000/20000 train_loss: 2.8039 train_time: 1.6m tok/s: 8159237 +1500/20000 train_loss: 2.6189 train_time: 2.4m tok/s: 8145394 +2000/20000 train_loss: 2.6519 train_time: 3.2m tok/s: 8139971 +layer_loop:enabled step:2160 frac:0.351 encoder:[0, 1, 2, 3, 4, 5, 3, 4] decoder:[5, 3, 4, 5, 6, 7, 8, 9, 10] +2500/20000 train_loss: 2.5413 train_time: 4.3m tok/s: 7655376 +3000/20000 train_loss: 2.5535 train_time: 5.5m tok/s: 7204765 +3500/20000 train_loss: 2.5573 train_time: 6.6m tok/s: 6915156 +4000/20000 train_loss: 2.4014 train_time: 7.8m tok/s: 6713810 +4500/20000 train_loss: 2.2688 train_time: 9.0m tok/s: 6549612 +4879/20000 val_loss: 2.3488 val_bpb: 1.0732 +stopping_early: wallclock_cap train_time: 596086ms step: 4879/20000 +peak memory allocated: 41707 MiB reserved: 46952 MiB +ema:applying EMA weights +diagnostic pre-quantization post-ema val_loss:2.32425991 val_bpb:1.06200352 eval_time:10598ms +Serialized model: 135418111 bytes +Code size (uncompressed): 171669 bytes +Code size (compressed): 34125 bytes +GPTQ:collecting Hessians from calibration data... +GPTQ:collected 67 Hessians in 4.1s +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 (int6)+lqer_asym: blocks.mlp.fc.weight + gptq (int7)+awqgrpint8+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, softcap_neg, softcap_pos +Serialize: per-group lrzip compression... +Serialize: per-group compression done in 119.8s +Serialized model quantized+pergroup: 15942549 bytes +Total submission size quantized+pergroup: 15976674 bytes +Deserialize: per-group lrzip decompression... +Deserialize: decompression done in 20.9s +diagnostic quantized val_loss:2.34250808 val_bpb:1.07034149 eval_time:11847ms +Deserialize: per-group lrzip decompression... +Deserialize: decompression done in 20.9s +ttt_lora:warming up compile (random tokens, no val data) +ttt_lora:compile warmup done (93.9s) + +beginning TTT eval timer +ttt_phased: total_docs:50000 prefix_docs:3000 suffix_docs:47000 num_phases:3 boundaries:[1000, 2000, 3000] +ttp: b775/782 bl:2.2708 bb:1.0557 rl:2.2708 rb:1.0557 dl:6892-7524 gd:0 +ttp: b774/782 bl:2.2808 bb:1.0619 rl:2.2756 rb:1.0587 dl:6447-6872 gd:0 +ttp: b769/782 bl:2.3214 bb:1.0807 rl:2.2881 rb:1.0647 dl:5097-5309 gd:0 +ttp: b764/782 bl:2.2856 bb:1.0706 rl:2.2877 rb:1.0658 dl:4284-4392 gd:0 +ttpp: phase:1/3 pd:1424 gd:1000 t:217.0s +tttg: c1/154 lr:0.001000 t:0.3s +tttg: c2/154 lr:0.001000 t:0.4s +tttg: c3/154 lr:0.001000 t:0.5s +tttg: c4/154 lr:0.000999 t:0.6s +tttg: c5/154 lr:0.000998 t:0.6s +tttg: c6/154 lr:0.000997 t:0.7s +tttg: c7/154 lr:0.000996 t:0.8s +tttg: c8/154 lr:0.000995 t:0.9s +tttg: c9/154 lr:0.000993 t:0.9s +tttg: c10/154 lr:0.000991 t:1.0s +tttg: c11/154 lr:0.000989 t:1.1s +tttg: c12/154 lr:0.000987 t:1.2s +tttg: c13/154 lr:0.000985 t:1.3s +tttg: c14/154 lr:0.000982 t:1.3s +tttg: c15/154 lr:0.000979 t:1.4s +tttg: c16/154 lr:0.000976 t:1.5s +tttg: c17/154 lr:0.000973 t:1.6s +tttg: c18/154 lr:0.000970 t:1.7s +tttg: c19/154 lr:0.000966 t:1.7s +tttg: c20/154 lr:0.000962 t:1.8s +tttg: c21/154 lr:0.000958 t:1.9s +tttg: c22/154 lr:0.000954 t:2.0s +tttg: c23/154 lr:0.000950 t:2.0s +tttg: c24/154 lr:0.000945 t:2.1s +tttg: c25/154 lr:0.000941 t:2.2s +tttg: c26/154 lr:0.000936 t:2.3s +tttg: c27/154 lr:0.000930 t:2.4s +tttg: c28/154 lr:0.000925 t:2.4s +tttg: c29/154 lr:0.000920 t:2.5s +tttg: c30/154 lr:0.000914 t:2.6s +tttg: c31/154 lr:0.000908 t:2.7s +tttg: c32/154 lr:0.000902 t:2.7s +tttg: c33/154 lr:0.000896 t:2.8s +tttg: c34/154 lr:0.000890 t:2.9s +tttg: c35/154 lr:0.000883 t:3.0s +tttg: c36/154 lr:0.000876 t:3.1s +tttg: c37/154 lr:0.000870 t:3.1s +tttg: c38/154 lr:0.000863 t:3.2s +tttg: c39/154 lr:0.000855 t:3.3s +tttg: c40/154 lr:0.000848 t:3.4s +tttg: c41/154 lr:0.000841 t:3.5s +tttg: c42/154 lr:0.000833 t:3.5s +tttg: c43/154 lr:0.000825 t:3.6s +tttg: c44/154 lr:0.000817 t:3.7s +tttg: c45/154 lr:0.000809 t:3.8s +tttg: c46/154 lr:0.000801 t:3.9s +tttg: c47/154 lr:0.000793 t:3.9s +tttg: c48/154 lr:0.000785 t:4.0s +tttg: c49/154 lr:0.000776 t:4.1s +tttg: c50/154 lr:0.000768 t:4.2s +tttg: c51/154 lr:0.000759 t:4.3s +tttg: c52/154 lr:0.000750 t:4.3s +tttg: c53/154 lr:0.000741 t:4.4s +tttg: c54/154 lr:0.000732 t:4.5s +tttg: c55/154 lr:0.000723 t:4.6s +tttg: c56/154 lr:0.000714 t:4.7s +tttg: c57/154 lr:0.000704 t:4.7s +tttg: c58/154 lr:0.000695 t:4.8s +tttg: c59/154 lr:0.000685 t:4.9s +tttg: c60/154 lr:0.000676 t:5.0s +tttg: c61/154 lr:0.000666 t:5.1s +tttg: c62/154 lr:0.000656 t:5.1s +tttg: c63/154 lr:0.000647 t:5.2s +tttg: c64/154 lr:0.000637 t:5.3s +tttg: c65/154 lr:0.000627 t:5.4s +tttg: c66/154 lr:0.000617 t:5.4s +tttg: c67/154 lr:0.000607 t:5.5s +tttg: c68/154 lr:0.000597 t:5.6s +tttg: c69/154 lr:0.000587 t:5.7s +tttg: c70/154 lr:0.000577 t:5.8s +tttg: c71/154 lr:0.000567 t:5.8s +tttg: c72/154 lr:0.000556 t:5.9s +tttg: c73/154 lr:0.000546 t:6.0s +tttg: c74/154 lr:0.000536 t:6.1s +tttg: c75/154 lr:0.000526 t:6.2s +tttg: c76/154 lr:0.000515 t:6.2s +tttg: c77/154 lr:0.000505 t:6.3s +tttg: c78/154 lr:0.000495 t:6.4s +tttg: c79/154 lr:0.000485 t:6.5s +tttg: c80/154 lr:0.000474 t:6.6s +tttg: c81/154 lr:0.000464 t:6.6s +tttg: c82/154 lr:0.000454 t:6.7s +tttg: c83/154 lr:0.000444 t:6.8s +tttg: c84/154 lr:0.000433 t:6.9s +tttg: c85/154 lr:0.000423 t:6.9s +tttg: c86/154 lr:0.000413 t:7.0s +tttg: c87/154 lr:0.000403 t:7.1s +tttg: c88/154 lr:0.000393 t:7.2s +tttg: c89/154 lr:0.000383 t:7.3s +tttg: c90/154 lr:0.000373 t:7.3s +tttg: c91/154 lr:0.000363 t:7.4s +tttg: c92/154 lr:0.000353 t:7.5s +tttg: c93/154 lr:0.000344 t:7.6s +tttg: c94/154 lr:0.000334 t:7.6s +tttg: c95/154 lr:0.000324 t:7.7s +tttg: c96/154 lr:0.000315 t:7.8s +tttg: c97/154 lr:0.000305 t:7.9s +tttg: c98/154 lr:0.000296 t:8.0s +tttg: c99/154 lr:0.000286 t:8.1s +tttg: c100/154 lr:0.000277 t:8.1s +tttg: c101/154 lr:0.000268 t:8.2s +tttg: c102/154 lr:0.000259 t:8.3s +tttg: c103/154 lr:0.000250 t:8.4s +tttg: c104/154 lr:0.000241 t:8.5s +tttg: c105/154 lr:0.000232 t:8.5s +tttg: c106/154 lr:0.000224 t:8.6s +tttg: c107/154 lr:0.000215 t:8.7s +tttg: c108/154 lr:0.000207 t:8.8s +tttg: c109/154 lr:0.000199 t:8.8s +tttg: c110/154 lr:0.000191 t:8.9s +tttg: c111/154 lr:0.000183 t:9.0s +tttg: c112/154 lr:0.000175 t:9.1s +tttg: c113/154 lr:0.000167 t:9.2s +tttg: c114/154 lr:0.000159 t:9.2s +tttg: c115/154 lr:0.000152 t:9.3s +tttg: c116/154 lr:0.000145 t:9.4s +tttg: c117/154 lr:0.000137 t:9.5s +tttg: c118/154 lr:0.000130 t:9.6s +tttg: c119/154 lr:0.000124 t:9.6s +tttg: c120/154 lr:0.000117 t:9.7s +tttg: c121/154 lr:0.000110 t:9.8s +tttg: c122/154 lr:0.000104 t:9.9s +tttg: c123/154 lr:0.000098 t:10.0s +tttg: c124/154 lr:0.000092 t:10.0s +tttg: c125/154 lr:0.000086 t:10.1s +tttg: c126/154 lr:0.000080 t:10.2s +tttg: c127/154 lr:0.000075 t:10.3s +tttg: c128/154 lr:0.000070 t:10.3s +tttg: c129/154 lr:0.000064 t:10.4s +tttg: c130/154 lr:0.000059 t:10.5s +tttg: c131/154 lr:0.000055 t:10.6s +tttg: c132/154 lr:0.000050 t:10.7s +tttg: c133/154 lr:0.000046 t:10.7s +tttg: c134/154 lr:0.000042 t:10.8s +tttg: c135/154 lr:0.000038 t:10.9s +tttg: c136/154 lr:0.000034 t:11.0s +tttg: c137/154 lr:0.000030 t:11.1s +tttg: c138/154 lr:0.000027 t:11.1s +tttg: c139/154 lr:0.000024 t:11.2s +tttg: c140/154 lr:0.000021 t:11.3s +tttg: c141/154 lr:0.000018 t:11.4s +tttg: c142/154 lr:0.000015 t:11.4s +tttg: c143/154 lr:0.000013 t:11.5s +tttg: c144/154 lr:0.000011 t:11.6s +tttg: c145/154 lr:0.000009 t:11.7s +tttg: c146/154 lr:0.000007 t:11.8s +tttg: c147/154 lr:0.000005 t:11.8s +tttg: c148/154 lr:0.000004 t:11.9s +tttg: c149/154 lr:0.000003 t:12.0s +tttg: c150/154 lr:0.000002 t:12.1s +tttg: c151/154 lr:0.000001 t:12.2s +tttg: c152/154 lr:0.000000 t:12.2s +tttg: c153/154 lr:0.000000 t:12.3s +ttpr: phase:1/3 t:230.1s +ttp: b752/782 bl:2.3284 bb:1.0703 rl:2.2926 rb:1.0663 dl:3222-3283 gd:0 +ttp: b751/782 bl:2.3029 bb:1.0309 rl:2.2937 rb:1.0624 dl:3150-3221 gd:0 +ttp: b748/782 bl:2.3104 bb:1.0782 rl:2.2953 rb:1.0639 dl:2992-3039 gd:0 +ttp: b745/782 bl:2.2273 bb:1.0197 rl:2.2898 rb:1.0603 dl:2842-2883 gd:0 +ttpp: phase:2/3 pd:2448 gd:2000 t:287.3s +tttg: c1/250 lr:0.001000 t:0.1s +tttg: c2/250 lr:0.001000 t:0.2s +tttg: c3/250 lr:0.001000 t:0.2s +tttg: c4/250 lr:0.001000 t:0.3s +tttg: c5/250 lr:0.000999 t:0.4s +tttg: c6/250 lr:0.000999 t:0.5s +tttg: c7/250 lr:0.000999 t:0.6s +tttg: c8/250 lr:0.000998 t:0.6s +tttg: c9/250 lr:0.000997 t:0.7s +tttg: c10/250 lr:0.000997 t:0.8s +tttg: c11/250 lr:0.000996 t:0.9s +tttg: c12/250 lr:0.000995 t:0.9s +tttg: c13/250 lr:0.000994 t:1.0s +tttg: c14/250 lr:0.000993 t:1.1s +tttg: c15/250 lr:0.000992 t:1.2s +tttg: c16/250 lr:0.000991 t:1.3s +tttg: c17/250 lr:0.000990 t:1.3s +tttg: c18/250 lr:0.000989 t:1.4s +tttg: c19/250 lr:0.000987 t:1.5s +tttg: c20/250 lr:0.000986 t:1.6s +tttg: c21/250 lr:0.000984 t:1.6s +tttg: c22/250 lr:0.000983 t:1.7s +tttg: c23/250 lr:0.000981 t:1.8s +tttg: c24/250 lr:0.000979 t:1.9s +tttg: c25/250 lr:0.000977 t:2.0s +tttg: c26/250 lr:0.000975 t:2.0s +tttg: c27/250 lr:0.000973 t:2.1s +tttg: c28/250 lr:0.000971 t:2.2s +tttg: c29/250 lr:0.000969 t:2.3s +tttg: c30/250 lr:0.000967 t:2.4s +tttg: c31/250 lr:0.000965 t:2.4s +tttg: c32/250 lr:0.000962 t:2.5s +tttg: c33/250 lr:0.000960 t:2.6s +tttg: c34/250 lr:0.000957 t:2.7s +tttg: c35/250 lr:0.000955 t:2.8s +tttg: c36/250 lr:0.000952 t:2.8s +tttg: c37/250 lr:0.000949 t:2.9s +tttg: c38/250 lr:0.000947 t:3.0s +tttg: c39/250 lr:0.000944 t:3.1s +tttg: c40/250 lr:0.000941 t:3.2s +tttg: c41/250 lr:0.000938 t:3.2s +tttg: c42/250 lr:0.000935 t:3.3s +tttg: c43/250 lr:0.000931 t:3.4s +tttg: c44/250 lr:0.000928 t:3.5s +tttg: c45/250 lr:0.000925 t:3.6s +tttg: c46/250 lr:0.000922 t:3.6s +tttg: c47/250 lr:0.000918 t:3.7s +tttg: c48/250 lr:0.000915 t:3.8s +tttg: c49/250 lr:0.000911 t:3.9s +tttg: c50/250 lr:0.000907 t:4.0s +tttg: c51/250 lr:0.000904 t:4.0s +tttg: c52/250 lr:0.000900 t:4.1s +tttg: c53/250 lr:0.000896 t:4.2s +tttg: c54/250 lr:0.000892 t:4.3s +tttg: c55/250 lr:0.000888 t:4.3s +tttg: c56/250 lr:0.000884 t:4.4s +tttg: c57/250 lr:0.000880 t:4.5s +tttg: c58/250 lr:0.000876 t:4.6s +tttg: c59/250 lr:0.000872 t:4.7s +tttg: c60/250 lr:0.000868 t:4.7s +tttg: c61/250 lr:0.000863 t:4.8s +tttg: c62/250 lr:0.000859 t:4.9s +tttg: c63/250 lr:0.000855 t:5.0s +tttg: c64/250 lr:0.000850 t:5.1s +tttg: c65/250 lr:0.000846 t:5.1s +tttg: c66/250 lr:0.000841 t:5.2s +tttg: c67/250 lr:0.000836 t:5.3s +tttg: c68/250 lr:0.000832 t:5.4s +tttg: c69/250 lr:0.000827 t:5.5s +tttg: c70/250 lr:0.000822 t:5.5s +tttg: c71/250 lr:0.000817 t:5.6s +tttg: c72/250 lr:0.000812 t:5.7s +tttg: c73/250 lr:0.000807 t:5.8s +tttg: c74/250 lr:0.000803 t:5.8s +tttg: c75/250 lr:0.000797 t:5.9s +tttg: c76/250 lr:0.000792 t:6.0s +tttg: c77/250 lr:0.000787 t:6.1s +tttg: c78/250 lr:0.000782 t:6.2s +tttg: c79/250 lr:0.000777 t:6.2s +tttg: c80/250 lr:0.000772 t:6.3s +tttg: c81/250 lr:0.000766 t:6.4s +tttg: c82/250 lr:0.000761 t:6.5s +tttg: c83/250 lr:0.000755 t:6.6s +tttg: c84/250 lr:0.000750 t:6.6s +tttg: c85/250 lr:0.000745 t:6.7s +tttg: c86/250 lr:0.000739 t:6.8s +tttg: c87/250 lr:0.000733 t:6.9s +tttg: c88/250 lr:0.000728 t:7.0s +tttg: c89/250 lr:0.000722 t:7.0s +tttg: c90/250 lr:0.000717 t:7.1s +tttg: c91/250 lr:0.000711 t:7.2s +tttg: c92/250 lr:0.000705 t:7.3s +tttg: c93/250 lr:0.000699 t:7.3s +tttg: c94/250 lr:0.000694 t:7.4s +tttg: c95/250 lr:0.000688 t:7.5s +tttg: c96/250 lr:0.000682 t:7.6s +tttg: c97/250 lr:0.000676 t:7.7s +tttg: c98/250 lr:0.000670 t:7.7s +tttg: c99/250 lr:0.000664 t:7.8s +tttg: c100/250 lr:0.000658 t:7.9s +tttg: c101/250 lr:0.000652 t:8.0s +tttg: c102/250 lr:0.000646 t:8.1s +tttg: c103/250 lr:0.000640 t:8.1s +tttg: c104/250 lr:0.000634 t:8.2s +tttg: c105/250 lr:0.000628 t:8.3s +tttg: c106/250 lr:0.000622 t:8.4s +tttg: c107/250 lr:0.000616 t:8.5s +tttg: c108/250 lr:0.000610 t:8.5s +tttg: c109/250 lr:0.000603 t:8.6s +tttg: c110/250 lr:0.000597 t:8.7s +tttg: c111/250 lr:0.000591 t:8.8s +tttg: c112/250 lr:0.000585 t:8.9s +tttg: c113/250 lr:0.000579 t:8.9s +tttg: c114/250 lr:0.000572 t:9.0s +tttg: c115/250 lr:0.000566 t:9.1s +tttg: c116/250 lr:0.000560 t:9.2s +tttg: c117/250 lr:0.000554 t:9.2s +tttg: c118/250 lr:0.000547 t:9.3s +tttg: c119/250 lr:0.000541 t:9.4s +tttg: c120/250 lr:0.000535 t:9.5s +tttg: c121/250 lr:0.000528 t:9.6s +tttg: c122/250 lr:0.000522 t:9.6s +tttg: c123/250 lr:0.000516 t:9.7s +tttg: c124/250 lr:0.000509 t:9.8s +tttg: c125/250 lr:0.000503 t:9.9s +tttg: c126/250 lr:0.000497 t:10.0s +tttg: c127/250 lr:0.000491 t:10.0s +tttg: c128/250 lr:0.000484 t:10.1s +tttg: c129/250 lr:0.000478 t:10.2s +tttg: c130/250 lr:0.000472 t:10.3s +tttg: c131/250 lr:0.000465 t:10.4s +tttg: c132/250 lr:0.000459 t:10.4s +tttg: c133/250 lr:0.000453 t:10.5s +tttg: c134/250 lr:0.000446 t:10.6s +tttg: c135/250 lr:0.000440 t:10.7s +tttg: c136/250 lr:0.000434 t:10.8s +tttg: c137/250 lr:0.000428 t:10.8s +tttg: c138/250 lr:0.000421 t:10.9s +tttg: c139/250 lr:0.000415 t:11.0s +tttg: c140/250 lr:0.000409 t:11.1s +tttg: c141/250 lr:0.000403 t:11.1s +tttg: c142/250 lr:0.000397 t:11.2s +tttg: c143/250 lr:0.000390 t:11.3s +tttg: c144/250 lr:0.000384 t:11.4s +tttg: c145/250 lr:0.000378 t:11.5s +tttg: c146/250 lr:0.000372 t:11.5s +tttg: c147/250 lr:0.000366 t:11.6s +tttg: c148/250 lr:0.000360 t:11.7s +tttg: c149/250 lr:0.000354 t:11.8s +tttg: c150/250 lr:0.000348 t:11.9s +tttg: c151/250 lr:0.000342 t:11.9s +tttg: c152/250 lr:0.000336 t:12.0s +tttg: c153/250 lr:0.000330 t:12.1s +tttg: c154/250 lr:0.000324 t:12.2s +tttg: c155/250 lr:0.000318 t:12.2s +tttg: c156/250 lr:0.000312 t:12.3s +tttg: c157/250 lr:0.000306 t:12.4s +tttg: c158/250 lr:0.000301 t:12.5s +tttg: c159/250 lr:0.000295 t:12.6s +tttg: c160/250 lr:0.000289 t:12.6s +tttg: c161/250 lr:0.000283 t:12.7s +tttg: c162/250 lr:0.000278 t:12.8s +tttg: c163/250 lr:0.000272 t:12.9s +tttg: c164/250 lr:0.000267 t:13.0s +tttg: c165/250 lr:0.000261 t:13.0s +tttg: c166/250 lr:0.000255 t:13.1s +tttg: c167/250 lr:0.000250 t:13.2s +tttg: c168/250 lr:0.000245 t:13.3s +tttg: c169/250 lr:0.000239 t:13.4s +tttg: c170/250 lr:0.000234 t:13.4s +tttg: c171/250 lr:0.000228 t:13.5s +tttg: c172/250 lr:0.000223 t:13.6s +tttg: c173/250 lr:0.000218 t:13.7s +tttg: c174/250 lr:0.000213 t:13.7s +tttg: c175/250 lr:0.000208 t:13.8s +tttg: c176/250 lr:0.000203 t:13.9s +tttg: c177/250 lr:0.000197 t:14.0s +tttg: c178/250 lr:0.000193 t:14.1s +tttg: c179/250 lr:0.000188 t:14.1s +tttg: c180/250 lr:0.000183 t:14.2s +tttg: c181/250 lr:0.000178 t:14.3s +tttg: c182/250 lr:0.000173 t:14.4s +tttg: c183/250 lr:0.000168 t:14.5s +tttg: c184/250 lr:0.000164 t:14.5s +tttg: c185/250 lr:0.000159 t:14.6s +tttg: c186/250 lr:0.000154 t:14.7s +tttg: c187/250 lr:0.000150 t:14.8s +tttg: c188/250 lr:0.000145 t:14.8s +tttg: c189/250 lr:0.000141 t:14.9s +tttg: c190/250 lr:0.000137 t:15.0s +tttg: c191/250 lr:0.000132 t:15.1s +tttg: c192/250 lr:0.000128 t:15.2s +tttg: c193/250 lr:0.000124 t:15.2s +tttg: c194/250 lr:0.000120 t:15.3s +tttg: c195/250 lr:0.000116 t:15.4s +tttg: c196/250 lr:0.000112 t:15.5s +tttg: c197/250 lr:0.000108 t:15.6s +tttg: c198/250 lr:0.000104 t:15.6s +tttg: c199/250 lr:0.000100 t:15.7s +tttg: c200/250 lr:0.000096 t:15.8s +tttg: c201/250 lr:0.000093 t:15.9s +tttg: c202/250 lr:0.000089 t:15.9s +tttg: c203/250 lr:0.000085 t:16.0s +tttg: c204/250 lr:0.000082 t:16.1s +tttg: c205/250 lr:0.000078 t:16.2s +tttg: c206/250 lr:0.000075 t:16.3s +tttg: c207/250 lr:0.000072 t:16.3s +tttg: c208/250 lr:0.000069 t:16.4s +tttg: c209/250 lr:0.000065 t:16.5s +tttg: c210/250 lr:0.000062 t:16.6s +tttg: c211/250 lr:0.000059 t:16.7s +tttg: c212/250 lr:0.000056 t:16.8s +tttg: c213/250 lr:0.000053 t:16.8s +tttg: c214/250 lr:0.000051 t:16.9s +tttg: c215/250 lr:0.000048 t:17.0s +tttg: c216/250 lr:0.000045 t:17.1s +tttg: c217/250 lr:0.000043 t:17.1s +tttg: c218/250 lr:0.000040 t:17.2s +tttg: c219/250 lr:0.000038 t:17.3s +tttg: c220/250 lr:0.000035 t:17.4s +tttg: c221/250 lr:0.000033 t:17.5s +tttg: c222/250 lr:0.000031 t:17.5s +tttg: c223/250 lr:0.000029 t:17.6s +tttg: c224/250 lr:0.000027 t:17.7s +tttg: c225/250 lr:0.000025 t:17.8s +tttg: c226/250 lr:0.000023 t:17.8s +tttg: c227/250 lr:0.000021 t:17.9s +tttg: c228/250 lr:0.000019 t:18.0s +tttg: c229/250 lr:0.000017 t:18.1s +tttg: c230/250 lr:0.000016 t:18.2s +tttg: c231/250 lr:0.000014 t:18.2s +tttg: c232/250 lr:0.000013 t:18.3s +tttg: c233/250 lr:0.000011 t:18.4s +tttg: c234/250 lr:0.000010 t:18.5s +tttg: c235/250 lr:0.000009 t:18.6s +tttg: c236/250 lr:0.000008 t:18.6s +tttg: c237/250 lr:0.000007 t:18.7s +tttg: c238/250 lr:0.000006 t:18.8s +tttg: c239/250 lr:0.000005 t:18.9s +tttg: c240/250 lr:0.000004 t:19.0s +tttg: c241/250 lr:0.000003 t:19.0s +tttg: c242/250 lr:0.000003 t:19.1s +tttg: c243/250 lr:0.000002 t:19.2s +tttg: c244/250 lr:0.000001 t:19.3s +tttg: c245/250 lr:0.000001 t:19.3s +tttg: c246/250 lr:0.000001 t:19.4s +tttg: c247/250 lr:0.000000 t:19.5s +tttg: c248/250 lr:0.000000 t:19.6s +tttg: c249/250 lr:0.000000 t:19.7s +ttpr: phase:2/3 t:307.8s +ttp: b736/782 bl:2.2403 bb:1.0555 rl:2.2865 rb:1.0600 dl:2526-2550 gd:0 +ttp: b735/782 bl:2.3819 bb:1.0958 rl:2.2924 rb:1.0622 dl:2495-2526 gd:0 +ttpp: phase:3/3 pd:3472 gd:3000 t:320.7s +tttg: c1/325 lr:0.001000 t:0.1s +tttg: c2/325 lr:0.001000 t:0.2s +tttg: c3/325 lr:0.001000 t:0.2s +tttg: c4/325 lr:0.001000 t:0.3s +tttg: c5/325 lr:0.001000 t:0.4s +tttg: c6/325 lr:0.000999 t:0.5s +tttg: c7/325 lr:0.000999 t:0.5s +tttg: c8/325 lr:0.000999 t:0.6s +tttg: c9/325 lr:0.000998 t:0.7s +tttg: c10/325 lr:0.000998 t:0.8s +tttg: c11/325 lr:0.000998 t:0.8s +tttg: c12/325 lr:0.000997 t:0.9s +tttg: c13/325 lr:0.000997 t:1.0s +tttg: c14/325 lr:0.000996 t:1.1s +tttg: c15/325 lr:0.000995 t:1.2s +tttg: c16/325 lr:0.000995 t:1.2s +tttg: c17/325 lr:0.000994 t:1.3s +tttg: c18/325 lr:0.000993 t:1.4s +tttg: c19/325 lr:0.000992 t:1.5s +tttg: c20/325 lr:0.000992 t:1.6s +tttg: c21/325 lr:0.000991 t:1.6s +tttg: c22/325 lr:0.000990 t:1.7s +tttg: c23/325 lr:0.000989 t:1.8s +tttg: c24/325 lr:0.000988 t:1.9s +tttg: c25/325 lr:0.000987 t:2.0s +tttg: c26/325 lr:0.000985 t:2.0s +tttg: c27/325 lr:0.000984 t:2.1s +tttg: c28/325 lr:0.000983 t:2.2s +tttg: c29/325 lr:0.000982 t:2.3s +tttg: c30/325 lr:0.000980 t:2.3s +tttg: c31/325 lr:0.000979 t:2.4s +tttg: c32/325 lr:0.000978 t:2.5s +tttg: c33/325 lr:0.000976 t:2.6s +tttg: c34/325 lr:0.000975 t:2.7s +tttg: c35/325 lr:0.000973 t:2.7s +tttg: c36/325 lr:0.000971 t:2.8s +tttg: c37/325 lr:0.000970 t:2.9s +tttg: c38/325 lr:0.000968 t:3.0s +tttg: c39/325 lr:0.000966 t:3.1s +tttg: c40/325 lr:0.000965 t:3.1s +tttg: c41/325 lr:0.000963 t:3.2s +tttg: c42/325 lr:0.000961 t:3.3s +tttg: c43/325 lr:0.000959 t:3.4s +tttg: c44/325 lr:0.000957 t:3.4s +tttg: c45/325 lr:0.000955 t:3.5s +tttg: c46/325 lr:0.000953 t:3.6s +tttg: c47/325 lr:0.000951 t:3.7s +tttg: c48/325 lr:0.000949 t:3.8s +tttg: c49/325 lr:0.000947 t:3.8s +tttg: c50/325 lr:0.000945 t:3.9s +tttg: c51/325 lr:0.000942 t:4.0s +tttg: c52/325 lr:0.000940 t:4.1s +tttg: c53/325 lr:0.000938 t:4.2s +tttg: c54/325 lr:0.000935 t:4.2s +tttg: c55/325 lr:0.000933 t:4.3s +tttg: c56/325 lr:0.000931 t:4.4s +tttg: c57/325 lr:0.000928 t:4.5s +tttg: c58/325 lr:0.000926 t:4.5s +tttg: c59/325 lr:0.000923 t:4.6s +tttg: c60/325 lr:0.000920 t:4.7s +tttg: c61/325 lr:0.000918 t:4.8s +tttg: c62/325 lr:0.000915 t:4.9s +tttg: c63/325 lr:0.000912 t:4.9s +tttg: c64/325 lr:0.000910 t:5.0s +tttg: c65/325 lr:0.000907 t:5.1s +tttg: c66/325 lr:0.000904 t:5.2s +tttg: c67/325 lr:0.000901 t:5.3s +tttg: c68/325 lr:0.000898 t:5.3s +tttg: c69/325 lr:0.000895 t:5.4s +tttg: c70/325 lr:0.000892 t:5.5s +tttg: c71/325 lr:0.000889 t:5.6s +tttg: c72/325 lr:0.000886 t:5.6s +tttg: c73/325 lr:0.000883 t:5.7s +tttg: c74/325 lr:0.000880 t:5.8s +tttg: c75/325 lr:0.000877 t:5.9s +tttg: c76/325 lr:0.000874 t:6.0s +tttg: c77/325 lr:0.000870 t:6.0s +tttg: c78/325 lr:0.000867 t:6.1s +tttg: c79/325 lr:0.000864 t:6.2s +tttg: c80/325 lr:0.000860 t:6.3s +tttg: c81/325 lr:0.000857 t:6.4s +tttg: c82/325 lr:0.000854 t:6.4s +tttg: c83/325 lr:0.000850 t:6.5s +tttg: c84/325 lr:0.000847 t:6.6s +tttg: c85/325 lr:0.000843 t:6.7s +tttg: c86/325 lr:0.000840 t:6.7s +tttg: c87/325 lr:0.000836 t:6.8s +tttg: c88/325 lr:0.000832 t:6.9s +tttg: c89/325 lr:0.000829 t:7.0s +tttg: c90/325 lr:0.000825 t:7.1s +tttg: c91/325 lr:0.000821 t:7.1s +tttg: c92/325 lr:0.000818 t:7.2s +tttg: c93/325 lr:0.000814 t:7.3s +tttg: c94/325 lr:0.000810 t:7.4s +tttg: c95/325 lr:0.000806 t:7.5s +tttg: c96/325 lr:0.000802 t:7.5s +tttg: c97/325 lr:0.000799 t:7.6s +tttg: c98/325 lr:0.000795 t:7.7s +tttg: c99/325 lr:0.000791 t:7.8s +tttg: c100/325 lr:0.000787 t:7.8s +tttg: c101/325 lr:0.000783 t:7.9s +tttg: c102/325 lr:0.000779 t:8.0s +tttg: c103/325 lr:0.000775 t:8.1s +tttg: c104/325 lr:0.000771 t:8.2s +tttg: c105/325 lr:0.000767 t:8.2s +tttg: c106/325 lr:0.000762 t:8.3s +tttg: c107/325 lr:0.000758 t:8.4s +tttg: c108/325 lr:0.000754 t:8.5s +tttg: c109/325 lr:0.000750 t:8.6s +tttg: c110/325 lr:0.000746 t:8.6s +tttg: c111/325 lr:0.000742 t:8.7s +tttg: c112/325 lr:0.000737 t:8.8s +tttg: c113/325 lr:0.000733 t:8.9s +tttg: c114/325 lr:0.000729 t:8.9s +tttg: c115/325 lr:0.000724 t:9.0s +tttg: c116/325 lr:0.000720 t:9.1s +tttg: c117/325 lr:0.000716 t:9.2s +tttg: c118/325 lr:0.000711 t:9.3s +tttg: c119/325 lr:0.000707 t:9.3s +tttg: c120/325 lr:0.000702 t:9.4s +tttg: c121/325 lr:0.000698 t:9.5s +tttg: c122/325 lr:0.000694 t:9.6s +tttg: c123/325 lr:0.000689 t:9.6s +tttg: c124/325 lr:0.000685 t:9.7s +tttg: c125/325 lr:0.000680 t:9.8s +tttg: c126/325 lr:0.000676 t:9.9s +tttg: c127/325 lr:0.000671 t:10.0s +tttg: c128/325 lr:0.000666 t:10.0s +tttg: c129/325 lr:0.000662 t:10.1s +tttg: c130/325 lr:0.000657 t:10.2s +tttg: c131/325 lr:0.000653 t:10.3s +tttg: c132/325 lr:0.000648 t:10.4s +tttg: c133/325 lr:0.000643 t:10.4s +tttg: c134/325 lr:0.000639 t:10.5s +tttg: c135/325 lr:0.000634 t:10.6s +tttg: c136/325 lr:0.000629 t:10.7s +tttg: c137/325 lr:0.000625 t:10.8s +tttg: c138/325 lr:0.000620 t:10.8s +tttg: c139/325 lr:0.000615 t:10.9s +tttg: c140/325 lr:0.000611 t:11.0s +tttg: c141/325 lr:0.000606 t:11.1s +tttg: c142/325 lr:0.000601 t:11.2s +tttg: c143/325 lr:0.000596 t:11.2s +tttg: c144/325 lr:0.000592 t:11.3s +tttg: c145/325 lr:0.000587 t:11.4s +tttg: c146/325 lr:0.000582 t:11.5s +tttg: c147/325 lr:0.000577 t:11.5s +tttg: c148/325 lr:0.000572 t:11.6s +tttg: c149/325 lr:0.000568 t:11.7s +tttg: c150/325 lr:0.000563 t:11.8s +tttg: c151/325 lr:0.000558 t:11.9s +tttg: c152/325 lr:0.000553 t:11.9s +tttg: c153/325 lr:0.000548 t:12.0s +tttg: c154/325 lr:0.000544 t:12.1s +tttg: c155/325 lr:0.000539 t:12.2s +tttg: c156/325 lr:0.000534 t:12.2s +tttg: c157/325 lr:0.000529 t:12.3s +tttg: c158/325 lr:0.000524 t:12.4s +tttg: c159/325 lr:0.000519 t:12.5s +tttg: c160/325 lr:0.000515 t:12.6s +tttg: c161/325 lr:0.000510 t:12.6s +tttg: c162/325 lr:0.000505 t:12.7s +tttg: c163/325 lr:0.000500 t:12.8s +tttg: c164/325 lr:0.000495 t:12.9s +tttg: c165/325 lr:0.000490 t:13.0s +tttg: c166/325 lr:0.000485 t:13.0s +tttg: c167/325 lr:0.000481 t:13.1s +tttg: c168/325 lr:0.000476 t:13.2s +tttg: c169/325 lr:0.000471 t:13.3s +tttg: c170/325 lr:0.000466 t:13.4s +tttg: c171/325 lr:0.000461 t:13.4s +tttg: c172/325 lr:0.000456 t:13.5s +tttg: c173/325 lr:0.000452 t:13.6s +tttg: c174/325 lr:0.000447 t:13.7s +tttg: c175/325 lr:0.000442 t:13.7s +tttg: c176/325 lr:0.000437 t:13.8s +tttg: c177/325 lr:0.000432 t:13.9s +tttg: c178/325 lr:0.000428 t:14.0s +tttg: c179/325 lr:0.000423 t:14.1s +tttg: c180/325 lr:0.000418 t:14.2s +tttg: c181/325 lr:0.000413 t:14.2s +tttg: c182/325 lr:0.000408 t:14.3s +tttg: c183/325 lr:0.000404 t:14.4s +tttg: c184/325 lr:0.000399 t:14.5s +tttg: c185/325 lr:0.000394 t:14.5s +tttg: c186/325 lr:0.000389 t:14.6s +tttg: c187/325 lr:0.000385 t:14.7s +tttg: c188/325 lr:0.000380 t:14.8s +tttg: c189/325 lr:0.000375 t:14.8s +tttg: c190/325 lr:0.000371 t:14.9s +tttg: c191/325 lr:0.000366 t:15.0s +tttg: c192/325 lr:0.000361 t:15.1s +tttg: c193/325 lr:0.000357 t:15.2s +tttg: c194/325 lr:0.000352 t:15.2s +tttg: c195/325 lr:0.000347 t:15.3s +tttg: c196/325 lr:0.000343 t:15.4s +tttg: c197/325 lr:0.000338 t:15.5s +tttg: c198/325 lr:0.000334 t:15.6s +tttg: c199/325 lr:0.000329 t:15.6s +tttg: c200/325 lr:0.000324 t:15.7s +tttg: c201/325 lr:0.000320 t:15.8s +tttg: c202/325 lr:0.000315 t:15.9s +tttg: c203/325 lr:0.000311 t:16.0s +tttg: c204/325 lr:0.000306 t:16.0s +tttg: c205/325 lr:0.000302 t:16.1s +tttg: c206/325 lr:0.000298 t:16.2s +tttg: c207/325 lr:0.000293 t:16.3s +tttg: c208/325 lr:0.000289 t:16.3s +tttg: c209/325 lr:0.000284 t:16.4s +tttg: c210/325 lr:0.000280 t:16.5s +tttg: c211/325 lr:0.000276 t:16.6s +tttg: c212/325 lr:0.000271 t:16.7s +tttg: c213/325 lr:0.000267 t:16.7s +tttg: c214/325 lr:0.000263 t:16.8s +tttg: c215/325 lr:0.000258 t:16.9s +tttg: c216/325 lr:0.000254 t:17.0s +tttg: c217/325 lr:0.000250 t:17.0s +tttg: c218/325 lr:0.000246 t:17.1s +tttg: c219/325 lr:0.000242 t:17.2s +tttg: c220/325 lr:0.000238 t:17.3s +tttg: c221/325 lr:0.000233 t:17.4s +tttg: c222/325 lr:0.000229 t:17.4s +tttg: c223/325 lr:0.000225 t:17.5s +tttg: c224/325 lr:0.000221 t:17.6s +tttg: c225/325 lr:0.000217 t:17.7s +tttg: c226/325 lr:0.000213 t:17.8s +tttg: c227/325 lr:0.000209 t:17.8s +tttg: c228/325 lr:0.000205 t:17.9s +tttg: c229/325 lr:0.000201 t:18.0s +tttg: c230/325 lr:0.000198 t:18.1s +tttg: c231/325 lr:0.000194 t:18.2s +tttg: c232/325 lr:0.000190 t:18.2s +tttg: c233/325 lr:0.000186 t:18.3s +tttg: c234/325 lr:0.000182 t:18.4s +tttg: c235/325 lr:0.000179 t:18.5s +tttg: c236/325 lr:0.000175 t:18.5s +tttg: c237/325 lr:0.000171 t:18.6s +tttg: c238/325 lr:0.000168 t:18.7s +tttg: c239/325 lr:0.000164 t:18.8s +tttg: c240/325 lr:0.000160 t:18.9s +tttg: c241/325 lr:0.000157 t:18.9s +tttg: c242/325 lr:0.000153 t:19.0s +tttg: c243/325 lr:0.000150 t:19.1s +tttg: c244/325 lr:0.000146 t:19.2s +tttg: c245/325 lr:0.000143 t:19.3s +tttg: c246/325 lr:0.000140 t:19.3s +tttg: c247/325 lr:0.000136 t:19.4s +tttg: c248/325 lr:0.000133 t:19.5s +tttg: c249/325 lr:0.000130 t:19.6s +tttg: c250/325 lr:0.000126 t:19.6s +tttg: c251/325 lr:0.000123 t:19.7s +tttg: c252/325 lr:0.000120 t:19.8s +tttg: c253/325 lr:0.000117 t:19.9s +tttg: c254/325 lr:0.000114 t:20.0s +tttg: c255/325 lr:0.000111 t:20.0s +tttg: c256/325 lr:0.000108 t:20.1s +tttg: c257/325 lr:0.000105 t:20.2s +tttg: c258/325 lr:0.000102 t:20.3s +tttg: c259/325 lr:0.000099 t:20.3s +tttg: c260/325 lr:0.000096 t:20.4s +tttg: c261/325 lr:0.000093 t:20.5s +tttg: c262/325 lr:0.000090 t:20.6s +tttg: c263/325 lr:0.000088 t:20.7s +tttg: c264/325 lr:0.000085 t:20.7s +tttg: c265/325 lr:0.000082 t:20.8s +tttg: c266/325 lr:0.000080 t:20.9s +tttg: c267/325 lr:0.000077 t:21.0s +tttg: c268/325 lr:0.000074 t:21.1s +tttg: c269/325 lr:0.000072 t:21.1s +tttg: c270/325 lr:0.000069 t:21.2s +tttg: c271/325 lr:0.000067 t:21.3s +tttg: c272/325 lr:0.000065 t:21.4s +tttg: c273/325 lr:0.000062 t:21.5s +tttg: c274/325 lr:0.000060 t:21.5s +tttg: c275/325 lr:0.000058 t:21.6s +tttg: c276/325 lr:0.000055 t:21.7s +tttg: c277/325 lr:0.000053 t:21.8s +tttg: c278/325 lr:0.000051 t:21.9s +tttg: c279/325 lr:0.000049 t:21.9s +tttg: c280/325 lr:0.000047 t:22.0s +tttg: c281/325 lr:0.000045 t:22.1s +tttg: c282/325 lr:0.000043 t:22.2s +tttg: c283/325 lr:0.000041 t:22.3s +tttg: c284/325 lr:0.000039 t:22.3s +tttg: c285/325 lr:0.000037 t:22.4s +tttg: c286/325 lr:0.000035 t:22.5s +tttg: c287/325 lr:0.000034 t:22.6s +tttg: c288/325 lr:0.000032 t:22.6s +tttg: c289/325 lr:0.000030 t:22.7s +tttg: c290/325 lr:0.000029 t:22.8s +tttg: c291/325 lr:0.000027 t:22.9s +tttg: c292/325 lr:0.000025 t:23.0s +tttg: c293/325 lr:0.000024 t:23.1s +tttg: c294/325 lr:0.000022 t:23.1s +tttg: c295/325 lr:0.000021 t:23.2s +tttg: c296/325 lr:0.000020 t:23.3s +tttg: c297/325 lr:0.000018 t:23.4s +tttg: c298/325 lr:0.000017 t:23.4s +tttg: c299/325 lr:0.000016 t:23.5s +tttg: c300/325 lr:0.000015 t:23.6s +tttg: c301/325 lr:0.000013 t:23.7s +tttg: c302/325 lr:0.000012 t:23.7s +tttg: c303/325 lr:0.000011 t:23.8s +tttg: c304/325 lr:0.000010 t:23.9s +tttg: c305/325 lr:0.000009 t:24.0s +tttg: c306/325 lr:0.000008 t:24.1s +tttg: c307/325 lr:0.000008 t:24.1s +tttg: c308/325 lr:0.000007 t:24.2s +tttg: c309/325 lr:0.000006 t:24.3s +tttg: c310/325 lr:0.000005 t:24.4s +tttg: c311/325 lr:0.000005 t:24.5s +tttg: c312/325 lr:0.000004 t:24.6s +tttg: c313/325 lr:0.000003 t:24.6s +tttg: c314/325 lr:0.000003 t:24.7s +tttg: c315/325 lr:0.000002 t:24.8s +tttg: c316/325 lr:0.000002 t:24.9s +tttg: c317/325 lr:0.000002 t:24.9s +tttg: c318/325 lr:0.000001 t:25.0s +tttg: c319/325 lr:0.000001 t:25.1s +tttg: c320/325 lr:0.000001 t:25.2s +tttg: c321/325 lr:0.000000 t:25.2s +tttg: c322/325 lr:0.000000 t:25.3s +tttg: c323/325 lr:0.000000 t:25.4s +tttg: c324/325 lr:0.000000 t:25.5s +ttpr: phase:3/3 t:346.9s +ttp: b723/782 bl:2.2903 bb:1.0282 rl:2.2923 rb:1.0604 dl:2185-2203 gd:1 +ttp: b718/782 bl:2.2872 bb:1.0265 rl:2.2921 rb:1.0588 dl:2089-2106 gd:1 +ttp: b711/782 bl:2.2803 bb:1.0197 rl:2.2916 rb:1.0571 dl:1966-1983 gd:1 +ttp: b697/782 bl:2.3217 bb:1.0301 rl:2.2927 rb:1.0561 dl:1790-1803 gd:1 +ttp: b692/782 bl:2.2885 bb:1.0273 rl:2.2925 rb:1.0550 dl:1737-1746 gd:1 +ttp: b683/782 bl:2.2654 bb:1.0545 rl:2.2917 rb:1.0550 dl:1646-1657 gd:1 +ttp: b676/782 bl:2.3288 bb:1.0475 rl:2.2928 rb:1.0548 dl:1586-1595 gd:1 +ttp: b666/782 bl:2.4095 bb:1.0636 rl:2.2960 rb:1.0550 dl:1507-1514 gd:1 +ttp: b661/782 bl:2.3972 bb:1.0837 rl:2.2986 rb:1.0558 dl:1474-1480 gd:1 +ttp: b654/782 bl:2.2886 bb:1.0351 rl:2.2984 rb:1.0553 dl:1425-1432 gd:1 +ttp: b644/782 bl:2.3567 bb:1.0463 rl:2.2997 rb:1.0551 dl:1362-1367 gd:1 +ttp: b637/782 bl:2.3608 bb:1.0766 rl:2.3010 rb:1.0555 dl:1320-1325 gd:1 +ttp: b629/782 bl:2.3450 bb:1.0091 rl:2.3019 rb:1.0545 dl:1276-1280 gd:1 +ttp: b621/782 bl:2.2952 bb:1.0481 rl:2.3018 rb:1.0544 dl:1231-1237 gd:1 +ttp: b612/782 bl:2.2329 bb:1.0116 rl:2.3005 rb:1.0536 dl:1186-1190 gd:1 +ttp: b605/782 bl:2.2424 bb:1.0226 rl:2.2995 rb:1.0531 dl:1154-1159 gd:1 +ttp: b593/782 bl:2.2880 bb:1.0100 rl:2.2993 rb:1.0523 dl:1103-1107 gd:1 +ttp: b585/782 bl:2.2775 bb:1.0330 rl:2.2990 rb:1.0520 dl:1069-1073 gd:1 +ttp: b576/782 bl:2.3723 bb:1.0912 rl:2.3001 rb:1.0526 dl:1033-1037 gd:1 +ttp: b569/782 bl:2.3015 bb:1.0407 rl:2.3001 rb:1.0524 dl:1007-1010 gd:1 +ttp: b562/782 bl:2.3010 bb:1.0306 rl:2.3001 rb:1.0521 dl:983-987 gd:1 +ttp: b552/782 bl:2.2698 bb:1.0168 rl:2.2997 rb:1.0516 dl:949-952 gd:1 +ttp: b544/782 bl:2.3420 bb:1.0672 rl:2.3002 rb:1.0518 dl:924-927 gd:1 +ttp: b535/782 bl:2.3731 bb:1.0292 rl:2.3011 rb:1.0516 dl:896-899 gd:1 +ttp: b527/782 bl:2.3402 bb:1.0271 rl:2.3016 rb:1.0513 dl:872-875 gd:1 +ttp: b519/782 bl:2.2900 bb:1.0389 rl:2.3015 rb:1.0511 dl:850-852 gd:1 +ttp: b511/782 bl:2.3818 bb:1.0478 rl:2.3023 rb:1.0511 dl:826-829 gd:1 +ttp: b503/782 bl:2.3454 bb:1.0626 rl:2.3028 rb:1.0512 dl:804-807 gd:1 +ttp: b495/782 bl:2.3041 bb:1.0292 rl:2.3028 rb:1.0510 dl:783-785 gd:1 +ttp: b487/782 bl:2.2745 bb:1.0650 rl:2.3025 rb:1.0511 dl:764-766 gd:1 +ttp: b480/782 bl:2.4286 bb:1.0813 rl:2.3037 rb:1.0514 dl:747-749 gd:1 +ttp: b472/782 bl:2.3787 bb:1.0797 rl:2.3044 rb:1.0517 dl:728-730 gd:1 +ttp: b464/782 bl:2.2602 bb:1.0129 rl:2.3040 rb:1.0513 dl:710-712 gd:1 +ttp: b456/782 bl:2.3445 bb:1.0385 rl:2.3043 rb:1.0512 dl:693-695 gd:1 +ttp: b448/782 bl:2.3094 bb:1.0068 rl:2.3044 rb:1.0508 dl:677-678 gd:1 +ttp: b438/782 bl:2.3063 bb:1.0525 rl:2.3044 rb:1.0508 dl:655-657 gd:1 +ttp: b430/782 bl:2.3774 bb:1.0396 rl:2.3050 rb:1.0507 dl:640-642 gd:1 +ttp: b422/782 bl:2.3037 bb:1.0871 rl:2.3049 rb:1.0510 dl:624-626 gd:1 +ttp: b413/782 bl:2.3653 bb:1.0601 rl:2.3054 rb:1.0511 dl:607-609 gd:1 +ttp: b405/782 bl:2.3529 bb:1.0559 rl:2.3057 rb:1.0511 dl:592-593 gd:1 +ttp: b396/782 bl:2.2815 bb:1.0732 rl:2.3055 rb:1.0512 dl:575-577 gd:1 +ttp: b390/782 bl:2.3441 bb:1.0561 rl:2.3058 rb:1.0513 dl:564-566 gd:1 +ttp: b382/782 bl:2.2901 bb:1.0820 rl:2.3057 rb:1.0515 dl:550-552 gd:1 +ttp: b374/782 bl:2.2954 bb:1.0348 rl:2.3056 rb:1.0514 dl:537-538 gd:1 +ttp: b366/782 bl:2.3325 bb:1.0685 rl:2.3058 rb:1.0515 dl:524-525 gd:1 +ttp: b359/782 bl:2.2492 bb:1.0328 rl:2.3055 rb:1.0514 dl:512-513 gd:1 +ttp: b350/782 bl:2.3207 bb:1.0547 rl:2.3056 rb:1.0514 dl:497-498 gd:1 +ttp: b342/782 bl:2.3640 bb:1.1183 rl:2.3059 rb:1.0517 dl:485-486 gd:1 +ttp: b334/782 bl:2.3799 bb:1.0698 rl:2.3063 rb:1.0518 dl:472-474 gd:1 +ttp: b326/782 bl:2.3080 bb:1.0569 rl:2.3063 rb:1.0518 dl:461-462 gd:1 +ttp: b318/782 bl:2.3362 bb:1.0676 rl:2.3064 rb:1.0519 dl:448-450 gd:1 +ttp: b309/782 bl:2.4096 bb:1.1057 rl:2.3069 rb:1.0522 dl:435-437 gd:1 +ttp: b300/782 bl:2.3309 bb:1.0529 rl:2.3070 rb:1.0522 dl:421-422 gd:1 +ttp: b294/782 bl:2.3018 bb:1.0752 rl:2.3070 rb:1.0523 dl:412-414 gd:1 +ttp: b287/782 bl:2.4005 bb:1.0936 rl:2.3074 rb:1.0525 dl:402-403 gd:1 +ttp: b280/782 bl:2.3323 bb:1.0874 rl:2.3075 rb:1.0526 dl:392-394 gd:1 +ttp: b273/782 bl:2.3261 bb:1.0723 rl:2.3076 rb:1.0527 dl:383-384 gd:1 +ttp: b267/782 bl:2.4147 bb:1.1412 rl:2.3080 rb:1.0530 dl:375-376 gd:1 +ttp: b261/782 bl:2.4213 bb:1.1145 rl:2.3084 rb:1.0533 dl:367-369 gd:1 +ttp: b252/782 bl:2.3817 bb:1.0675 rl:2.3087 rb:1.0533 dl:356-357 gd:1 +ttp: b243/782 bl:2.3514 bb:1.0790 rl:2.3089 rb:1.0534 dl:345-346 gd:1 +ttp: b234/782 bl:2.4125 bb:1.1432 rl:2.3092 rb:1.0537 dl:334-335 gd:1 +ttp: b226/782 bl:2.3536 bb:1.0916 rl:2.3094 rb:1.0538 dl:324-325 gd:1 +ttp: b218/782 bl:2.4564 bb:1.1079 rl:2.3099 rb:1.0540 dl:315-316 gd:1 +ttp: b210/782 bl:2.2520 bb:1.0798 rl:2.3097 rb:1.0541 dl:306-307 gd:1 +ttp: b202/782 bl:2.3534 bb:1.1015 rl:2.3098 rb:1.0542 dl:298-299 gd:1 +ttp: b194/782 bl:2.4286 bb:1.1126 rl:2.3102 rb:1.0544 dl:289-290 gd:1 +ttp: b185/782 bl:2.4276 bb:1.1131 rl:2.3105 rb:1.0546 dl:279-280 gd:1 +ttp: b178/782 bl:2.3389 bb:1.0941 rl:2.3106 rb:1.0547 dl:272-273 gd:1 +ttp: b170/782 bl:2.3758 bb:1.1266 rl:2.3108 rb:1.0549 dl:264-265 gd:1 +ttp: b162/782 bl:2.3976 bb:1.1163 rl:2.3110 rb:1.0550 dl:256-257 gd:1 +ttp: b154/782 bl:2.4658 bb:1.2026 rl:2.3114 rb:1.0554 dl:249-250 gd:1 +ttp: b146/782 bl:2.4532 bb:1.1721 rl:2.3117 rb:1.0557 dl:241-242 gd:1 +ttp: b137/782 bl:2.4056 bb:1.1493 rl:2.3120 rb:1.0559 dl:233-233 gd:1 +ttp: b128/782 bl:2.3900 bb:1.1551 rl:2.3121 rb:1.0561 dl:224-225 gd:1 +ttp: b120/782 bl:2.3852 bb:1.1083 rl:2.3123 rb:1.0562 dl:217-218 gd:1 +ttp: b111/782 bl:2.4011 bb:1.1708 rl:2.3125 rb:1.0564 dl:208-210 gd:1 +ttp: b104/782 bl:2.4888 bb:1.1749 rl:2.3128 rb:1.0567 dl:202-203 gd:1 +ttp: b96/782 bl:2.4654 bb:1.1969 rl:2.3131 rb:1.0569 dl:195-196 gd:1 +ttp: b88/782 bl:2.4752 bb:1.1812 rl:2.3134 rb:1.0571 dl:188-189 gd:1 +ttp: b79/782 bl:2.3890 bb:1.1421 rl:2.3136 rb:1.0573 dl:180-181 gd:1 +ttp: b70/782 bl:2.5118 bb:1.2241 rl:2.3139 rb:1.0575 dl:172-173 gd:1 +ttp: b64/782 bl:2.5266 bb:1.1524 rl:2.3143 rb:1.0577 dl:166-167 gd:1 +ttp: b56/782 bl:2.5450 bb:1.2199 rl:2.3146 rb:1.0579 dl:159-160 gd:1 +ttp: b48/782 bl:2.5119 bb:1.2110 rl:2.3149 rb:1.0582 dl:151-152 gd:1 +ttp: b40/782 bl:2.4848 bb:1.1511 rl:2.3152 rb:1.0583 dl:143-144 gd:1 +ttp: b32/782 bl:2.6024 bb:1.2134 rl:2.3156 rb:1.0585 dl:135-136 gd:1 +ttp: b24/782 bl:2.4484 bb:1.1547 rl:2.3157 rb:1.0586 dl:127-128 gd:1 +ttp: b16/782 bl:2.6205 bb:1.2556 rl:2.3161 rb:1.0588 dl:117-118 gd:1 +ttp: b8/782 bl:2.7871 bb:1.2937 rl:2.3166 rb:1.0591 dl:103-105 gd:1 +quantized_ttt_phased val_loss:2.31627197 val_bpb:1.05844590 eval_time:425873ms +total_eval_time:425.9s diff --git a/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed1234.log b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed1234.log new file mode 100644 index 0000000000..d31e8105bf --- /dev/null +++ b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed1234.log @@ -0,0 +1,1040 @@ +W0430 06:01:38.924000 335014 torch/distributed/run.py:803] +W0430 06:01:38.924000 335014 torch/distributed/run.py:803] ***************************************** +W0430 06:01:38.924000 335014 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. +W0430 06:01:38.924000 335014 torch/distributed/run.py:803] ***************************************** +Hyperparameters: + adam_eps: 1e-08 + adam_wd: 0.02 + artifact_dir: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed1234 + attn_clip_sigmas: 13.0 + attn_out_gate_enabled: False + attn_out_gate_src: proj + awq_lite_bits: 8 + awq_lite_enabled: True + awq_lite_group_size: 64 + awq_lite_group_top_k: 1 + beta1: 0.9 + beta2: 0.99 + caseops_enabled: True + compressor: pergroup + data_dir: ./data/ + datasets_dir: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved + distributed: True + ema_decay: 0.9965 + embed_bits: 7 + embed_clip_sigmas: 14.0 + embed_lr: 0.6 + embed_wd: 0.085 + enable_looping_at: 0.35 + eval_seq_len: 2560 + eval_stride: 64 + fused_ce_enabled: True + gate_window: 12 + gated_attn_enabled: False + gated_attn_init_std: 0.01 + gated_attn_quant_gate: True + global_ttt_batch_seqs: 32 + global_ttt_chunk_tokens: 32768 + global_ttt_epochs: 1 + global_ttt_grad_clip: 1.0 + global_ttt_lr: 0.001 + global_ttt_momentum: 0.9 + global_ttt_respect_doc_boundaries: True + global_ttt_warmup_chunks: 0 + global_ttt_warmup_start_lr: 0.0 + gptq_calibration_batches: 16 + gptq_reserve_seconds: 4.0 + grad_accum_steps: 1 + grad_clip_norm: 0.3 + is_main_process: True + iterations: 20000 + ln_scale: True + local_rank: 0 + logfile: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed1234/rank56_prefix3000_seed1234.txt + logit_softcap: 30.0 + loop_end: 5 + loop_start: 3 + lqer_asym_enabled: True + lqer_asym_group: 64 + lqer_enabled: True + lqer_factor_bits: 4 + lqer_gain_select: False + lqer_rank: 4 + lqer_scope: all + lqer_top_k: 3 + matrix_bits: 6 + matrix_clip_sigmas: 12.85 + matrix_lr: 0.026 + max_wallclock_seconds: 600.0 + min_lr: 0.1 + mlp_clip_sigmas: 11.5 + mlp_mult: 4.0 + model_dim: 512 + model_path: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed1234/final_model.pt + muon_backend_steps: 5 + muon_momentum: 0.97 + muon_momentum_warmup_start: 0.92 + muon_momentum_warmup_steps: 1500 + muon_row_normalize: True + muon_wd: 0.095 + num_heads: 8 + num_kv_heads: 4 + num_layers: 11 + num_loops: 2 + parallel_final_lane: mean + parallel_start_layer: 8 + phased_ttt_num_phases: 3 + phased_ttt_prefix_docs: 3000 + qk_gain_init: 5.25 + quantized_model_path: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed1234/final_model.int6.ptz + rank: 0 + rope_base: 10000.0 + rope_dims: 16 + rope_train_seq_len: 2048 + rope_yarn: False + run_id: rank56_prefix3000_seed1234 + scalar_lr: 0.02 + seed: 1234 + skip_gates_enabled: True + smear_gate_enabled: True + sparse_attn_gate_enabled: True + sparse_attn_gate_init_std: 0.0 + sparse_attn_gate_scale: 0.5 + tie_embeddings: True + tied_embed_init_std: 0.005 + tied_embed_lr: 0.03 + tokenizer_path: ./data/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model + train_batch_tokens: 786432 + train_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_train_*.bin + train_log_every: 500 + train_seq_len: 2048 + ttt_batch_size: 64 + ttt_beta1: 0.0 + ttt_beta2: 0.99 + ttt_chunk_size: 48 + ttt_enabled: True + ttt_eval_batches: + ttt_eval_seq_len: 2560 + ttt_grad_steps: 1 + ttt_k_lora: True + ttt_local_lr_mult: 0.75 + ttt_lora_lr: 0.0001 + ttt_lora_rank: 56 + ttt_mask: no_qv + ttt_mlp_lora: True + ttt_o_lora: True + ttt_optimizer: adam + ttt_q_lora: False + ttt_v_lora: False + ttt_weight_decay: 0.5 + val_batch_tokens: 524288 + val_bytes_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_val_bytes_*.bin + val_doc_fraction: 1.0 + val_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_val_*.bin + val_loss_every: 0 + vocab_size: 8192 + warmdown_frac: 0.85 + warmup_steps: 20 + world_size: 8 + xsa_last_n: 11 +train_shards: 80 +val_tokens: 47851520 +model_params:35945673 +gptq:reserving 4s, effective=596000ms +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 +1/20000 train_loss: 9.0017 train_time: 0.0m tok/s: 17959522 +2/20000 train_loss: 12.9407 train_time: 0.0m tok/s: 11253963 +3/20000 train_loss: 10.2332 train_time: 0.0m tok/s: 10160787 +4/20000 train_loss: 8.7483 train_time: 0.0m tok/s: 9656973 +5/20000 train_loss: 7.9481 train_time: 0.0m tok/s: 9366548 +500/20000 train_loss: 2.5705 train_time: 0.8m tok/s: 8203424 +1000/20000 train_loss: 2.8064 train_time: 1.6m tok/s: 8163289 +1500/20000 train_loss: 2.6219 train_time: 2.4m tok/s: 8150277 +2000/20000 train_loss: 2.6603 train_time: 3.2m tok/s: 8145208 +layer_loop:enabled step:2160 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: 2.5430 train_time: 4.3m tok/s: 7661430 +3000/20000 train_loss: 2.5577 train_time: 5.5m tok/s: 7210638 +3500/20000 train_loss: 2.5606 train_time: 6.6m tok/s: 6916574 +4000/20000 train_loss: 2.4024 train_time: 7.8m tok/s: 6714805 +4500/20000 train_loss: 2.2727 train_time: 9.0m tok/s: 6549943 +4880/20000 val_loss: 2.3518 val_bpb: 1.0746 +stopping_early: wallclock_cap train_time: 596146ms step: 4880/20000 +peak memory allocated: 41707 MiB reserved: 46952 MiB +ema:applying EMA weights +diagnostic pre-quantization post-ema val_loss:2.32685853 val_bpb:1.06319088 eval_time:10770ms +Serialized model: 135418111 bytes +Code size (uncompressed): 171669 bytes +Code size (compressed): 34125 bytes +GPTQ:collecting Hessians from calibration data... +GPTQ:collected 67 Hessians in 4.1s +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 (int6)+lqer_asym: blocks.mlp.fc.weight + gptq (int7)+awqgrpint8+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, softcap_neg, softcap_pos +Serialize: per-group lrzip compression... +Serialize: per-group compression done in 120.8s +Serialized model quantized+pergroup: 15945985 bytes +Total submission size quantized+pergroup: 15980110 bytes +Deserialize: per-group lrzip decompression... +Deserialize: decompression done in 20.9s +diagnostic quantized val_loss:2.34562535 val_bpb:1.07176584 eval_time:11618ms +Deserialize: per-group lrzip decompression... +Deserialize: decompression done in 20.7s +ttt_lora:warming up compile (random tokens, no val data) +ttt_lora:compile warmup done (91.7s) + +beginning TTT eval timer +ttt_phased: total_docs:50000 prefix_docs:3000 suffix_docs:47000 num_phases:3 boundaries:[1000, 2000, 3000] +ttp: b776/782 bl:2.2543 bb:1.0688 rl:2.2543 rb:1.0688 dl:7534-8350 gd:0 +ttp: b773/782 bl:2.1969 bb:1.0346 rl:2.2289 rb:1.0536 dl:6104-6447 gd:0 +ttp: b768/782 bl:2.2361 bb:1.0414 rl:2.2307 rb:1.0504 dl:4859-5083 gd:0 +ttp: b764/782 bl:2.2893 bb:1.0723 rl:2.2415 rb:1.0544 dl:4284-4392 gd:0 +ttpp: phase:1/3 pd:1424 gd:1000 t:187.5s +tttg: c1/154 lr:0.001000 t:0.3s +tttg: c2/154 lr:0.001000 t:0.4s +tttg: c3/154 lr:0.001000 t:0.5s +tttg: c4/154 lr:0.000999 t:0.6s +tttg: c5/154 lr:0.000998 t:0.6s +tttg: c6/154 lr:0.000997 t:0.7s +tttg: c7/154 lr:0.000996 t:0.8s +tttg: c8/154 lr:0.000995 t:0.9s +tttg: c9/154 lr:0.000993 t:0.9s +tttg: c10/154 lr:0.000991 t:1.0s +tttg: c11/154 lr:0.000989 t:1.1s +tttg: c12/154 lr:0.000987 t:1.2s +tttg: c13/154 lr:0.000985 t:1.2s +tttg: c14/154 lr:0.000982 t:1.3s +tttg: c15/154 lr:0.000979 t:1.4s +tttg: c16/154 lr:0.000976 t:1.5s +tttg: c17/154 lr:0.000973 t:1.6s +tttg: c18/154 lr:0.000970 t:1.6s +tttg: c19/154 lr:0.000966 t:1.7s +tttg: c20/154 lr:0.000962 t:1.8s +tttg: c21/154 lr:0.000958 t:1.9s +tttg: c22/154 lr:0.000954 t:1.9s +tttg: c23/154 lr:0.000950 t:2.0s +tttg: c24/154 lr:0.000945 t:2.1s +tttg: c25/154 lr:0.000941 t:2.2s +tttg: c26/154 lr:0.000936 t:2.2s +tttg: c27/154 lr:0.000930 t:2.3s +tttg: c28/154 lr:0.000925 t:2.4s +tttg: c29/154 lr:0.000920 t:2.5s +tttg: c30/154 lr:0.000914 t:2.6s +tttg: c31/154 lr:0.000908 t:2.6s +tttg: c32/154 lr:0.000902 t:2.7s +tttg: c33/154 lr:0.000896 t:2.8s +tttg: c34/154 lr:0.000890 t:2.9s +tttg: c35/154 lr:0.000883 t:2.9s +tttg: c36/154 lr:0.000876 t:3.0s +tttg: c37/154 lr:0.000870 t:3.1s +tttg: c38/154 lr:0.000863 t:3.2s +tttg: c39/154 lr:0.000855 t:3.2s +tttg: c40/154 lr:0.000848 t:3.3s +tttg: c41/154 lr:0.000841 t:3.4s +tttg: c42/154 lr:0.000833 t:3.5s +tttg: c43/154 lr:0.000825 t:3.5s +tttg: c44/154 lr:0.000817 t:3.6s +tttg: c45/154 lr:0.000809 t:3.7s +tttg: c46/154 lr:0.000801 t:3.8s +tttg: c47/154 lr:0.000793 t:3.9s +tttg: c48/154 lr:0.000785 t:3.9s +tttg: c49/154 lr:0.000776 t:4.0s +tttg: c50/154 lr:0.000768 t:4.1s +tttg: c51/154 lr:0.000759 t:4.2s +tttg: c52/154 lr:0.000750 t:4.3s +tttg: c53/154 lr:0.000741 t:4.3s +tttg: c54/154 lr:0.000732 t:4.4s +tttg: c55/154 lr:0.000723 t:4.5s +tttg: c56/154 lr:0.000714 t:4.6s +tttg: c57/154 lr:0.000704 t:4.6s +tttg: c58/154 lr:0.000695 t:4.7s +tttg: c59/154 lr:0.000685 t:4.8s +tttg: c60/154 lr:0.000676 t:4.9s +tttg: c61/154 lr:0.000666 t:4.9s +tttg: c62/154 lr:0.000656 t:5.0s +tttg: c63/154 lr:0.000647 t:5.1s +tttg: c64/154 lr:0.000637 t:5.2s +tttg: c65/154 lr:0.000627 t:5.2s +tttg: c66/154 lr:0.000617 t:5.3s +tttg: c67/154 lr:0.000607 t:5.4s +tttg: c68/154 lr:0.000597 t:5.5s +tttg: c69/154 lr:0.000587 t:5.6s +tttg: c70/154 lr:0.000577 t:5.6s +tttg: c71/154 lr:0.000567 t:5.7s +tttg: c72/154 lr:0.000556 t:5.8s +tttg: c73/154 lr:0.000546 t:5.9s +tttg: c74/154 lr:0.000536 t:5.9s +tttg: c75/154 lr:0.000526 t:6.0s +tttg: c76/154 lr:0.000515 t:6.1s +tttg: c77/154 lr:0.000505 t:6.2s +tttg: c78/154 lr:0.000495 t:6.2s +tttg: c79/154 lr:0.000485 t:6.3s +tttg: c80/154 lr:0.000474 t:6.4s +tttg: c81/154 lr:0.000464 t:6.5s +tttg: c82/154 lr:0.000454 t:6.5s +tttg: c83/154 lr:0.000444 t:6.6s +tttg: c84/154 lr:0.000433 t:6.7s +tttg: c85/154 lr:0.000423 t:6.8s +tttg: c86/154 lr:0.000413 t:6.8s +tttg: c87/154 lr:0.000403 t:6.9s +tttg: c88/154 lr:0.000393 t:7.0s +tttg: c89/154 lr:0.000383 t:7.1s +tttg: c90/154 lr:0.000373 t:7.2s +tttg: c91/154 lr:0.000363 t:7.2s +tttg: c92/154 lr:0.000353 t:7.3s +tttg: c93/154 lr:0.000344 t:7.4s +tttg: c94/154 lr:0.000334 t:7.5s +tttg: c95/154 lr:0.000324 t:7.5s +tttg: c96/154 lr:0.000315 t:7.6s +tttg: c97/154 lr:0.000305 t:7.7s +tttg: c98/154 lr:0.000296 t:7.8s +tttg: c99/154 lr:0.000286 t:7.9s +tttg: c100/154 lr:0.000277 t:7.9s +tttg: c101/154 lr:0.000268 t:8.0s +tttg: c102/154 lr:0.000259 t:8.1s +tttg: c103/154 lr:0.000250 t:8.2s +tttg: c104/154 lr:0.000241 t:8.2s +tttg: c105/154 lr:0.000232 t:8.3s +tttg: c106/154 lr:0.000224 t:8.4s +tttg: c107/154 lr:0.000215 t:8.5s +tttg: c108/154 lr:0.000207 t:8.5s +tttg: c109/154 lr:0.000199 t:8.6s +tttg: c110/154 lr:0.000191 t:8.7s +tttg: c111/154 lr:0.000183 t:8.8s +tttg: c112/154 lr:0.000175 t:8.9s +tttg: c113/154 lr:0.000167 t:8.9s +tttg: c114/154 lr:0.000159 t:9.0s +tttg: c115/154 lr:0.000152 t:9.1s +tttg: c116/154 lr:0.000145 t:9.2s +tttg: c117/154 lr:0.000137 t:9.2s +tttg: c118/154 lr:0.000130 t:9.3s +tttg: c119/154 lr:0.000124 t:9.4s +tttg: c120/154 lr:0.000117 t:9.5s +tttg: c121/154 lr:0.000110 t:9.5s +tttg: c122/154 lr:0.000104 t:9.6s +tttg: c123/154 lr:0.000098 t:9.7s +tttg: c124/154 lr:0.000092 t:9.8s +tttg: c125/154 lr:0.000086 t:9.9s +tttg: c126/154 lr:0.000080 t:9.9s +tttg: c127/154 lr:0.000075 t:10.0s +tttg: c128/154 lr:0.000070 t:10.1s +tttg: c129/154 lr:0.000064 t:10.2s +tttg: c130/154 lr:0.000059 t:10.2s +tttg: c131/154 lr:0.000055 t:10.3s +tttg: c132/154 lr:0.000050 t:10.4s +tttg: c133/154 lr:0.000046 t:10.5s +tttg: c134/154 lr:0.000042 t:10.5s +tttg: c135/154 lr:0.000038 t:10.6s +tttg: c136/154 lr:0.000034 t:10.7s +tttg: c137/154 lr:0.000030 t:10.8s +tttg: c138/154 lr:0.000027 t:10.9s +tttg: c139/154 lr:0.000024 t:10.9s +tttg: c140/154 lr:0.000021 t:11.0s +tttg: c141/154 lr:0.000018 t:11.1s +tttg: c142/154 lr:0.000015 t:11.2s +tttg: c143/154 lr:0.000013 t:11.2s +tttg: c144/154 lr:0.000011 t:11.3s +tttg: c145/154 lr:0.000009 t:11.4s +tttg: c146/154 lr:0.000007 t:11.5s +tttg: c147/154 lr:0.000005 t:11.6s +tttg: c148/154 lr:0.000004 t:11.6s +tttg: c149/154 lr:0.000003 t:11.7s +tttg: c150/154 lr:0.000002 t:11.8s +tttg: c151/154 lr:0.000001 t:11.9s +tttg: c152/154 lr:0.000000 t:11.9s +tttg: c153/154 lr:0.000000 t:12.0s +ttpr: phase:1/3 t:200.2s +ttp: b758/782 bl:2.3035 bb:1.0736 rl:2.2499 rb:1.0571 dl:3634-3740 gd:0 +ttp: b749/782 bl:2.3899 bb:1.0845 rl:2.2641 rb:1.0599 dl:3039-3089 gd:0 +ttp: b746/782 bl:2.4092 bb:1.0616 rl:2.2769 rb:1.0601 dl:2884-2943 gd:0 +ttp: b745/782 bl:2.2330 bb:1.0223 rl:2.2734 rb:1.0570 dl:2842-2883 gd:0 +ttpp: phase:2/3 pd:2448 gd:2000 t:259.5s +tttg: c1/250 lr:0.001000 t:0.1s +tttg: c2/250 lr:0.001000 t:0.2s +tttg: c3/250 lr:0.001000 t:0.2s +tttg: c4/250 lr:0.001000 t:0.3s +tttg: c5/250 lr:0.000999 t:0.4s +tttg: c6/250 lr:0.000999 t:0.5s +tttg: c7/250 lr:0.000999 t:0.5s +tttg: c8/250 lr:0.000998 t:0.6s +tttg: c9/250 lr:0.000997 t:0.7s +tttg: c10/250 lr:0.000997 t:0.8s +tttg: c11/250 lr:0.000996 t:0.8s +tttg: c12/250 lr:0.000995 t:0.9s +tttg: c13/250 lr:0.000994 t:1.0s +tttg: c14/250 lr:0.000993 t:1.1s +tttg: c15/250 lr:0.000992 t:1.2s +tttg: c16/250 lr:0.000991 t:1.2s +tttg: c17/250 lr:0.000990 t:1.3s +tttg: c18/250 lr:0.000989 t:1.4s +tttg: c19/250 lr:0.000987 t:1.5s +tttg: c20/250 lr:0.000986 t:1.5s +tttg: c21/250 lr:0.000984 t:1.6s +tttg: c22/250 lr:0.000983 t:1.7s +tttg: c23/250 lr:0.000981 t:1.8s +tttg: c24/250 lr:0.000979 t:1.8s +tttg: c25/250 lr:0.000977 t:1.9s +tttg: c26/250 lr:0.000975 t:2.0s +tttg: c27/250 lr:0.000973 t:2.1s +tttg: c28/250 lr:0.000971 t:2.1s +tttg: c29/250 lr:0.000969 t:2.2s +tttg: c30/250 lr:0.000967 t:2.3s +tttg: c31/250 lr:0.000965 t:2.4s +tttg: c32/250 lr:0.000962 t:2.5s +tttg: c33/250 lr:0.000960 t:2.5s +tttg: c34/250 lr:0.000957 t:2.6s +tttg: c35/250 lr:0.000955 t:2.7s +tttg: c36/250 lr:0.000952 t:2.8s +tttg: c37/250 lr:0.000949 t:2.8s +tttg: c38/250 lr:0.000947 t:2.9s +tttg: c39/250 lr:0.000944 t:3.0s +tttg: c40/250 lr:0.000941 t:3.1s +tttg: c41/250 lr:0.000938 t:3.2s +tttg: c42/250 lr:0.000935 t:3.2s +tttg: c43/250 lr:0.000931 t:3.3s +tttg: c44/250 lr:0.000928 t:3.4s +tttg: c45/250 lr:0.000925 t:3.5s +tttg: c46/250 lr:0.000922 t:3.6s +tttg: c47/250 lr:0.000918 t:3.6s +tttg: c48/250 lr:0.000915 t:3.7s +tttg: c49/250 lr:0.000911 t:3.8s +tttg: c50/250 lr:0.000907 t:3.9s +tttg: c51/250 lr:0.000904 t:3.9s +tttg: c52/250 lr:0.000900 t:4.0s +tttg: c53/250 lr:0.000896 t:4.1s +tttg: c54/250 lr:0.000892 t:4.2s +tttg: c55/250 lr:0.000888 t:4.3s +tttg: c56/250 lr:0.000884 t:4.3s +tttg: c57/250 lr:0.000880 t:4.4s +tttg: c58/250 lr:0.000876 t:4.5s +tttg: c59/250 lr:0.000872 t:4.6s +tttg: c60/250 lr:0.000868 t:4.6s +tttg: c61/250 lr:0.000863 t:4.7s +tttg: c62/250 lr:0.000859 t:4.8s +tttg: c63/250 lr:0.000855 t:4.9s +tttg: c64/250 lr:0.000850 t:5.0s +tttg: c65/250 lr:0.000846 t:5.0s +tttg: c66/250 lr:0.000841 t:5.1s +tttg: c67/250 lr:0.000836 t:5.2s +tttg: c68/250 lr:0.000832 t:5.3s +tttg: c69/250 lr:0.000827 t:5.3s +tttg: c70/250 lr:0.000822 t:5.4s +tttg: c71/250 lr:0.000817 t:5.5s +tttg: c72/250 lr:0.000812 t:5.6s +tttg: c73/250 lr:0.000807 t:5.6s +tttg: c74/250 lr:0.000803 t:5.7s +tttg: c75/250 lr:0.000797 t:5.8s +tttg: c76/250 lr:0.000792 t:5.9s +tttg: c77/250 lr:0.000787 t:6.0s +tttg: c78/250 lr:0.000782 t:6.0s +tttg: c79/250 lr:0.000777 t:6.1s +tttg: c80/250 lr:0.000772 t:6.2s +tttg: c81/250 lr:0.000766 t:6.3s +tttg: c82/250 lr:0.000761 t:6.3s +tttg: c83/250 lr:0.000755 t:6.4s +tttg: c84/250 lr:0.000750 t:6.5s +tttg: c85/250 lr:0.000745 t:6.6s +tttg: c86/250 lr:0.000739 t:6.6s +tttg: c87/250 lr:0.000733 t:6.7s +tttg: c88/250 lr:0.000728 t:6.8s +tttg: c89/250 lr:0.000722 t:6.9s +tttg: c90/250 lr:0.000717 t:7.0s +tttg: c91/250 lr:0.000711 t:7.0s +tttg: c92/250 lr:0.000705 t:7.1s +tttg: c93/250 lr:0.000699 t:7.2s +tttg: c94/250 lr:0.000694 t:7.3s +tttg: c95/250 lr:0.000688 t:7.3s +tttg: c96/250 lr:0.000682 t:7.4s +tttg: c97/250 lr:0.000676 t:7.5s +tttg: c98/250 lr:0.000670 t:7.6s +tttg: c99/250 lr:0.000664 t:7.6s +tttg: c100/250 lr:0.000658 t:7.7s +tttg: c101/250 lr:0.000652 t:7.8s +tttg: c102/250 lr:0.000646 t:7.9s +tttg: c103/250 lr:0.000640 t:8.0s +tttg: c104/250 lr:0.000634 t:8.0s +tttg: c105/250 lr:0.000628 t:8.1s +tttg: c106/250 lr:0.000622 t:8.2s +tttg: c107/250 lr:0.000616 t:8.3s +tttg: c108/250 lr:0.000610 t:8.4s +tttg: c109/250 lr:0.000603 t:8.4s +tttg: c110/250 lr:0.000597 t:8.5s +tttg: c111/250 lr:0.000591 t:8.6s +tttg: c112/250 lr:0.000585 t:8.7s +tttg: c113/250 lr:0.000579 t:8.7s +tttg: c114/250 lr:0.000572 t:8.8s +tttg: c115/250 lr:0.000566 t:8.9s +tttg: c116/250 lr:0.000560 t:9.0s +tttg: c117/250 lr:0.000554 t:9.0s +tttg: c118/250 lr:0.000547 t:9.1s +tttg: c119/250 lr:0.000541 t:9.2s +tttg: c120/250 lr:0.000535 t:9.3s +tttg: c121/250 lr:0.000528 t:9.4s +tttg: c122/250 lr:0.000522 t:9.4s +tttg: c123/250 lr:0.000516 t:9.5s +tttg: c124/250 lr:0.000509 t:9.6s +tttg: c125/250 lr:0.000503 t:9.7s +tttg: c126/250 lr:0.000497 t:9.8s +tttg: c127/250 lr:0.000491 t:9.8s +tttg: c128/250 lr:0.000484 t:9.9s +tttg: c129/250 lr:0.000478 t:10.0s +tttg: c130/250 lr:0.000472 t:10.1s +tttg: c131/250 lr:0.000465 t:10.1s +tttg: c132/250 lr:0.000459 t:10.2s +tttg: c133/250 lr:0.000453 t:10.3s +tttg: c134/250 lr:0.000446 t:10.4s +tttg: c135/250 lr:0.000440 t:10.4s +tttg: c136/250 lr:0.000434 t:10.5s +tttg: c137/250 lr:0.000428 t:10.6s +tttg: c138/250 lr:0.000421 t:10.7s +tttg: c139/250 lr:0.000415 t:10.8s +tttg: c140/250 lr:0.000409 t:10.8s +tttg: c141/250 lr:0.000403 t:10.9s +tttg: c142/250 lr:0.000397 t:11.0s +tttg: c143/250 lr:0.000390 t:11.1s +tttg: c144/250 lr:0.000384 t:11.1s +tttg: c145/250 lr:0.000378 t:11.2s +tttg: c146/250 lr:0.000372 t:11.3s +tttg: c147/250 lr:0.000366 t:11.4s +tttg: c148/250 lr:0.000360 t:11.5s +tttg: c149/250 lr:0.000354 t:11.5s +tttg: c150/250 lr:0.000348 t:11.6s +tttg: c151/250 lr:0.000342 t:11.7s +tttg: c152/250 lr:0.000336 t:11.8s +tttg: c153/250 lr:0.000330 t:11.8s +tttg: c154/250 lr:0.000324 t:11.9s +tttg: c155/250 lr:0.000318 t:12.0s +tttg: c156/250 lr:0.000312 t:12.1s +tttg: c157/250 lr:0.000306 t:12.2s +tttg: c158/250 lr:0.000301 t:12.2s +tttg: c159/250 lr:0.000295 t:12.3s +tttg: c160/250 lr:0.000289 t:12.4s +tttg: c161/250 lr:0.000283 t:12.5s +tttg: c162/250 lr:0.000278 t:12.5s +tttg: c163/250 lr:0.000272 t:12.6s +tttg: c164/250 lr:0.000267 t:12.7s +tttg: c165/250 lr:0.000261 t:12.8s +tttg: c166/250 lr:0.000255 t:12.8s +tttg: c167/250 lr:0.000250 t:12.9s +tttg: c168/250 lr:0.000245 t:13.0s +tttg: c169/250 lr:0.000239 t:13.1s +tttg: c170/250 lr:0.000234 t:13.2s +tttg: c171/250 lr:0.000228 t:13.2s +tttg: c172/250 lr:0.000223 t:13.3s +tttg: c173/250 lr:0.000218 t:13.4s +tttg: c174/250 lr:0.000213 t:13.5s +tttg: c175/250 lr:0.000208 t:13.5s +tttg: c176/250 lr:0.000203 t:13.6s +tttg: c177/250 lr:0.000197 t:13.7s +tttg: c178/250 lr:0.000193 t:13.8s +tttg: c179/250 lr:0.000188 t:13.8s +tttg: c180/250 lr:0.000183 t:13.9s +tttg: c181/250 lr:0.000178 t:14.0s +tttg: c182/250 lr:0.000173 t:14.1s +tttg: c183/250 lr:0.000168 t:14.2s +tttg: c184/250 lr:0.000164 t:14.2s +tttg: c185/250 lr:0.000159 t:14.3s +tttg: c186/250 lr:0.000154 t:14.4s +tttg: c187/250 lr:0.000150 t:14.5s +tttg: c188/250 lr:0.000145 t:14.5s +tttg: c189/250 lr:0.000141 t:14.6s +tttg: c190/250 lr:0.000137 t:14.7s +tttg: c191/250 lr:0.000132 t:14.8s +tttg: c192/250 lr:0.000128 t:14.8s +tttg: c193/250 lr:0.000124 t:14.9s +tttg: c194/250 lr:0.000120 t:15.0s +tttg: c195/250 lr:0.000116 t:15.1s +tttg: c196/250 lr:0.000112 t:15.1s +tttg: c197/250 lr:0.000108 t:15.2s +tttg: c198/250 lr:0.000104 t:15.3s +tttg: c199/250 lr:0.000100 t:15.4s +tttg: c200/250 lr:0.000096 t:15.5s +tttg: c201/250 lr:0.000093 t:15.5s +tttg: c202/250 lr:0.000089 t:15.6s +tttg: c203/250 lr:0.000085 t:15.7s +tttg: c204/250 lr:0.000082 t:15.8s +tttg: c205/250 lr:0.000078 t:15.8s +tttg: c206/250 lr:0.000075 t:15.9s +tttg: c207/250 lr:0.000072 t:16.0s +tttg: c208/250 lr:0.000069 t:16.1s +tttg: c209/250 lr:0.000065 t:16.2s +tttg: c210/250 lr:0.000062 t:16.2s +tttg: c211/250 lr:0.000059 t:16.3s +tttg: c212/250 lr:0.000056 t:16.4s +tttg: c213/250 lr:0.000053 t:16.5s +tttg: c214/250 lr:0.000051 t:16.5s +tttg: c215/250 lr:0.000048 t:16.6s +tttg: c216/250 lr:0.000045 t:16.7s +tttg: c217/250 lr:0.000043 t:16.8s +tttg: c218/250 lr:0.000040 t:16.9s +tttg: c219/250 lr:0.000038 t:17.0s +tttg: c220/250 lr:0.000035 t:17.0s +tttg: c221/250 lr:0.000033 t:17.1s +tttg: c222/250 lr:0.000031 t:17.2s +tttg: c223/250 lr:0.000029 t:17.3s +tttg: c224/250 lr:0.000027 t:17.4s +tttg: c225/250 lr:0.000025 t:17.4s +tttg: c226/250 lr:0.000023 t:17.5s +tttg: c227/250 lr:0.000021 t:17.6s +tttg: c228/250 lr:0.000019 t:17.7s +tttg: c229/250 lr:0.000017 t:17.7s +tttg: c230/250 lr:0.000016 t:17.8s +tttg: c231/250 lr:0.000014 t:17.9s +tttg: c232/250 lr:0.000013 t:18.0s +tttg: c233/250 lr:0.000011 t:18.0s +tttg: c234/250 lr:0.000010 t:18.1s +tttg: c235/250 lr:0.000009 t:18.2s +tttg: c236/250 lr:0.000008 t:18.3s +tttg: c237/250 lr:0.000007 t:18.4s +tttg: c238/250 lr:0.000006 t:18.4s +tttg: c239/250 lr:0.000005 t:18.5s +tttg: c240/250 lr:0.000004 t:18.6s +tttg: c241/250 lr:0.000003 t:18.7s +tttg: c242/250 lr:0.000003 t:18.7s +tttg: c243/250 lr:0.000002 t:18.8s +tttg: c244/250 lr:0.000001 t:18.9s +tttg: c245/250 lr:0.000001 t:19.0s +tttg: c246/250 lr:0.000001 t:19.1s +tttg: c247/250 lr:0.000000 t:19.1s +tttg: c248/250 lr:0.000000 t:19.2s +tttg: c249/250 lr:0.000000 t:19.3s +ttpr: phase:2/3 t:279.6s +ttp: b738/782 bl:2.3103 bb:1.0462 rl:2.2759 rb:1.0563 dl:2583-2618 gd:0 +ttp: b733/782 bl:2.3770 bb:1.0643 rl:2.2819 rb:1.0568 dl:2441-2468 gd:0 +ttpp: phase:3/3 pd:3472 gd:3000 t:292.4s +tttg: c1/325 lr:0.001000 t:0.1s +tttg: c2/325 lr:0.001000 t:0.2s +tttg: c3/325 lr:0.001000 t:0.2s +tttg: c4/325 lr:0.001000 t:0.3s +tttg: c5/325 lr:0.001000 t:0.4s +tttg: c6/325 lr:0.000999 t:0.5s +tttg: c7/325 lr:0.000999 t:0.5s +tttg: c8/325 lr:0.000999 t:0.6s +tttg: c9/325 lr:0.000998 t:0.7s +tttg: c10/325 lr:0.000998 t:0.8s +tttg: c11/325 lr:0.000998 t:0.8s +tttg: c12/325 lr:0.000997 t:0.9s +tttg: c13/325 lr:0.000997 t:1.0s +tttg: c14/325 lr:0.000996 t:1.1s +tttg: c15/325 lr:0.000995 t:1.2s +tttg: c16/325 lr:0.000995 t:1.2s +tttg: c17/325 lr:0.000994 t:1.3s +tttg: c18/325 lr:0.000993 t:1.4s +tttg: c19/325 lr:0.000992 t:1.5s +tttg: c20/325 lr:0.000992 t:1.5s +tttg: c21/325 lr:0.000991 t:1.6s +tttg: c22/325 lr:0.000990 t:1.7s +tttg: c23/325 lr:0.000989 t:1.8s +tttg: c24/325 lr:0.000988 t:1.8s +tttg: c25/325 lr:0.000987 t:1.9s +tttg: c26/325 lr:0.000985 t:2.0s +tttg: c27/325 lr:0.000984 t:2.1s +tttg: c28/325 lr:0.000983 t:2.2s +tttg: c29/325 lr:0.000982 t:2.2s +tttg: c30/325 lr:0.000980 t:2.3s +tttg: c31/325 lr:0.000979 t:2.4s +tttg: c32/325 lr:0.000978 t:2.5s +tttg: c33/325 lr:0.000976 t:2.5s +tttg: c34/325 lr:0.000975 t:2.6s +tttg: c35/325 lr:0.000973 t:2.7s +tttg: c36/325 lr:0.000971 t:2.8s +tttg: c37/325 lr:0.000970 t:2.9s +tttg: c38/325 lr:0.000968 t:2.9s +tttg: c39/325 lr:0.000966 t:3.0s +tttg: c40/325 lr:0.000965 t:3.1s +tttg: c41/325 lr:0.000963 t:3.2s +tttg: c42/325 lr:0.000961 t:3.2s +tttg: c43/325 lr:0.000959 t:3.3s +tttg: c44/325 lr:0.000957 t:3.4s +tttg: c45/325 lr:0.000955 t:3.5s +tttg: c46/325 lr:0.000953 t:3.5s +tttg: c47/325 lr:0.000951 t:3.6s +tttg: c48/325 lr:0.000949 t:3.7s +tttg: c49/325 lr:0.000947 t:3.8s +tttg: c50/325 lr:0.000945 t:3.8s +tttg: c51/325 lr:0.000942 t:3.9s +tttg: c52/325 lr:0.000940 t:4.0s +tttg: c53/325 lr:0.000938 t:4.1s +tttg: c54/325 lr:0.000935 t:4.2s +tttg: c55/325 lr:0.000933 t:4.2s +tttg: c56/325 lr:0.000931 t:4.3s +tttg: c57/325 lr:0.000928 t:4.4s +tttg: c58/325 lr:0.000926 t:4.5s +tttg: c59/325 lr:0.000923 t:4.5s +tttg: c60/325 lr:0.000920 t:4.6s +tttg: c61/325 lr:0.000918 t:4.7s +tttg: c62/325 lr:0.000915 t:4.8s +tttg: c63/325 lr:0.000912 t:4.9s +tttg: c64/325 lr:0.000910 t:4.9s +tttg: c65/325 lr:0.000907 t:5.0s +tttg: c66/325 lr:0.000904 t:5.1s +tttg: c67/325 lr:0.000901 t:5.2s +tttg: c68/325 lr:0.000898 t:5.2s +tttg: c69/325 lr:0.000895 t:5.3s +tttg: c70/325 lr:0.000892 t:5.4s +tttg: c71/325 lr:0.000889 t:5.5s +tttg: c72/325 lr:0.000886 t:5.6s +tttg: c73/325 lr:0.000883 t:5.6s +tttg: c74/325 lr:0.000880 t:5.7s +tttg: c75/325 lr:0.000877 t:5.8s +tttg: c76/325 lr:0.000874 t:5.9s +tttg: c77/325 lr:0.000870 t:5.9s +tttg: c78/325 lr:0.000867 t:6.0s +tttg: c79/325 lr:0.000864 t:6.1s +tttg: c80/325 lr:0.000860 t:6.2s +tttg: c81/325 lr:0.000857 t:6.2s +tttg: c82/325 lr:0.000854 t:6.3s +tttg: c83/325 lr:0.000850 t:6.4s +tttg: c84/325 lr:0.000847 t:6.5s +tttg: c85/325 lr:0.000843 t:6.6s +tttg: c86/325 lr:0.000840 t:6.6s +tttg: c87/325 lr:0.000836 t:6.7s +tttg: c88/325 lr:0.000832 t:6.8s +tttg: c89/325 lr:0.000829 t:6.9s +tttg: c90/325 lr:0.000825 t:6.9s +tttg: c91/325 lr:0.000821 t:7.0s +tttg: c92/325 lr:0.000818 t:7.1s +tttg: c93/325 lr:0.000814 t:7.2s +tttg: c94/325 lr:0.000810 t:7.2s +tttg: c95/325 lr:0.000806 t:7.3s +tttg: c96/325 lr:0.000802 t:7.4s +tttg: c97/325 lr:0.000799 t:7.5s +tttg: c98/325 lr:0.000795 t:7.6s +tttg: c99/325 lr:0.000791 t:7.6s +tttg: c100/325 lr:0.000787 t:7.7s +tttg: c101/325 lr:0.000783 t:7.8s +tttg: c102/325 lr:0.000779 t:7.9s +tttg: c103/325 lr:0.000775 t:7.9s +tttg: c104/325 lr:0.000771 t:8.0s +tttg: c105/325 lr:0.000767 t:8.1s +tttg: c106/325 lr:0.000762 t:8.2s +tttg: c107/325 lr:0.000758 t:8.2s +tttg: c108/325 lr:0.000754 t:8.3s +tttg: c109/325 lr:0.000750 t:8.4s +tttg: c110/325 lr:0.000746 t:8.5s +tttg: c111/325 lr:0.000742 t:8.6s +tttg: c112/325 lr:0.000737 t:8.6s +tttg: c113/325 lr:0.000733 t:8.7s +tttg: c114/325 lr:0.000729 t:8.8s +tttg: c115/325 lr:0.000724 t:8.9s +tttg: c116/325 lr:0.000720 t:8.9s +tttg: c117/325 lr:0.000716 t:9.0s +tttg: c118/325 lr:0.000711 t:9.1s +tttg: c119/325 lr:0.000707 t:9.2s +tttg: c120/325 lr:0.000702 t:9.2s +tttg: c121/325 lr:0.000698 t:9.3s +tttg: c122/325 lr:0.000694 t:9.4s +tttg: c123/325 lr:0.000689 t:9.5s +tttg: c124/325 lr:0.000685 t:9.6s +tttg: c125/325 lr:0.000680 t:9.6s +tttg: c126/325 lr:0.000676 t:9.7s +tttg: c127/325 lr:0.000671 t:9.8s +tttg: c128/325 lr:0.000666 t:9.9s +tttg: c129/325 lr:0.000662 t:9.9s +tttg: c130/325 lr:0.000657 t:10.0s +tttg: c131/325 lr:0.000653 t:10.1s +tttg: c132/325 lr:0.000648 t:10.2s +tttg: c133/325 lr:0.000643 t:10.2s +tttg: c134/325 lr:0.000639 t:10.3s +tttg: c135/325 lr:0.000634 t:10.4s +tttg: c136/325 lr:0.000629 t:10.5s +tttg: c137/325 lr:0.000625 t:10.5s +tttg: c138/325 lr:0.000620 t:10.6s +tttg: c139/325 lr:0.000615 t:10.7s +tttg: c140/325 lr:0.000611 t:10.8s +tttg: c141/325 lr:0.000606 t:10.9s +tttg: c142/325 lr:0.000601 t:10.9s +tttg: c143/325 lr:0.000596 t:11.0s +tttg: c144/325 lr:0.000592 t:11.1s +tttg: c145/325 lr:0.000587 t:11.2s +tttg: c146/325 lr:0.000582 t:11.3s +tttg: c147/325 lr:0.000577 t:11.3s +tttg: c148/325 lr:0.000572 t:11.4s +tttg: c149/325 lr:0.000568 t:11.5s +tttg: c150/325 lr:0.000563 t:11.6s +tttg: c151/325 lr:0.000558 t:11.6s +tttg: c152/325 lr:0.000553 t:11.7s +tttg: c153/325 lr:0.000548 t:11.8s +tttg: c154/325 lr:0.000544 t:11.9s +tttg: c155/325 lr:0.000539 t:11.9s +tttg: c156/325 lr:0.000534 t:12.0s +tttg: c157/325 lr:0.000529 t:12.1s +tttg: c158/325 lr:0.000524 t:12.2s +tttg: c159/325 lr:0.000519 t:12.3s +tttg: c160/325 lr:0.000515 t:12.3s +tttg: c161/325 lr:0.000510 t:12.4s +tttg: c162/325 lr:0.000505 t:12.5s +tttg: c163/325 lr:0.000500 t:12.6s +tttg: c164/325 lr:0.000495 t:12.6s +tttg: c165/325 lr:0.000490 t:12.7s +tttg: c166/325 lr:0.000485 t:12.8s +tttg: c167/325 lr:0.000481 t:12.9s +tttg: c168/325 lr:0.000476 t:12.9s +tttg: c169/325 lr:0.000471 t:13.0s +tttg: c170/325 lr:0.000466 t:13.1s +tttg: c171/325 lr:0.000461 t:13.2s +tttg: c172/325 lr:0.000456 t:13.3s +tttg: c173/325 lr:0.000452 t:13.3s +tttg: c174/325 lr:0.000447 t:13.4s +tttg: c175/325 lr:0.000442 t:13.5s +tttg: c176/325 lr:0.000437 t:13.6s +tttg: c177/325 lr:0.000432 t:13.6s +tttg: c178/325 lr:0.000428 t:13.7s +tttg: c179/325 lr:0.000423 t:13.8s +tttg: c180/325 lr:0.000418 t:13.9s +tttg: c181/325 lr:0.000413 t:13.9s +tttg: c182/325 lr:0.000408 t:14.0s +tttg: c183/325 lr:0.000404 t:14.1s +tttg: c184/325 lr:0.000399 t:14.2s +tttg: c185/325 lr:0.000394 t:14.3s +tttg: c186/325 lr:0.000389 t:14.3s +tttg: c187/325 lr:0.000385 t:14.4s +tttg: c188/325 lr:0.000380 t:14.5s +tttg: c189/325 lr:0.000375 t:14.6s +tttg: c190/325 lr:0.000371 t:14.7s +tttg: c191/325 lr:0.000366 t:14.7s +tttg: c192/325 lr:0.000361 t:14.8s +tttg: c193/325 lr:0.000357 t:14.9s +tttg: c194/325 lr:0.000352 t:15.0s +tttg: c195/325 lr:0.000347 t:15.0s +tttg: c196/325 lr:0.000343 t:15.1s +tttg: c197/325 lr:0.000338 t:15.2s +tttg: c198/325 lr:0.000334 t:15.3s +tttg: c199/325 lr:0.000329 t:15.3s +tttg: c200/325 lr:0.000324 t:15.4s +tttg: c201/325 lr:0.000320 t:15.5s +tttg: c202/325 lr:0.000315 t:15.6s +tttg: c203/325 lr:0.000311 t:15.7s +tttg: c204/325 lr:0.000306 t:15.7s +tttg: c205/325 lr:0.000302 t:15.8s +tttg: c206/325 lr:0.000298 t:15.9s +tttg: c207/325 lr:0.000293 t:16.0s +tttg: c208/325 lr:0.000289 t:16.0s +tttg: c209/325 lr:0.000284 t:16.1s +tttg: c210/325 lr:0.000280 t:16.2s +tttg: c211/325 lr:0.000276 t:16.3s +tttg: c212/325 lr:0.000271 t:16.4s +tttg: c213/325 lr:0.000267 t:16.4s +tttg: c214/325 lr:0.000263 t:16.5s +tttg: c215/325 lr:0.000258 t:16.6s +tttg: c216/325 lr:0.000254 t:16.7s +tttg: c217/325 lr:0.000250 t:16.7s +tttg: c218/325 lr:0.000246 t:16.8s +tttg: c219/325 lr:0.000242 t:16.9s +tttg: c220/325 lr:0.000238 t:17.0s +tttg: c221/325 lr:0.000233 t:17.1s +tttg: c222/325 lr:0.000229 t:17.1s +tttg: c223/325 lr:0.000225 t:17.2s +tttg: c224/325 lr:0.000221 t:17.3s +tttg: c225/325 lr:0.000217 t:17.4s +tttg: c226/325 lr:0.000213 t:17.4s +tttg: c227/325 lr:0.000209 t:17.5s +tttg: c228/325 lr:0.000205 t:17.6s +tttg: c229/325 lr:0.000201 t:17.7s +tttg: c230/325 lr:0.000198 t:17.7s +tttg: c231/325 lr:0.000194 t:17.8s +tttg: c232/325 lr:0.000190 t:17.9s +tttg: c233/325 lr:0.000186 t:18.0s +tttg: c234/325 lr:0.000182 t:18.1s +tttg: c235/325 lr:0.000179 t:18.1s +tttg: c236/325 lr:0.000175 t:18.2s +tttg: c237/325 lr:0.000171 t:18.3s +tttg: c238/325 lr:0.000168 t:18.4s +tttg: c239/325 lr:0.000164 t:18.4s +tttg: c240/325 lr:0.000160 t:18.5s +tttg: c241/325 lr:0.000157 t:18.6s +tttg: c242/325 lr:0.000153 t:18.7s +tttg: c243/325 lr:0.000150 t:18.8s +tttg: c244/325 lr:0.000146 t:18.8s +tttg: c245/325 lr:0.000143 t:18.9s +tttg: c246/325 lr:0.000140 t:19.0s +tttg: c247/325 lr:0.000136 t:19.1s +tttg: c248/325 lr:0.000133 t:19.1s +tttg: c249/325 lr:0.000130 t:19.2s +tttg: c250/325 lr:0.000126 t:19.3s +tttg: c251/325 lr:0.000123 t:19.4s +tttg: c252/325 lr:0.000120 t:19.4s +tttg: c253/325 lr:0.000117 t:19.5s +tttg: c254/325 lr:0.000114 t:19.6s +tttg: c255/325 lr:0.000111 t:19.7s +tttg: c256/325 lr:0.000108 t:19.8s +tttg: c257/325 lr:0.000105 t:19.8s +tttg: c258/325 lr:0.000102 t:19.9s +tttg: c259/325 lr:0.000099 t:20.0s +tttg: c260/325 lr:0.000096 t:20.1s +tttg: c261/325 lr:0.000093 t:20.1s +tttg: c262/325 lr:0.000090 t:20.2s +tttg: c263/325 lr:0.000088 t:20.3s +tttg: c264/325 lr:0.000085 t:20.4s +tttg: c265/325 lr:0.000082 t:20.4s +tttg: c266/325 lr:0.000080 t:20.5s +tttg: c267/325 lr:0.000077 t:20.6s +tttg: c268/325 lr:0.000074 t:20.7s +tttg: c269/325 lr:0.000072 t:20.7s +tttg: c270/325 lr:0.000069 t:20.8s +tttg: c271/325 lr:0.000067 t:20.9s +tttg: c272/325 lr:0.000065 t:21.0s +tttg: c273/325 lr:0.000062 t:21.1s +tttg: c274/325 lr:0.000060 t:21.1s +tttg: c275/325 lr:0.000058 t:21.2s +tttg: c276/325 lr:0.000055 t:21.3s +tttg: c277/325 lr:0.000053 t:21.4s +tttg: c278/325 lr:0.000051 t:21.4s +tttg: c279/325 lr:0.000049 t:21.5s +tttg: c280/325 lr:0.000047 t:21.6s +tttg: c281/325 lr:0.000045 t:21.7s +tttg: c282/325 lr:0.000043 t:21.7s +tttg: c283/325 lr:0.000041 t:21.8s +tttg: c284/325 lr:0.000039 t:21.9s +tttg: c285/325 lr:0.000037 t:22.0s +tttg: c286/325 lr:0.000035 t:22.1s +tttg: c287/325 lr:0.000034 t:22.1s +tttg: c288/325 lr:0.000032 t:22.2s +tttg: c289/325 lr:0.000030 t:22.3s +tttg: c290/325 lr:0.000029 t:22.4s +tttg: c291/325 lr:0.000027 t:22.4s +tttg: c292/325 lr:0.000025 t:22.5s +tttg: c293/325 lr:0.000024 t:22.6s +tttg: c294/325 lr:0.000022 t:22.7s +tttg: c295/325 lr:0.000021 t:22.7s +tttg: c296/325 lr:0.000020 t:22.8s +tttg: c297/325 lr:0.000018 t:22.9s +tttg: c298/325 lr:0.000017 t:23.0s +tttg: c299/325 lr:0.000016 t:23.1s +tttg: c300/325 lr:0.000015 t:23.1s +tttg: c301/325 lr:0.000013 t:23.2s +tttg: c302/325 lr:0.000012 t:23.3s +tttg: c303/325 lr:0.000011 t:23.4s +tttg: c304/325 lr:0.000010 t:23.4s +tttg: c305/325 lr:0.000009 t:23.5s +tttg: c306/325 lr:0.000008 t:23.6s +tttg: c307/325 lr:0.000008 t:23.7s +tttg: c308/325 lr:0.000007 t:23.7s +tttg: c309/325 lr:0.000006 t:23.8s +tttg: c310/325 lr:0.000005 t:23.9s +tttg: c311/325 lr:0.000005 t:24.0s +tttg: c312/325 lr:0.000004 t:24.1s +tttg: c313/325 lr:0.000003 t:24.1s +tttg: c314/325 lr:0.000003 t:24.2s +tttg: c315/325 lr:0.000002 t:24.3s +tttg: c316/325 lr:0.000002 t:24.4s +tttg: c317/325 lr:0.000002 t:24.4s +tttg: c318/325 lr:0.000001 t:24.5s +tttg: c319/325 lr:0.000001 t:24.6s +tttg: c320/325 lr:0.000001 t:24.7s +tttg: c321/325 lr:0.000000 t:24.7s +tttg: c322/325 lr:0.000000 t:24.8s +tttg: c323/325 lr:0.000000 t:24.9s +tttg: c324/325 lr:0.000000 t:25.0s +ttpr: phase:3/3 t:318.2s +ttp: b727/782 bl:2.2655 bb:1.0441 rl:2.2810 rb:1.0561 dl:2277-2305 gd:1 +ttp: b713/782 bl:2.2541 bb:1.0130 rl:2.2799 rb:1.0541 dl:2002-2017 gd:1 +ttp: b705/782 bl:2.3627 bb:1.0620 rl:2.2832 rb:1.0545 dl:1885-1898 gd:1 +ttp: b703/782 bl:2.3389 bb:1.0289 rl:2.2853 rb:1.0534 dl:1859-1872 gd:1 +ttp: b694/782 bl:2.3104 bb:1.0565 rl:2.2861 rb:1.0536 dl:1758-1769 gd:1 +ttp: b684/782 bl:2.3677 bb:1.0431 rl:2.2887 rb:1.0532 dl:1658-1665 gd:1 +ttp: b678/782 bl:2.3443 bb:1.0261 rl:2.2904 rb:1.0524 dl:1601-1610 gd:1 +ttp: b667/782 bl:2.3638 bb:1.0685 rl:2.2924 rb:1.0528 dl:1514-1521 gd:1 +ttp: b661/782 bl:2.3985 bb:1.0843 rl:2.2951 rb:1.0536 dl:1474-1480 gd:1 +ttp: b654/782 bl:2.2897 bb:1.0356 rl:2.2950 rb:1.0532 dl:1425-1432 gd:1 +ttp: b645/782 bl:2.3008 bb:1.0294 rl:2.2951 rb:1.0526 dl:1367-1375 gd:1 +ttp: b636/782 bl:2.3800 bb:1.0667 rl:2.2969 rb:1.0529 dl:1314-1320 gd:1 +ttp: b627/782 bl:2.3770 bb:1.0702 rl:2.2986 rb:1.0533 dl:1266-1271 gd:1 +ttp: b619/782 bl:2.3213 bb:1.0586 rl:2.2990 rb:1.0534 dl:1221-1226 gd:1 +ttp: b610/782 bl:2.2529 bb:1.0074 rl:2.2982 rb:1.0525 dl:1177-1182 gd:1 +ttp: b602/782 bl:2.3741 bb:1.0472 rl:2.2995 rb:1.0525 dl:1141-1146 gd:1 +ttp: b598/782 bl:2.3543 bb:1.0649 rl:2.3004 rb:1.0527 dl:1124-1129 gd:1 +ttp: b590/782 bl:2.3081 bb:1.0576 rl:2.3005 rb:1.0527 dl:1089-1093 gd:1 +ttp: b576/782 bl:2.3754 bb:1.0926 rl:2.3016 rb:1.0533 dl:1033-1037 gd:1 +ttp: b568/782 bl:2.3542 bb:1.0808 rl:2.3024 rb:1.0537 dl:1004-1007 gd:1 +ttp: b560/782 bl:2.2695 bb:1.0099 rl:2.3019 rb:1.0531 dl:975-979 gd:1 +ttp: b552/782 bl:2.2728 bb:1.0182 rl:2.3015 rb:1.0526 dl:949-952 gd:1 +ttp: b544/782 bl:2.3424 bb:1.0674 rl:2.3021 rb:1.0528 dl:924-927 gd:1 +ttp: b536/782 bl:2.3143 bb:1.0421 rl:2.3022 rb:1.0527 dl:899-902 gd:1 +ttp: b529/782 bl:2.3133 bb:1.0162 rl:2.3023 rb:1.0522 dl:878-882 gd:1 +ttp: b522/782 bl:2.3040 bb:1.0333 rl:2.3024 rb:1.0520 dl:858-860 gd:1 +ttp: b515/782 bl:2.3415 bb:1.0426 rl:2.3028 rb:1.0519 dl:838-841 gd:1 +ttp: b508/782 bl:2.3948 bb:1.0529 rl:2.3038 rb:1.0519 dl:817-820 gd:1 +ttp: b501/782 bl:2.3788 bb:1.0509 rl:2.3045 rb:1.0519 dl:799-802 gd:1 +ttp: b493/782 bl:2.3669 bb:1.0448 rl:2.3051 rb:1.0519 dl:778-780 gd:1 +ttp: b485/782 bl:2.2948 bb:1.0337 rl:2.3050 rb:1.0517 dl:759-761 gd:1 +ttp: b477/782 bl:2.4030 bb:1.0349 rl:2.3059 rb:1.0515 dl:740-742 gd:1 +ttp: b469/782 bl:2.3291 bb:1.0242 rl:2.3061 rb:1.0513 dl:721-724 gd:1 +ttp: b461/782 bl:2.3754 bb:1.0392 rl:2.3067 rb:1.0512 dl:703-706 gd:1 +ttp: b453/782 bl:2.3353 bb:1.0551 rl:2.3070 rb:1.0512 dl:687-689 gd:1 +ttp: b445/782 bl:2.3582 bb:1.0481 rl:2.3074 rb:1.0512 dl:670-672 gd:1 +ttp: b437/782 bl:2.2914 bb:1.0543 rl:2.3073 rb:1.0512 dl:653-655 gd:1 +ttp: b429/782 bl:2.2392 bb:1.0212 rl:2.3068 rb:1.0510 dl:638-640 gd:1 +ttp: b421/782 bl:2.2905 bb:1.0028 rl:2.3066 rb:1.0506 dl:622-624 gd:1 +ttp: b414/782 bl:2.2023 bb:1.0083 rl:2.3059 rb:1.0503 dl:609-611 gd:1 +ttp: b405/782 bl:2.3597 bb:1.0589 rl:2.3063 rb:1.0504 dl:592-593 gd:1 +ttp: b397/782 bl:2.3541 bb:1.0441 rl:2.3066 rb:1.0503 dl:577-579 gd:1 +ttp: b388/782 bl:2.3139 bb:1.0435 rl:2.3066 rb:1.0503 dl:561-562 gd:1 +ttp: b380/782 bl:2.3594 bb:1.0880 rl:2.3070 rb:1.0505 dl:547-549 gd:1 +ttp: b372/782 bl:2.3384 bb:1.0503 rl:2.3071 rb:1.0505 dl:533-535 gd:1 +ttp: b364/782 bl:2.3424 bb:1.0592 rl:2.3074 rb:1.0506 dl:521-522 gd:1 +ttp: b356/782 bl:2.3446 bb:1.0558 rl:2.3076 rb:1.0506 dl:506-508 gd:1 +ttp: b348/782 bl:2.3561 bb:1.0567 rl:2.3078 rb:1.0506 dl:494-495 gd:1 +ttp: b340/782 bl:2.4509 bb:1.0774 rl:2.3086 rb:1.0508 dl:482-483 gd:1 +ttp: b330/782 bl:2.2356 bb:1.0653 rl:2.3082 rb:1.0508 dl:466-468 gd:1 +ttp: b322/782 bl:2.3643 bb:1.0553 rl:2.3085 rb:1.0509 dl:455-457 gd:1 +ttp: b313/782 bl:2.2872 bb:1.0777 rl:2.3084 rb:1.0510 dl:440-442 gd:1 +ttp: b305/782 bl:2.3316 bb:1.0837 rl:2.3085 rb:1.0511 dl:429-430 gd:1 +ttp: b298/782 bl:2.4208 bb:1.1023 rl:2.3090 rb:1.0514 dl:418-420 gd:1 +ttp: b291/782 bl:2.2662 bb:1.0133 rl:2.3088 rb:1.0512 dl:407-409 gd:1 +ttp: b282/782 bl:2.3140 bb:1.0679 rl:2.3088 rb:1.0513 dl:395-396 gd:1 +ttp: b274/782 bl:2.2988 bb:1.0686 rl:2.3088 rb:1.0513 dl:384-385 gd:1 +ttp: b268/782 bl:2.3593 bb:1.0779 rl:2.3090 rb:1.0514 dl:376-378 gd:1 +ttp: b260/782 bl:2.3757 bb:1.0824 rl:2.3093 rb:1.0516 dl:366-367 gd:1 +ttp: b251/782 bl:2.3657 bb:1.0937 rl:2.3095 rb:1.0517 dl:355-356 gd:1 +ttp: b244/782 bl:2.3380 bb:1.1125 rl:2.3096 rb:1.0519 dl:346-347 gd:1 +ttp: b235/782 bl:2.2816 bb:1.0984 rl:2.3095 rb:1.0521 dl:335-336 gd:1 +ttp: b228/782 bl:2.3299 bb:1.0847 rl:2.3095 rb:1.0522 dl:327-328 gd:1 +ttp: b219/782 bl:2.3349 bb:1.1172 rl:2.3096 rb:1.0524 dl:316-317 gd:1 +ttp: b210/782 bl:2.2563 bb:1.0818 rl:2.3095 rb:1.0525 dl:306-307 gd:1 +ttp: b203/782 bl:2.4181 bb:1.1031 rl:2.3098 rb:1.0526 dl:299-300 gd:1 +ttp: b195/782 bl:2.4224 bb:1.1298 rl:2.3101 rb:1.0529 dl:290-291 gd:1 +ttp: b185/782 bl:2.4294 bb:1.1139 rl:2.3105 rb:1.0530 dl:279-280 gd:1 +ttp: b177/782 bl:2.4006 bb:1.1060 rl:2.3107 rb:1.0532 dl:271-272 gd:1 +ttp: b169/782 bl:2.3691 bb:1.1135 rl:2.3109 rb:1.0533 dl:263-264 gd:1 +ttp: b161/782 bl:2.3489 bb:1.1306 rl:2.3110 rb:1.0535 dl:256-256 gd:1 +ttp: b153/782 bl:2.2511 bb:1.0413 rl:2.3108 rb:1.0535 dl:248-249 gd:1 +ttp: b145/782 bl:2.5270 bb:1.1681 rl:2.3113 rb:1.0538 dl:240-241 gd:1 +ttp: b141/782 bl:2.4708 bb:1.1275 rl:2.3117 rb:1.0539 dl:236-237 gd:1 +ttp: b133/782 bl:2.3631 bb:1.1335 rl:2.3118 rb:1.0541 dl:229-230 gd:1 +ttp: b124/782 bl:2.3795 bb:1.1623 rl:2.3120 rb:1.0543 dl:220-222 gd:1 +ttp: b116/782 bl:2.4856 bb:1.1286 rl:2.3124 rb:1.0545 dl:213-214 gd:1 +ttp: b108/782 bl:2.3868 bb:1.1507 rl:2.3125 rb:1.0547 dl:206-207 gd:1 +ttp: b100/782 bl:2.4189 bb:1.1572 rl:2.3127 rb:1.0549 dl:199-200 gd:1 +ttp: b92/782 bl:2.4285 bb:1.1555 rl:2.3129 rb:1.0551 dl:191-192 gd:1 +ttp: b84/782 bl:2.5193 bb:1.1979 rl:2.3133 rb:1.0553 dl:184-185 gd:1 +ttp: b76/782 bl:2.4966 bb:1.1726 rl:2.3136 rb:1.0555 dl:177-178 gd:1 +ttp: b67/782 bl:2.5399 bb:1.2024 rl:2.3140 rb:1.0557 dl:169-170 gd:1 +ttp: b59/782 bl:2.5040 bb:1.1929 rl:2.3143 rb:1.0560 dl:162-163 gd:1 +ttp: b51/782 bl:2.4757 bb:1.1844 rl:2.3146 rb:1.0561 dl:154-155 gd:1 +ttp: b43/782 bl:2.5074 bb:1.2241 rl:2.3148 rb:1.0564 dl:146-147 gd:1 +ttp: b33/782 bl:2.5916 bb:1.2212 rl:2.3152 rb:1.0566 dl:136-137 gd:1 +ttp: b24/782 bl:2.4551 bb:1.1579 rl:2.3154 rb:1.0567 dl:127-128 gd:1 +ttp: b16/782 bl:2.6220 bb:1.2563 rl:2.3157 rb:1.0569 dl:117-118 gd:1 +ttp: b8/782 bl:2.7935 bb:1.2967 rl:2.3162 rb:1.0572 dl:103-105 gd:1 +quantized_ttt_phased val_loss:2.31965536 val_bpb:1.05999198 eval_time:400555ms +total_eval_time:400.6s diff --git a/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed42.log b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed42.log new file mode 100644 index 0000000000..1f39fd3f62 --- /dev/null +++ b/records/track_10min_16mb/2026-04-30_LongCtx_NoQV_Rank56Prefix3000/train_seed42.log @@ -0,0 +1,1041 @@ +W0430 03:56:32.122000 16981 torch/distributed/run.py:803] +W0430 03:56:32.122000 16981 torch/distributed/run.py:803] ***************************************** +W0430 03:56:32.122000 16981 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. +W0430 03:56:32.122000 16981 torch/distributed/run.py:803] ***************************************** +Hyperparameters: + adam_eps: 1e-08 + adam_wd: 0.02 + artifact_dir: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed42 + attn_clip_sigmas: 13.0 + attn_out_gate_enabled: False + attn_out_gate_src: proj + awq_lite_bits: 8 + awq_lite_enabled: True + awq_lite_group_size: 64 + awq_lite_group_top_k: 1 + beta1: 0.9 + beta2: 0.99 + caseops_enabled: True + compressor: pergroup + data_dir: ./data/ + datasets_dir: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved + distributed: True + ema_decay: 0.9965 + embed_bits: 7 + embed_clip_sigmas: 14.0 + embed_lr: 0.6 + embed_wd: 0.085 + enable_looping_at: 0.35 + eval_seq_len: 2560 + eval_stride: 64 + fused_ce_enabled: True + gate_window: 12 + gated_attn_enabled: False + gated_attn_init_std: 0.01 + gated_attn_quant_gate: True + global_ttt_batch_seqs: 32 + global_ttt_chunk_tokens: 32768 + global_ttt_epochs: 1 + global_ttt_grad_clip: 1.0 + global_ttt_lr: 0.001 + global_ttt_momentum: 0.9 + global_ttt_respect_doc_boundaries: True + global_ttt_warmup_chunks: 0 + global_ttt_warmup_start_lr: 0.0 + gptq_calibration_batches: 16 + gptq_reserve_seconds: 4.0 + grad_accum_steps: 1 + grad_clip_norm: 0.3 + is_main_process: True + iterations: 20000 + ln_scale: True + local_rank: 0 + logfile: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed42/rank56_prefix3000_seed42.txt + logit_softcap: 30.0 + loop_end: 5 + loop_start: 3 + lqer_asym_enabled: True + lqer_asym_group: 64 + lqer_enabled: True + lqer_factor_bits: 4 + lqer_gain_select: False + lqer_rank: 4 + lqer_scope: all + lqer_top_k: 3 + matrix_bits: 6 + matrix_clip_sigmas: 12.85 + matrix_lr: 0.026 + max_wallclock_seconds: 600.0 + min_lr: 0.1 + mlp_clip_sigmas: 11.5 + mlp_mult: 4.0 + model_dim: 512 + model_path: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed42/final_model.pt + muon_backend_steps: 5 + muon_momentum: 0.97 + muon_momentum_warmup_start: 0.92 + muon_momentum_warmup_steps: 1500 + muon_row_normalize: True + muon_wd: 0.095 + num_heads: 8 + num_kv_heads: 4 + num_layers: 11 + num_loops: 2 + parallel_final_lane: mean + parallel_start_layer: 8 + phased_ttt_num_phases: 3 + phased_ttt_prefix_docs: 3000 + qk_gain_init: 5.25 + quantized_model_path: /workspace/pg1953_rank56_scout/runs/rank56_prefix3000_seed42/final_model.int6.ptz + rank: 0 + rope_base: 10000.0 + rope_dims: 16 + rope_train_seq_len: 2048 + rope_yarn: False + run_id: rank56_prefix3000_seed42 + scalar_lr: 0.02 + seed: 42 + skip_gates_enabled: True + smear_gate_enabled: True + sparse_attn_gate_enabled: True + sparse_attn_gate_init_std: 0.0 + sparse_attn_gate_scale: 0.5 + tie_embeddings: True + tied_embed_init_std: 0.005 + tied_embed_lr: 0.03 + tokenizer_path: ./data/tokenizers/fineweb_8192_bpe_lossless_caps_caseops_v1_reserved.model + train_batch_tokens: 786432 + train_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_train_*.bin + train_log_every: 500 + train_seq_len: 2048 + ttt_batch_size: 64 + ttt_beta1: 0.0 + ttt_beta2: 0.99 + ttt_chunk_size: 48 + ttt_enabled: True + ttt_eval_batches: + ttt_eval_seq_len: 2560 + ttt_grad_steps: 1 + ttt_k_lora: True + ttt_local_lr_mult: 0.75 + ttt_lora_lr: 0.0001 + ttt_lora_rank: 56 + ttt_mask: no_qv + ttt_mlp_lora: True + ttt_o_lora: True + ttt_optimizer: adam + ttt_q_lora: False + ttt_v_lora: False + ttt_weight_decay: 0.5 + val_batch_tokens: 524288 + val_bytes_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_val_bytes_*.bin + val_doc_fraction: 1.0 + val_files: ./data/datasets/fineweb10B_sp8192_lossless_caps_caseops_v1_reserved/fineweb_val_*.bin + val_loss_every: 0 + vocab_size: 8192 + warmdown_frac: 0.85 + warmup_steps: 20 + world_size: 8 + xsa_last_n: 11 +train_shards: 80 +val_tokens: 47851520 +model_params:35945673 +gptq:reserving 4s, effective=596000ms +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 +1/20000 train_loss: 9.0087 train_time: 0.0m tok/s: 17876429 +2/20000 train_loss: 12.8307 train_time: 0.0m tok/s: 10668705 +3/20000 train_loss: 10.2095 train_time: 0.0m tok/s: 9766828 +4/20000 train_loss: 8.6887 train_time: 0.0m tok/s: 9399234 +5/20000 train_loss: 7.9427 train_time: 0.0m tok/s: 9199230 +500/20000 train_loss: 2.5645 train_time: 0.8m tok/s: 8187080 +1000/20000 train_loss: 2.7962 train_time: 1.6m tok/s: 8145578 +1500/20000 train_loss: 2.6170 train_time: 2.4m tok/s: 8134325 +2000/20000 train_loss: 2.6493 train_time: 3.2m tok/s: 8131326 +layer_loop:enabled step:2157 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: 2.5398 train_time: 4.3m tok/s: 7647461 +3000/20000 train_loss: 2.5534 train_time: 5.5m tok/s: 7198579 +3500/20000 train_loss: 2.5531 train_time: 6.6m tok/s: 6909739 +4000/20000 train_loss: 2.4007 train_time: 7.8m tok/s: 6706639 +4500/20000 train_loss: 2.2707 train_time: 9.0m tok/s: 6558454 +4900/20000 val_loss: 2.3480 val_bpb: 1.0729 +stopping_early: wallclock_cap train_time: 596051ms step: 4900/20000 +peak memory allocated: 41711 MiB reserved: 47040 MiB +ema:applying EMA weights +diagnostic pre-quantization post-ema val_loss:2.32225953 val_bpb:1.06108950 eval_time:9523ms +Serialized model: 135418111 bytes +Code size (uncompressed): 171669 bytes +Code size (compressed): 34125 bytes +GPTQ:collecting Hessians from calibration data... +GPTQ:collected 67 Hessians in 4.1s +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 (int6)+lqer_asym: blocks.mlp.fc.weight + gptq (int7)+awqgrpint8+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, softcap_neg, softcap_pos +Serialize: per-group lrzip compression... +Serialize: per-group compression done in 120.4s +Serialized model quantized+pergroup: 15941864 bytes +Total submission size quantized+pergroup: 15975989 bytes +Deserialize: per-group lrzip decompression... +Deserialize: decompression done in 20.9s +diagnostic quantized val_loss:2.34065949 val_bpb:1.06949683 eval_time:76086ms +Deserialize: per-group lrzip decompression... +Deserialize: decompression done in 20.9s +ttt_lora:warming up compile (random tokens, no val data) +ttt_lora:compile warmup done (149.9s) + +beginning TTT eval timer +ttt_phased: total_docs:50000 prefix_docs:3000 suffix_docs:47000 num_phases:3 boundaries:[1000, 2000, 3000] +ttp: b775/782 bl:2.2711 bb:1.0558 rl:2.2711 rb:1.0558 dl:6892-7524 gd:0 +ttp: b774/782 bl:2.2803 bb:1.0617 rl:2.2755 rb:1.0586 dl:6447-6872 gd:0 +ttp: b768/782 bl:2.2297 bb:1.0384 rl:2.2634 rb:1.0533 dl:4859-5083 gd:0 +ttp: b764/782 bl:2.2855 bb:1.0706 rl:2.2675 rb:1.0565 dl:4284-4392 gd:0 +ttpp: phase:1/3 pd:1424 gd:1000 t:243.4s +tttg: c1/154 lr:0.001000 t:1.9s +tttg: c2/154 lr:0.001000 t:2.0s +tttg: c3/154 lr:0.001000 t:2.1s +tttg: c4/154 lr:0.000999 t:2.2s +tttg: c5/154 lr:0.000998 t:2.2s +tttg: c6/154 lr:0.000997 t:2.3s +tttg: c7/154 lr:0.000996 t:2.4s +tttg: c8/154 lr:0.000995 t:2.5s +tttg: c9/154 lr:0.000993 t:2.5s +tttg: c10/154 lr:0.000991 t:2.6s +tttg: c11/154 lr:0.000989 t:2.7s +tttg: c12/154 lr:0.000987 t:2.8s +tttg: c13/154 lr:0.000985 t:2.9s +tttg: c14/154 lr:0.000982 t:2.9s +tttg: c15/154 lr:0.000979 t:3.0s +tttg: c16/154 lr:0.000976 t:3.1s +tttg: c17/154 lr:0.000973 t:3.2s +tttg: c18/154 lr:0.000970 t:3.3s +tttg: c19/154 lr:0.000966 t:3.3s +tttg: c20/154 lr:0.000962 t:3.4s +tttg: c21/154 lr:0.000958 t:3.5s +tttg: c22/154 lr:0.000954 t:3.6s +tttg: c23/154 lr:0.000950 t:3.7s +tttg: c24/154 lr:0.000945 t:3.7s +tttg: c25/154 lr:0.000941 t:3.8s +tttg: c26/154 lr:0.000936 t:3.9s +tttg: c27/154 lr:0.000930 t:4.0s +tttg: c28/154 lr:0.000925 t:4.1s +tttg: c29/154 lr:0.000920 t:4.1s +tttg: c30/154 lr:0.000914 t:4.2s +tttg: c31/154 lr:0.000908 t:4.3s +tttg: c32/154 lr:0.000902 t:4.4s +tttg: c33/154 lr:0.000896 t:4.4s +tttg: c34/154 lr:0.000890 t:4.5s +tttg: c35/154 lr:0.000883 t:4.6s +tttg: c36/154 lr:0.000876 t:4.7s +tttg: c37/154 lr:0.000870 t:4.8s +tttg: c38/154 lr:0.000863 t:4.8s +tttg: c39/154 lr:0.000855 t:4.9s +tttg: c40/154 lr:0.000848 t:5.0s +tttg: c41/154 lr:0.000841 t:5.1s +tttg: c42/154 lr:0.000833 t:5.1s +tttg: c43/154 lr:0.000825 t:5.2s +tttg: c44/154 lr:0.000817 t:5.3s +tttg: c45/154 lr:0.000809 t:5.4s +tttg: c46/154 lr:0.000801 t:5.5s +tttg: c47/154 lr:0.000793 t:5.5s +tttg: c48/154 lr:0.000785 t:5.6s +tttg: c49/154 lr:0.000776 t:5.7s +tttg: c50/154 lr:0.000768 t:5.8s +tttg: c51/154 lr:0.000759 t:5.9s +tttg: c52/154 lr:0.000750 t:5.9s +tttg: c53/154 lr:0.000741 t:6.0s +tttg: c54/154 lr:0.000732 t:6.1s +tttg: c55/154 lr:0.000723 t:6.2s +tttg: c56/154 lr:0.000714 t:6.3s +tttg: c57/154 lr:0.000704 t:6.3s +tttg: c58/154 lr:0.000695 t:6.4s +tttg: c59/154 lr:0.000685 t:6.5s +tttg: c60/154 lr:0.000676 t:6.6s +tttg: c61/154 lr:0.000666 t:6.7s +tttg: c62/154 lr:0.000656 t:6.7s +tttg: c63/154 lr:0.000647 t:6.8s +tttg: c64/154 lr:0.000637 t:6.9s +tttg: c65/154 lr:0.000627 t:7.0s +tttg: c66/154 lr:0.000617 t:7.1s +tttg: c67/154 lr:0.000607 t:7.1s +tttg: c68/154 lr:0.000597 t:7.2s +tttg: c69/154 lr:0.000587 t:7.3s +tttg: c70/154 lr:0.000577 t:7.4s +tttg: c71/154 lr:0.000567 t:7.5s +tttg: c72/154 lr:0.000556 t:7.5s +tttg: c73/154 lr:0.000546 t:7.6s +tttg: c74/154 lr:0.000536 t:7.7s +tttg: c75/154 lr:0.000526 t:7.8s +tttg: c76/154 lr:0.000515 t:7.9s +tttg: c77/154 lr:0.000505 t:7.9s +tttg: c78/154 lr:0.000495 t:8.0s +tttg: c79/154 lr:0.000485 t:8.1s +tttg: c80/154 lr:0.000474 t:8.2s +tttg: c81/154 lr:0.000464 t:8.2s +tttg: c82/154 lr:0.000454 t:8.3s +tttg: c83/154 lr:0.000444 t:8.4s +tttg: c84/154 lr:0.000433 t:8.5s +tttg: c85/154 lr:0.000423 t:8.6s +tttg: c86/154 lr:0.000413 t:8.6s +tttg: c87/154 lr:0.000403 t:8.7s +tttg: c88/154 lr:0.000393 t:8.8s +tttg: c89/154 lr:0.000383 t:8.9s +tttg: c90/154 lr:0.000373 t:9.0s +tttg: c91/154 lr:0.000363 t:9.1s +tttg: c92/154 lr:0.000353 t:9.1s +tttg: c93/154 lr:0.000344 t:9.2s +tttg: c94/154 lr:0.000334 t:9.3s +tttg: c95/154 lr:0.000324 t:9.4s +tttg: c96/154 lr:0.000315 t:9.4s +tttg: c97/154 lr:0.000305 t:9.5s +tttg: c98/154 lr:0.000296 t:9.6s +tttg: c99/154 lr:0.000286 t:9.7s +tttg: c100/154 lr:0.000277 t:9.8s +tttg: c101/154 lr:0.000268 t:9.8s +tttg: c102/154 lr:0.000259 t:9.9s +tttg: c103/154 lr:0.000250 t:10.0s +tttg: c104/154 lr:0.000241 t:10.1s +tttg: c105/154 lr:0.000232 t:10.2s +tttg: c106/154 lr:0.000224 t:10.2s +tttg: c107/154 lr:0.000215 t:10.3s +tttg: c108/154 lr:0.000207 t:10.4s +tttg: c109/154 lr:0.000199 t:10.5s +tttg: c110/154 lr:0.000191 t:10.6s +tttg: c111/154 lr:0.000183 t:10.6s +tttg: c112/154 lr:0.000175 t:10.7s +tttg: c113/154 lr:0.000167 t:10.8s +tttg: c114/154 lr:0.000159 t:10.9s +tttg: c115/154 lr:0.000152 t:11.0s +tttg: c116/154 lr:0.000145 t:11.0s +tttg: c117/154 lr:0.000137 t:11.1s +tttg: c118/154 lr:0.000130 t:11.2s +tttg: c119/154 lr:0.000124 t:11.3s +tttg: c120/154 lr:0.000117 t:11.3s +tttg: c121/154 lr:0.000110 t:11.4s +tttg: c122/154 lr:0.000104 t:11.5s +tttg: c123/154 lr:0.000098 t:11.6s +tttg: c124/154 lr:0.000092 t:11.7s +tttg: c125/154 lr:0.000086 t:11.7s +tttg: c126/154 lr:0.000080 t:11.8s +tttg: c127/154 lr:0.000075 t:11.9s +tttg: c128/154 lr:0.000070 t:12.0s +tttg: c129/154 lr:0.000064 t:12.1s +tttg: c130/154 lr:0.000059 t:12.1s +tttg: c131/154 lr:0.000055 t:12.2s +tttg: c132/154 lr:0.000050 t:12.3s +tttg: c133/154 lr:0.000046 t:12.4s +tttg: c134/154 lr:0.000042 t:12.5s +tttg: c135/154 lr:0.000038 t:12.5s +tttg: c136/154 lr:0.000034 t:12.6s +tttg: c137/154 lr:0.000030 t:12.7s +tttg: c138/154 lr:0.000027 t:12.8s +tttg: c139/154 lr:0.000024 t:12.9s +tttg: c140/154 lr:0.000021 t:12.9s +tttg: c141/154 lr:0.000018 t:13.0s +tttg: c142/154 lr:0.000015 t:13.1s +tttg: c143/154 lr:0.000013 t:13.2s +tttg: c144/154 lr:0.000011 t:13.3s +tttg: c145/154 lr:0.000009 t:13.3s +tttg: c146/154 lr:0.000007 t:13.4s +tttg: c147/154 lr:0.000005 t:13.5s +tttg: c148/154 lr:0.000004 t:13.6s +tttg: c149/154 lr:0.000003 t:13.7s +tttg: c150/154 lr:0.000002 t:13.7s +tttg: c151/154 lr:0.000001 t:13.8s +tttg: c152/154 lr:0.000000 t:13.9s +tttg: c153/154 lr:0.000000 t:14.0s +ttpr: phase:1/3 t:258.1s +ttp: b752/782 bl:2.3248 bb:1.0687 rl:2.2746 rb:1.0580 dl:3222-3283 gd:0 +ttp: b751/782 bl:2.3020 bb:1.0305 rl:2.2776 rb:1.0550 dl:3150-3221 gd:0 +ttp: b746/782 bl:2.4046 bb:1.0596 rl:2.2889 rb:1.0554 dl:2884-2943 gd:0 +ttp: b745/782 bl:2.2287 bb:1.0203 rl:2.2841 rb:1.0525 dl:2842-2883 gd:0 +ttpp: phase:2/3 pd:2448 gd:2000 t:369.5s +tttg: c1/250 lr:0.001000 t:0.1s +tttg: c2/250 lr:0.001000 t:0.2s +tttg: c3/250 lr:0.001000 t:0.3s +tttg: c4/250 lr:0.001000 t:0.3s +tttg: c5/250 lr:0.000999 t:0.4s +tttg: c6/250 lr:0.000999 t:0.5s +tttg: c7/250 lr:0.000999 t:0.6s +tttg: c8/250 lr:0.000998 t:0.6s +tttg: c9/250 lr:0.000997 t:0.7s +tttg: c10/250 lr:0.000997 t:0.8s +tttg: c11/250 lr:0.000996 t:0.9s +tttg: c12/250 lr:0.000995 t:1.0s +tttg: c13/250 lr:0.000994 t:1.0s +tttg: c14/250 lr:0.000993 t:1.1s +tttg: c15/250 lr:0.000992 t:1.2s +tttg: c16/250 lr:0.000991 t:1.3s +tttg: c17/250 lr:0.000990 t:1.4s +tttg: c18/250 lr:0.000989 t:1.4s +tttg: c19/250 lr:0.000987 t:1.5s +tttg: c20/250 lr:0.000986 t:1.6s +tttg: c21/250 lr:0.000984 t:1.7s +tttg: c22/250 lr:0.000983 t:1.8s +tttg: c23/250 lr:0.000981 t:1.8s +tttg: c24/250 lr:0.000979 t:1.9s +tttg: c25/250 lr:0.000977 t:2.0s +tttg: c26/250 lr:0.000975 t:2.1s +tttg: c27/250 lr:0.000973 t:2.2s +tttg: c28/250 lr:0.000971 t:2.2s +tttg: c29/250 lr:0.000969 t:2.3s +tttg: c30/250 lr:0.000967 t:2.4s +tttg: c31/250 lr:0.000965 t:2.5s +tttg: c32/250 lr:0.000962 t:2.5s +tttg: c33/250 lr:0.000960 t:2.6s +tttg: c34/250 lr:0.000957 t:2.7s +tttg: c35/250 lr:0.000955 t:2.8s +tttg: c36/250 lr:0.000952 t:2.9s +tttg: c37/250 lr:0.000949 t:3.0s +tttg: c38/250 lr:0.000947 t:3.0s +tttg: c39/250 lr:0.000944 t:3.1s +tttg: c40/250 lr:0.000941 t:3.2s +tttg: c41/250 lr:0.000938 t:3.3s +tttg: c42/250 lr:0.000935 t:3.4s +tttg: c43/250 lr:0.000931 t:3.4s +tttg: c44/250 lr:0.000928 t:3.5s +tttg: c45/250 lr:0.000925 t:3.6s +tttg: c46/250 lr:0.000922 t:3.7s +tttg: c47/250 lr:0.000918 t:3.8s +tttg: c48/250 lr:0.000915 t:3.8s +tttg: c49/250 lr:0.000911 t:3.9s +tttg: c50/250 lr:0.000907 t:4.0s +tttg: c51/250 lr:0.000904 t:4.1s +tttg: c52/250 lr:0.000900 t:4.1s +tttg: c53/250 lr:0.000896 t:4.2s +tttg: c54/250 lr:0.000892 t:4.3s +tttg: c55/250 lr:0.000888 t:4.4s +tttg: c56/250 lr:0.000884 t:4.5s +tttg: c57/250 lr:0.000880 t:4.5s +tttg: c58/250 lr:0.000876 t:4.6s +tttg: c59/250 lr:0.000872 t:4.7s +tttg: c60/250 lr:0.000868 t:4.8s +tttg: c61/250 lr:0.000863 t:4.9s +tttg: c62/250 lr:0.000859 t:5.0s +tttg: c63/250 lr:0.000855 t:5.0s +tttg: c64/250 lr:0.000850 t:5.1s +tttg: c65/250 lr:0.000846 t:5.2s +tttg: c66/250 lr:0.000841 t:5.3s +tttg: c67/250 lr:0.000836 t:5.4s +tttg: c68/250 lr:0.000832 t:5.4s +tttg: c69/250 lr:0.000827 t:5.5s +tttg: c70/250 lr:0.000822 t:5.6s +tttg: c71/250 lr:0.000817 t:5.7s +tttg: c72/250 lr:0.000812 t:5.8s +tttg: c73/250 lr:0.000807 t:5.8s +tttg: c74/250 lr:0.000803 t:5.9s +tttg: c75/250 lr:0.000797 t:6.0s +tttg: c76/250 lr:0.000792 t:6.1s +tttg: c77/250 lr:0.000787 t:6.1s +tttg: c78/250 lr:0.000782 t:6.2s +tttg: c79/250 lr:0.000777 t:6.3s +tttg: c80/250 lr:0.000772 t:6.4s +tttg: c81/250 lr:0.000766 t:6.5s +tttg: c82/250 lr:0.000761 t:6.5s +tttg: c83/250 lr:0.000755 t:6.6s +tttg: c84/250 lr:0.000750 t:6.7s +tttg: c85/250 lr:0.000745 t:6.8s +tttg: c86/250 lr:0.000739 t:6.8s +tttg: c87/250 lr:0.000733 t:6.9s +tttg: c88/250 lr:0.000728 t:7.0s +tttg: c89/250 lr:0.000722 t:7.1s +tttg: c90/250 lr:0.000717 t:7.2s +tttg: c91/250 lr:0.000711 t:7.2s +tttg: c92/250 lr:0.000705 t:7.3s +tttg: c93/250 lr:0.000699 t:7.4s +tttg: c94/250 lr:0.000694 t:7.5s +tttg: c95/250 lr:0.000688 t:7.6s +tttg: c96/250 lr:0.000682 t:7.6s +tttg: c97/250 lr:0.000676 t:7.7s +tttg: c98/250 lr:0.000670 t:7.8s +tttg: c99/250 lr:0.000664 t:7.9s +tttg: c100/250 lr:0.000658 t:8.0s +tttg: c101/250 lr:0.000652 t:8.0s +tttg: c102/250 lr:0.000646 t:8.1s +tttg: c103/250 lr:0.000640 t:8.2s +tttg: c104/250 lr:0.000634 t:8.3s +tttg: c105/250 lr:0.000628 t:8.4s +tttg: c106/250 lr:0.000622 t:8.4s +tttg: c107/250 lr:0.000616 t:8.5s +tttg: c108/250 lr:0.000610 t:8.6s +tttg: c109/250 lr:0.000603 t:8.7s +tttg: c110/250 lr:0.000597 t:8.8s +tttg: c111/250 lr:0.000591 t:8.8s +tttg: c112/250 lr:0.000585 t:8.9s +tttg: c113/250 lr:0.000579 t:9.0s +tttg: c114/250 lr:0.000572 t:9.1s +tttg: c115/250 lr:0.000566 t:9.2s +tttg: c116/250 lr:0.000560 t:9.2s +tttg: c117/250 lr:0.000554 t:9.3s +tttg: c118/250 lr:0.000547 t:9.4s +tttg: c119/250 lr:0.000541 t:9.5s +tttg: c120/250 lr:0.000535 t:9.6s +tttg: c121/250 lr:0.000528 t:9.6s +tttg: c122/250 lr:0.000522 t:9.7s +tttg: c123/250 lr:0.000516 t:9.8s +tttg: c124/250 lr:0.000509 t:9.9s +tttg: c125/250 lr:0.000503 t:10.0s +tttg: c126/250 lr:0.000497 t:10.0s +tttg: c127/250 lr:0.000491 t:10.1s +tttg: c128/250 lr:0.000484 t:10.2s +tttg: c129/250 lr:0.000478 t:10.3s +tttg: c130/250 lr:0.000472 t:10.4s +tttg: c131/250 lr:0.000465 t:10.4s +tttg: c132/250 lr:0.000459 t:10.5s +tttg: c133/250 lr:0.000453 t:10.6s +tttg: c134/250 lr:0.000446 t:10.7s +tttg: c135/250 lr:0.000440 t:10.7s +tttg: c136/250 lr:0.000434 t:10.8s +tttg: c137/250 lr:0.000428 t:10.9s +tttg: c138/250 lr:0.000421 t:11.0s +tttg: c139/250 lr:0.000415 t:11.1s +tttg: c140/250 lr:0.000409 t:11.1s +tttg: c141/250 lr:0.000403 t:11.2s +tttg: c142/250 lr:0.000397 t:11.3s +tttg: c143/250 lr:0.000390 t:11.4s +tttg: c144/250 lr:0.000384 t:11.5s +tttg: c145/250 lr:0.000378 t:11.6s +tttg: c146/250 lr:0.000372 t:11.6s +tttg: c147/250 lr:0.000366 t:11.7s +tttg: c148/250 lr:0.000360 t:11.8s +tttg: c149/250 lr:0.000354 t:11.9s +tttg: c150/250 lr:0.000348 t:12.0s +tttg: c151/250 lr:0.000342 t:12.0s +tttg: c152/250 lr:0.000336 t:12.1s +tttg: c153/250 lr:0.000330 t:12.2s +tttg: c154/250 lr:0.000324 t:12.3s +tttg: c155/250 lr:0.000318 t:12.4s +tttg: c156/250 lr:0.000312 t:12.4s +tttg: c157/250 lr:0.000306 t:12.5s +tttg: c158/250 lr:0.000301 t:12.6s +tttg: c159/250 lr:0.000295 t:12.7s +tttg: c160/250 lr:0.000289 t:12.7s +tttg: c161/250 lr:0.000283 t:12.8s +tttg: c162/250 lr:0.000278 t:12.9s +tttg: c163/250 lr:0.000272 t:13.0s +tttg: c164/250 lr:0.000267 t:13.1s +tttg: c165/250 lr:0.000261 t:13.1s +tttg: c166/250 lr:0.000255 t:13.2s +tttg: c167/250 lr:0.000250 t:13.3s +tttg: c168/250 lr:0.000245 t:13.4s +tttg: c169/250 lr:0.000239 t:13.5s +tttg: c170/250 lr:0.000234 t:13.5s +tttg: c171/250 lr:0.000228 t:13.6s +tttg: c172/250 lr:0.000223 t:13.7s +tttg: c173/250 lr:0.000218 t:13.8s +tttg: c174/250 lr:0.000213 t:13.9s +tttg: c175/250 lr:0.000208 t:13.9s +tttg: c176/250 lr:0.000203 t:14.0s +tttg: c177/250 lr:0.000197 t:14.1s +tttg: c178/250 lr:0.000193 t:14.2s +tttg: c179/250 lr:0.000188 t:14.3s +tttg: c180/250 lr:0.000183 t:14.3s +tttg: c181/250 lr:0.000178 t:14.4s +tttg: c182/250 lr:0.000173 t:14.5s +tttg: c183/250 lr:0.000168 t:14.6s +tttg: c184/250 lr:0.000164 t:14.7s +tttg: c185/250 lr:0.000159 t:14.7s +tttg: c186/250 lr:0.000154 t:14.8s +tttg: c187/250 lr:0.000150 t:14.9s +tttg: c188/250 lr:0.000145 t:15.0s +tttg: c189/250 lr:0.000141 t:15.1s +tttg: c190/250 lr:0.000137 t:15.1s +tttg: c191/250 lr:0.000132 t:15.2s +tttg: c192/250 lr:0.000128 t:15.3s +tttg: c193/250 lr:0.000124 t:15.4s +tttg: c194/250 lr:0.000120 t:15.4s +tttg: c195/250 lr:0.000116 t:15.5s +tttg: c196/250 lr:0.000112 t:15.6s +tttg: c197/250 lr:0.000108 t:15.7s +tttg: c198/250 lr:0.000104 t:15.8s +tttg: c199/250 lr:0.000100 t:15.9s +tttg: c200/250 lr:0.000096 t:15.9s +tttg: c201/250 lr:0.000093 t:16.0s +tttg: c202/250 lr:0.000089 t:16.1s +tttg: c203/250 lr:0.000085 t:16.2s +tttg: c204/250 lr:0.000082 t:16.2s +tttg: c205/250 lr:0.000078 t:16.3s +tttg: c206/250 lr:0.000075 t:16.4s +tttg: c207/250 lr:0.000072 t:16.5s +tttg: c208/250 lr:0.000069 t:16.6s +tttg: c209/250 lr:0.000065 t:16.6s +tttg: c210/250 lr:0.000062 t:16.7s +tttg: c211/250 lr:0.000059 t:16.8s +tttg: c212/250 lr:0.000056 t:16.9s +tttg: c213/250 lr:0.000053 t:17.0s +tttg: c214/250 lr:0.000051 t:17.0s +tttg: c215/250 lr:0.000048 t:17.1s +tttg: c216/250 lr:0.000045 t:17.2s +tttg: c217/250 lr:0.000043 t:17.3s +tttg: c218/250 lr:0.000040 t:17.4s +tttg: c219/250 lr:0.000038 t:17.5s +tttg: c220/250 lr:0.000035 t:17.5s +tttg: c221/250 lr:0.000033 t:17.6s +tttg: c222/250 lr:0.000031 t:17.7s +tttg: c223/250 lr:0.000029 t:17.8s +tttg: c224/250 lr:0.000027 t:17.8s +tttg: c225/250 lr:0.000025 t:17.9s +tttg: c226/250 lr:0.000023 t:18.0s +tttg: c227/250 lr:0.000021 t:18.1s +tttg: c228/250 lr:0.000019 t:18.2s +tttg: c229/250 lr:0.000017 t:18.2s +tttg: c230/250 lr:0.000016 t:18.3s +tttg: c231/250 lr:0.000014 t:18.4s +tttg: c232/250 lr:0.000013 t:18.5s +tttg: c233/250 lr:0.000011 t:18.6s +tttg: c234/250 lr:0.000010 t:18.6s +tttg: c235/250 lr:0.000009 t:18.7s +tttg: c236/250 lr:0.000008 t:18.8s +tttg: c237/250 lr:0.000007 t:18.9s +tttg: c238/250 lr:0.000006 t:19.0s +tttg: c239/250 lr:0.000005 t:19.0s +tttg: c240/250 lr:0.000004 t:19.1s +tttg: c241/250 lr:0.000003 t:19.2s +tttg: c242/250 lr:0.000003 t:19.3s +tttg: c243/250 lr:0.000002 t:19.4s +tttg: c244/250 lr:0.000001 t:19.4s +tttg: c245/250 lr:0.000001 t:19.5s +tttg: c246/250 lr:0.000001 t:19.6s +tttg: c247/250 lr:0.000000 t:19.7s +tttg: c248/250 lr:0.000000 t:19.7s +tttg: c249/250 lr:0.000000 t:19.8s +ttpr: phase:2/3 t:390.1s +ttp: b736/782 bl:2.2381 bb:1.0545 rl:2.2810 rb:1.0527 dl:2526-2550 gd:0 +ttp: b734/782 bl:2.2550 bb:1.0259 rl:2.2794 rb:1.0510 dl:2469-2495 gd:0 +ttpp: phase:3/3 pd:3472 gd:3000 t:403.0s +tttg: c1/325 lr:0.001000 t:0.1s +tttg: c2/325 lr:0.001000 t:0.2s +tttg: c3/325 lr:0.001000 t:0.2s +tttg: c4/325 lr:0.001000 t:0.3s +tttg: c5/325 lr:0.001000 t:0.4s +tttg: c6/325 lr:0.000999 t:0.5s +tttg: c7/325 lr:0.000999 t:0.5s +tttg: c8/325 lr:0.000999 t:0.6s +tttg: c9/325 lr:0.000998 t:0.7s +tttg: c10/325 lr:0.000998 t:0.8s +tttg: c11/325 lr:0.000998 t:0.9s +tttg: c12/325 lr:0.000997 t:0.9s +tttg: c13/325 lr:0.000997 t:1.0s +tttg: c14/325 lr:0.000996 t:1.1s +tttg: c15/325 lr:0.000995 t:1.2s +tttg: c16/325 lr:0.000995 t:1.2s +tttg: c17/325 lr:0.000994 t:1.3s +tttg: c18/325 lr:0.000993 t:6.9s +tttg: c19/325 lr:0.000992 t:7.0s +tttg: c20/325 lr:0.000992 t:7.1s +tttg: c21/325 lr:0.000991 t:7.2s +tttg: c22/325 lr:0.000990 t:7.2s +tttg: c23/325 lr:0.000989 t:7.3s +tttg: c24/325 lr:0.000988 t:7.4s +tttg: c25/325 lr:0.000987 t:7.5s +tttg: c26/325 lr:0.000985 t:7.5s +tttg: c27/325 lr:0.000984 t:7.6s +tttg: c28/325 lr:0.000983 t:7.7s +tttg: c29/325 lr:0.000982 t:7.8s +tttg: c30/325 lr:0.000980 t:7.9s +tttg: c31/325 lr:0.000979 t:7.9s +tttg: c32/325 lr:0.000978 t:8.0s +tttg: c33/325 lr:0.000976 t:8.1s +tttg: c34/325 lr:0.000975 t:8.2s +tttg: c35/325 lr:0.000973 t:8.3s +tttg: c36/325 lr:0.000971 t:8.3s +tttg: c37/325 lr:0.000970 t:8.4s +tttg: c38/325 lr:0.000968 t:8.5s +tttg: c39/325 lr:0.000966 t:8.6s +tttg: c40/325 lr:0.000965 t:8.6s +tttg: c41/325 lr:0.000963 t:8.7s +tttg: c42/325 lr:0.000961 t:8.8s +tttg: c43/325 lr:0.000959 t:8.9s +tttg: c44/325 lr:0.000957 t:9.0s +tttg: c45/325 lr:0.000955 t:9.0s +tttg: c46/325 lr:0.000953 t:9.1s +tttg: c47/325 lr:0.000951 t:9.2s +tttg: c48/325 lr:0.000949 t:9.3s +tttg: c49/325 lr:0.000947 t:9.3s +tttg: c50/325 lr:0.000945 t:9.4s +tttg: c51/325 lr:0.000942 t:9.5s +tttg: c52/325 lr:0.000940 t:9.6s +tttg: c53/325 lr:0.000938 t:9.7s +tttg: c54/325 lr:0.000935 t:9.7s +tttg: c55/325 lr:0.000933 t:9.8s +tttg: c56/325 lr:0.000931 t:9.9s +tttg: c57/325 lr:0.000928 t:10.0s +tttg: c58/325 lr:0.000926 t:10.1s +tttg: c59/325 lr:0.000923 t:10.1s +tttg: c60/325 lr:0.000920 t:10.2s +tttg: c61/325 lr:0.000918 t:10.3s +tttg: c62/325 lr:0.000915 t:10.4s +tttg: c63/325 lr:0.000912 t:10.4s +tttg: c64/325 lr:0.000910 t:10.5s +tttg: c65/325 lr:0.000907 t:10.6s +tttg: c66/325 lr:0.000904 t:10.7s +tttg: c67/325 lr:0.000901 t:10.8s +tttg: c68/325 lr:0.000898 t:10.8s +tttg: c69/325 lr:0.000895 t:10.9s +tttg: c70/325 lr:0.000892 t:11.0s +tttg: c71/325 lr:0.000889 t:11.1s +tttg: c72/325 lr:0.000886 t:11.2s +tttg: c73/325 lr:0.000883 t:11.2s +tttg: c74/325 lr:0.000880 t:11.3s +tttg: c75/325 lr:0.000877 t:11.4s +tttg: c76/325 lr:0.000874 t:11.5s +tttg: c77/325 lr:0.000870 t:11.5s +tttg: c78/325 lr:0.000867 t:11.6s +tttg: c79/325 lr:0.000864 t:11.7s +tttg: c80/325 lr:0.000860 t:11.8s +tttg: c81/325 lr:0.000857 t:11.9s +tttg: c82/325 lr:0.000854 t:11.9s +tttg: c83/325 lr:0.000850 t:12.0s +tttg: c84/325 lr:0.000847 t:12.1s +tttg: c85/325 lr:0.000843 t:12.2s +tttg: c86/325 lr:0.000840 t:12.2s +tttg: c87/325 lr:0.000836 t:12.3s +tttg: c88/325 lr:0.000832 t:12.4s +tttg: c89/325 lr:0.000829 t:12.5s +tttg: c90/325 lr:0.000825 t:12.6s +tttg: c91/325 lr:0.000821 t:12.6s +tttg: c92/325 lr:0.000818 t:12.7s +tttg: c93/325 lr:0.000814 t:12.8s +tttg: c94/325 lr:0.000810 t:12.9s +tttg: c95/325 lr:0.000806 t:13.0s +tttg: c96/325 lr:0.000802 t:13.0s +tttg: c97/325 lr:0.000799 t:13.1s +tttg: c98/325 lr:0.000795 t:13.2s +tttg: c99/325 lr:0.000791 t:13.3s +tttg: c100/325 lr:0.000787 t:13.3s +tttg: c101/325 lr:0.000783 t:13.4s +tttg: c102/325 lr:0.000779 t:13.5s +tttg: c103/325 lr:0.000775 t:13.6s +tttg: c104/325 lr:0.000771 t:13.7s +tttg: c105/325 lr:0.000767 t:13.7s +tttg: c106/325 lr:0.000762 t:13.8s +tttg: c107/325 lr:0.000758 t:13.9s +tttg: c108/325 lr:0.000754 t:14.0s +tttg: c109/325 lr:0.000750 t:14.0s +tttg: c110/325 lr:0.000746 t:14.1s +tttg: c111/325 lr:0.000742 t:14.2s +tttg: c112/325 lr:0.000737 t:14.3s +tttg: c113/325 lr:0.000733 t:14.4s +tttg: c114/325 lr:0.000729 t:14.4s +tttg: c115/325 lr:0.000724 t:14.5s +tttg: c116/325 lr:0.000720 t:14.6s +tttg: c117/325 lr:0.000716 t:14.7s +tttg: c118/325 lr:0.000711 t:14.8s +tttg: c119/325 lr:0.000707 t:14.8s +tttg: c120/325 lr:0.000702 t:14.9s +tttg: c121/325 lr:0.000698 t:15.0s +tttg: c122/325 lr:0.000694 t:15.1s +tttg: c123/325 lr:0.000689 t:15.2s +tttg: c124/325 lr:0.000685 t:15.2s +tttg: c125/325 lr:0.000680 t:15.3s +tttg: c126/325 lr:0.000676 t:15.4s +tttg: c127/325 lr:0.000671 t:15.5s +tttg: c128/325 lr:0.000666 t:15.6s +tttg: c129/325 lr:0.000662 t:15.7s +tttg: c130/325 lr:0.000657 t:15.7s +tttg: c131/325 lr:0.000653 t:15.8s +tttg: c132/325 lr:0.000648 t:15.9s +tttg: c133/325 lr:0.000643 t:16.0s +tttg: c134/325 lr:0.000639 t:16.1s +tttg: c135/325 lr:0.000634 t:16.1s +tttg: c136/325 lr:0.000629 t:16.2s +tttg: c137/325 lr:0.000625 t:16.3s +tttg: c138/325 lr:0.000620 t:16.4s +tttg: c139/325 lr:0.000615 t:16.5s +tttg: c140/325 lr:0.000611 t:16.5s +tttg: c141/325 lr:0.000606 t:16.6s +tttg: c142/325 lr:0.000601 t:16.7s +tttg: c143/325 lr:0.000596 t:16.8s +tttg: c144/325 lr:0.000592 t:16.9s +tttg: c145/325 lr:0.000587 t:16.9s +tttg: c146/325 lr:0.000582 t:17.0s +tttg: c147/325 lr:0.000577 t:17.1s +tttg: c148/325 lr:0.000572 t:17.2s +tttg: c149/325 lr:0.000568 t:17.3s +tttg: c150/325 lr:0.000563 t:17.3s +tttg: c151/325 lr:0.000558 t:17.4s +tttg: c152/325 lr:0.000553 t:17.5s +tttg: c153/325 lr:0.000548 t:17.6s +tttg: c154/325 lr:0.000544 t:17.6s +tttg: c155/325 lr:0.000539 t:17.7s +tttg: c156/325 lr:0.000534 t:17.8s +tttg: c157/325 lr:0.000529 t:17.9s +tttg: c158/325 lr:0.000524 t:18.0s +tttg: c159/325 lr:0.000519 t:18.0s +tttg: c160/325 lr:0.000515 t:18.1s +tttg: c161/325 lr:0.000510 t:18.2s +tttg: c162/325 lr:0.000505 t:18.3s +tttg: c163/325 lr:0.000500 t:18.4s +tttg: c164/325 lr:0.000495 t:18.4s +tttg: c165/325 lr:0.000490 t:18.5s +tttg: c166/325 lr:0.000485 t:18.6s +tttg: c167/325 lr:0.000481 t:18.7s +tttg: c168/325 lr:0.000476 t:18.8s +tttg: c169/325 lr:0.000471 t:18.8s +tttg: c170/325 lr:0.000466 t:18.9s +tttg: c171/325 lr:0.000461 t:19.0s +tttg: c172/325 lr:0.000456 t:19.1s +tttg: c173/325 lr:0.000452 t:19.2s +tttg: c174/325 lr:0.000447 t:19.2s +tttg: c175/325 lr:0.000442 t:19.3s +tttg: c176/325 lr:0.000437 t:19.4s +tttg: c177/325 lr:0.000432 t:19.5s +tttg: c178/325 lr:0.000428 t:19.6s +tttg: c179/325 lr:0.000423 t:19.6s +tttg: c180/325 lr:0.000418 t:19.7s +tttg: c181/325 lr:0.000413 t:19.8s +tttg: c182/325 lr:0.000408 t:19.9s +tttg: c183/325 lr:0.000404 t:20.0s +tttg: c184/325 lr:0.000399 t:20.0s +tttg: c185/325 lr:0.000394 t:20.1s +tttg: c186/325 lr:0.000389 t:20.2s +tttg: c187/325 lr:0.000385 t:20.3s +tttg: c188/325 lr:0.000380 t:20.4s +tttg: c189/325 lr:0.000375 t:20.4s +tttg: c190/325 lr:0.000371 t:20.5s +tttg: c191/325 lr:0.000366 t:20.6s +tttg: c192/325 lr:0.000361 t:20.7s +tttg: c193/325 lr:0.000357 t:20.7s +tttg: c194/325 lr:0.000352 t:20.8s +tttg: c195/325 lr:0.000347 t:20.9s +tttg: c196/325 lr:0.000343 t:21.0s +tttg: c197/325 lr:0.000338 t:21.1s +tttg: c198/325 lr:0.000334 t:21.2s +tttg: c199/325 lr:0.000329 t:21.2s +tttg: c200/325 lr:0.000324 t:21.3s +tttg: c201/325 lr:0.000320 t:21.4s +tttg: c202/325 lr:0.000315 t:21.5s +tttg: c203/325 lr:0.000311 t:21.6s +tttg: c204/325 lr:0.000306 t:21.6s +tttg: c205/325 lr:0.000302 t:21.7s +tttg: c206/325 lr:0.000298 t:21.8s +tttg: c207/325 lr:0.000293 t:21.9s +tttg: c208/325 lr:0.000289 t:22.0s +tttg: c209/325 lr:0.000284 t:22.0s +tttg: c210/325 lr:0.000280 t:22.1s +tttg: c211/325 lr:0.000276 t:22.2s +tttg: c212/325 lr:0.000271 t:22.3s +tttg: c213/325 lr:0.000267 t:22.4s +tttg: c214/325 lr:0.000263 t:22.4s +tttg: c215/325 lr:0.000258 t:22.5s +tttg: c216/325 lr:0.000254 t:22.6s +tttg: c217/325 lr:0.000250 t:22.7s +tttg: c218/325 lr:0.000246 t:22.8s +tttg: c219/325 lr:0.000242 t:22.8s +tttg: c220/325 lr:0.000238 t:22.9s +tttg: c221/325 lr:0.000233 t:23.0s +tttg: c222/325 lr:0.000229 t:23.1s +tttg: c223/325 lr:0.000225 t:23.2s +tttg: c224/325 lr:0.000221 t:23.2s +tttg: c225/325 lr:0.000217 t:23.3s +tttg: c226/325 lr:0.000213 t:23.4s +tttg: c227/325 lr:0.000209 t:23.5s +tttg: c228/325 lr:0.000205 t:23.6s +tttg: c229/325 lr:0.000201 t:23.6s +tttg: c230/325 lr:0.000198 t:23.7s +tttg: c231/325 lr:0.000194 t:23.8s +tttg: c232/325 lr:0.000190 t:23.9s +tttg: c233/325 lr:0.000186 t:24.0s +tttg: c234/325 lr:0.000182 t:24.0s +tttg: c235/325 lr:0.000179 t:24.1s +tttg: c236/325 lr:0.000175 t:24.2s +tttg: c237/325 lr:0.000171 t:24.3s +tttg: c238/325 lr:0.000168 t:24.4s +tttg: c239/325 lr:0.000164 t:24.5s +tttg: c240/325 lr:0.000160 t:24.5s +tttg: c241/325 lr:0.000157 t:24.6s +tttg: c242/325 lr:0.000153 t:24.7s +tttg: c243/325 lr:0.000150 t:24.8s +tttg: c244/325 lr:0.000146 t:24.9s +tttg: c245/325 lr:0.000143 t:24.9s +tttg: c246/325 lr:0.000140 t:25.0s +tttg: c247/325 lr:0.000136 t:25.1s +tttg: c248/325 lr:0.000133 t:25.2s +tttg: c249/325 lr:0.000130 t:25.3s +tttg: c250/325 lr:0.000126 t:25.4s +tttg: c251/325 lr:0.000123 t:25.4s +tttg: c252/325 lr:0.000120 t:25.5s +tttg: c253/325 lr:0.000117 t:25.6s +tttg: c254/325 lr:0.000114 t:25.7s +tttg: c255/325 lr:0.000111 t:25.7s +tttg: c256/325 lr:0.000108 t:25.8s +tttg: c257/325 lr:0.000105 t:25.9s +tttg: c258/325 lr:0.000102 t:26.0s +tttg: c259/325 lr:0.000099 t:26.1s +tttg: c260/325 lr:0.000096 t:26.1s +tttg: c261/325 lr:0.000093 t:26.2s +tttg: c262/325 lr:0.000090 t:26.3s +tttg: c263/325 lr:0.000088 t:26.4s +tttg: c264/325 lr:0.000085 t:26.5s +tttg: c265/325 lr:0.000082 t:26.5s +tttg: c266/325 lr:0.000080 t:26.6s +tttg: c267/325 lr:0.000077 t:26.7s +tttg: c268/325 lr:0.000074 t:26.8s +tttg: c269/325 lr:0.000072 t:26.9s +tttg: c270/325 lr:0.000069 t:26.9s +tttg: c271/325 lr:0.000067 t:27.0s +tttg: c272/325 lr:0.000065 t:27.1s +tttg: c273/325 lr:0.000062 t:27.2s +tttg: c274/325 lr:0.000060 t:27.3s +tttg: c275/325 lr:0.000058 t:27.3s +tttg: c276/325 lr:0.000055 t:27.4s +tttg: c277/325 lr:0.000053 t:27.5s +tttg: c278/325 lr:0.000051 t:27.6s +tttg: c279/325 lr:0.000049 t:27.7s +tttg: c280/325 lr:0.000047 t:27.7s +tttg: c281/325 lr:0.000045 t:27.8s +tttg: c282/325 lr:0.000043 t:27.9s +tttg: c283/325 lr:0.000041 t:28.0s +tttg: c284/325 lr:0.000039 t:28.1s +tttg: c285/325 lr:0.000037 t:28.1s +tttg: c286/325 lr:0.000035 t:28.2s +tttg: c287/325 lr:0.000034 t:28.3s +tttg: c288/325 lr:0.000032 t:28.4s +tttg: c289/325 lr:0.000030 t:28.5s +tttg: c290/325 lr:0.000029 t:28.5s +tttg: c291/325 lr:0.000027 t:28.6s +tttg: c292/325 lr:0.000025 t:28.7s +tttg: c293/325 lr:0.000024 t:28.8s +tttg: c294/325 lr:0.000022 t:28.9s +tttg: c295/325 lr:0.000021 t:28.9s +tttg: c296/325 lr:0.000020 t:29.0s +tttg: c297/325 lr:0.000018 t:29.1s +tttg: c298/325 lr:0.000017 t:29.2s +tttg: c299/325 lr:0.000016 t:29.3s +tttg: c300/325 lr:0.000015 t:29.3s +tttg: c301/325 lr:0.000013 t:29.4s +tttg: c302/325 lr:0.000012 t:29.5s +tttg: c303/325 lr:0.000011 t:29.6s +tttg: c304/325 lr:0.000010 t:29.6s +tttg: c305/325 lr:0.000009 t:29.7s +tttg: c306/325 lr:0.000008 t:29.8s +tttg: c307/325 lr:0.000008 t:29.9s +tttg: c308/325 lr:0.000007 t:30.0s +tttg: c309/325 lr:0.000006 t:30.1s +tttg: c310/325 lr:0.000005 t:30.1s +tttg: c311/325 lr:0.000005 t:30.2s +tttg: c312/325 lr:0.000004 t:30.3s +tttg: c313/325 lr:0.000003 t:30.4s +tttg: c314/325 lr:0.000003 t:30.4s +tttg: c315/325 lr:0.000002 t:30.5s +tttg: c316/325 lr:0.000002 t:30.6s +tttg: c317/325 lr:0.000002 t:30.7s +tttg: c318/325 lr:0.000001 t:30.8s +tttg: c319/325 lr:0.000001 t:30.8s +tttg: c320/325 lr:0.000001 t:30.9s +tttg: c321/325 lr:0.000000 t:31.0s +tttg: c322/325 lr:0.000000 t:31.1s +tttg: c323/325 lr:0.000000 t:31.2s +tttg: c324/325 lr:0.000000 t:31.2s +ttpr: phase:3/3 t:435.0s +ttp: b724/782 bl:2.3104 bb:1.0550 rl:2.2810 rb:1.0512 dl:2203-2231 gd:1 +ttp: b715/782 bl:2.3557 bb:1.0270 rl:2.2844 rb:1.0500 dl:2036-2053 gd:1 +ttp: b707/782 bl:2.3535 bb:1.0459 rl:2.2873 rb:1.0498 dl:1910-1923 gd:1 +ttp: b697/782 bl:2.3205 bb:1.0296 rl:2.2885 rb:1.0491 dl:1790-1803 gd:1 +ttp: b693/782 bl:2.3331 bb:1.0481 rl:2.2901 rb:1.0490 dl:1746-1757 gd:1 +ttp: b683/782 bl:2.2668 bb:1.0552 rl:2.2893 rb:1.0492 dl:1646-1657 gd:1 +ttp: b676/782 bl:2.3270 bb:1.0467 rl:2.2904 rb:1.0492 dl:1586-1595 gd:1 +ttp: b665/782 bl:2.3209 bb:1.0427 rl:2.2913 rb:1.0490 dl:1500-1507 gd:1 +ttp: b659/782 bl:2.2992 bb:1.0376 rl:2.2915 rb:1.0487 dl:1459-1466 gd:1 +ttp: b652/782 bl:2.2419 bb:1.0191 rl:2.2903 rb:1.0479 dl:1411-1419 gd:1 +ttp: b644/782 bl:2.3588 bb:1.0472 rl:2.2918 rb:1.0479 dl:1362-1367 gd:1 +ttp: b635/782 bl:2.3356 bb:1.0540 rl:2.2928 rb:1.0481 dl:1308-1314 gd:1 +ttp: b626/782 bl:2.3098 bb:1.0263 rl:2.2931 rb:1.0476 dl:1260-1265 gd:1 +ttp: b618/782 bl:2.4023 bb:1.0692 rl:2.2953 rb:1.0480 dl:1216-1221 gd:1 +ttp: b611/782 bl:2.2885 bb:1.0219 rl:2.2951 rb:1.0475 dl:1182-1186 gd:1 +ttp: b604/782 bl:2.3739 bb:1.0420 rl:2.2965 rb:1.0474 dl:1150-1154 gd:1 +ttp: b592/782 bl:2.2157 bb:0.9893 rl:2.2952 rb:1.0465 dl:1098-1103 gd:1 +ttp: b591/782 bl:2.3027 bb:1.0305 rl:2.2953 rb:1.0462 dl:1093-1098 gd:1 +ttp: b583/782 bl:2.3166 bb:1.0294 rl:2.2956 rb:1.0459 dl:1060-1064 gd:1 +ttp: b569/782 bl:2.3008 bb:1.0404 rl:2.2957 rb:1.0458 dl:1007-1010 gd:1 +ttp: b560/782 bl:2.2577 bb:1.0047 rl:2.2952 rb:1.0453 dl:975-979 gd:1 +ttp: b552/782 bl:2.2661 bb:1.0152 rl:2.2948 rb:1.0448 dl:949-952 gd:1 +ttp: b544/782 bl:2.3385 bb:1.0656 rl:2.2954 rb:1.0451 dl:924-927 gd:1 +ttp: b536/782 bl:2.3119 bb:1.0411 rl:2.2956 rb:1.0451 dl:899-902 gd:1 +ttp: b531/782 bl:2.2899 bb:1.0395 rl:2.2955 rb:1.0450 dl:884-887 gd:1 +ttp: b524/782 bl:2.3666 bb:1.0633 rl:2.2963 rb:1.0452 dl:863-866 gd:1 +ttp: b513/782 bl:2.3577 bb:1.0351 rl:2.2970 rb:1.0451 dl:832-835 gd:1 +ttp: b505/782 bl:2.3258 bb:1.0636 rl:2.2973 rb:1.0453 dl:809-812 gd:1 +ttp: b499/782 bl:2.3306 bb:1.0523 rl:2.2976 rb:1.0454 dl:794-796 gd:1 +ttp: b491/782 bl:2.2729 bb:1.0252 rl:2.2974 rb:1.0452 dl:773-776 gd:1 +ttp: b482/782 bl:2.3255 bb:1.0454 rl:2.2977 rb:1.0452 dl:752-754 gd:1 +ttp: b475/782 bl:2.3600 bb:1.0531 rl:2.2982 rb:1.0452 dl:735-737 gd:1 +ttp: b467/782 bl:2.3405 bb:1.0490 rl:2.2986 rb:1.0453 dl:717-719 gd:1 +ttp: b460/782 bl:2.2470 bb:1.0512 rl:2.2982 rb:1.0453 dl:701-703 gd:1 +ttp: b452/782 bl:2.2556 bb:1.0095 rl:2.2978 rb:1.0450 dl:685-687 gd:1 +ttp: b445/782 bl:2.3545 bb:1.0465 rl:2.2983 rb:1.0450 dl:670-672 gd:1 +ttp: b438/782 bl:2.2994 bb:1.0494 rl:2.2983 rb:1.0451 dl:655-657 gd:1 +ttp: b428/782 bl:2.2985 bb:1.0474 rl:2.2983 rb:1.0451 dl:636-638 gd:1 +ttp: b420/782 bl:2.3530 bb:1.0504 rl:2.2987 rb:1.0451 dl:620-622 gd:1 +ttp: b414/782 bl:2.1960 bb:1.0054 rl:2.2979 rb:1.0448 dl:609-611 gd:1 +ttp: b407/782 bl:2.2616 bb:1.0354 rl:2.2977 rb:1.0448 dl:595-597 gd:1 +ttp: b399/782 bl:2.2772 bb:1.0277 rl:2.2976 rb:1.0447 dl:581-582 gd:1 +ttp: b387/782 bl:2.3537 bb:1.0796 rl:2.2979 rb:1.0449 dl:559-561 gd:1 +ttp: b379/782 bl:2.4182 bb:1.0871 rl:2.2987 rb:1.0451 dl:545-547 gd:1 +ttp: b371/782 bl:2.2517 bb:1.0995 rl:2.2984 rb:1.0455 dl:532-533 gd:1 +ttp: b361/782 bl:2.3410 bb:1.0929 rl:2.2986 rb:1.0457 dl:515-517 gd:1 +ttp: b353/782 bl:2.1873 bb:1.0003 rl:2.2980 rb:1.0455 dl:501-503 gd:1 +ttp: b345/782 bl:2.3527 bb:1.0710 rl:2.2983 rb:1.0456 dl:489-491 gd:1 +ttp: b337/782 bl:2.3094 bb:1.0509 rl:2.2984 rb:1.0456 dl:477-478 gd:1 +ttp: b333/782 bl:2.4254 bb:1.0795 rl:2.2990 rb:1.0458 dl:471-472 gd:1 +ttp: b325/782 bl:2.3440 bb:1.0782 rl:2.2993 rb:1.0460 dl:459-461 gd:1 +ttp: b316/782 bl:2.3645 bb:1.0786 rl:2.2996 rb:1.0461 dl:445-446 gd:1 +ttp: b308/782 bl:2.3974 bb:1.0874 rl:2.3000 rb:1.0463 dl:433-435 gd:1 +ttp: b297/782 bl:2.3928 bb:1.0811 rl:2.3005 rb:1.0465 dl:417-418 gd:1 +ttp: b289/782 bl:2.3138 bb:1.0761 rl:2.3005 rb:1.0466 dl:405-406 gd:1 +ttp: b281/782 bl:2.2914 bb:1.0863 rl:2.3005 rb:1.0468 dl:394-395 gd:1 +ttp: b273/782 bl:2.3240 bb:1.0713 rl:2.3006 rb:1.0469 dl:383-384 gd:1 +ttp: b268/782 bl:2.3487 bb:1.0730 rl:2.3008 rb:1.0470 dl:376-378 gd:1 +ttp: b258/782 bl:2.4435 bb:1.0964 rl:2.3013 rb:1.0472 dl:364-365 gd:1 +ttp: b250/782 bl:2.3067 bb:1.0695 rl:2.3013 rb:1.0473 dl:354-355 gd:1 +ttp: b241/782 bl:2.3352 bb:1.0853 rl:2.3015 rb:1.0474 dl:342-344 gd:1 +ttp: b233/782 bl:2.3602 bb:1.1276 rl:2.3017 rb:1.0477 dl:333-334 gd:1 +ttp: b226/782 bl:2.3614 bb:1.0952 rl:2.3019 rb:1.0478 dl:324-325 gd:1 +ttp: b217/782 bl:2.3644 bb:1.1288 rl:2.3021 rb:1.0481 dl:314-315 gd:1 +ttp: b209/782 bl:2.4043 bb:1.1246 rl:2.3024 rb:1.0483 dl:305-306 gd:1 +ttp: b201/782 bl:2.2876 bb:1.0912 rl:2.3024 rb:1.0484 dl:297-298 gd:1 +ttp: b193/782 bl:2.3553 bb:1.1295 rl:2.3025 rb:1.0487 dl:288-289 gd:1 +ttp: b190/782 bl:2.3374 bb:1.0746 rl:2.3026 rb:1.0487 dl:284-285 gd:1 +ttp: b181/782 bl:2.3278 bb:1.1241 rl:2.3027 rb:1.0489 dl:275-276 gd:1 +ttp: b171/782 bl:2.4584 bb:1.1337 rl:2.3031 rb:1.0492 dl:266-266 gd:1 +ttp: b165/782 bl:2.3446 bb:1.1134 rl:2.3032 rb:1.0493 dl:260-260 gd:1 +ttp: b155/782 bl:2.4005 bb:1.1098 rl:2.3035 rb:1.0495 dl:250-251 gd:1 +ttp: b146/782 bl:2.4504 bb:1.1708 rl:2.3038 rb:1.0498 dl:241-242 gd:1 +ttp: b136/782 bl:2.4043 bb:1.1305 rl:2.3041 rb:1.0500 dl:232-233 gd:1 +ttp: b129/782 bl:2.3755 bb:1.1381 rl:2.3042 rb:1.0501 dl:225-226 gd:1 +ttp: b122/782 bl:2.4046 bb:1.1384 rl:2.3044 rb:1.0503 dl:219-219 gd:1 +ttp: b113/782 bl:2.5478 bb:1.1328 rl:2.3050 rb:1.0505 dl:210-211 gd:1 +ttp: b105/782 bl:2.4108 bb:1.1466 rl:2.3052 rb:1.0507 dl:203-204 gd:1 +ttp: b97/782 bl:2.4457 bb:1.1575 rl:2.3055 rb:1.0509 dl:196-197 gd:1 +ttp: b90/782 bl:2.4697 bb:1.2093 rl:2.3058 rb:1.0512 dl:190-190 gd:1 +ttp: b82/782 bl:2.4847 bb:1.1826 rl:2.3061 rb:1.0514 dl:183-183 gd:1 +ttp: b73/782 bl:2.5372 bb:1.2453 rl:2.3065 rb:1.0517 dl:174-175 gd:1 +ttp: b66/782 bl:2.6233 bb:1.2276 rl:2.3070 rb:1.0520 dl:169-169 gd:1 +ttp: b57/782 bl:2.4632 bb:1.1599 rl:2.3073 rb:1.0522 dl:160-161 gd:1 +ttp: b49/782 bl:2.4455 bb:1.1629 rl:2.3075 rb:1.0523 dl:152-153 gd:1 +ttp: b41/782 bl:2.5353 bb:1.2155 rl:2.3078 rb:1.0526 dl:144-145 gd:1 +ttp: b35/782 bl:2.6151 bb:1.2686 rl:2.3082 rb:1.0528 dl:138-139 gd:1 +ttp: b27/782 bl:2.5872 bb:1.2231 rl:2.3086 rb:1.0531 dl:130-131 gd:1 +ttp: b19/782 bl:2.6214 bb:1.2037 rl:2.3090 rb:1.0532 dl:121-122 gd:1 +ttp: b11/782 bl:2.6146 bb:1.2090 rl:2.3093 rb:1.0534 dl:109-110 gd:1 +ttp: b2/782 bl:2.8225 bb:1.2405 rl:2.3097 rb:1.0536 dl:83-89 gd:1 +quantized_ttt_phased val_loss:2.31487693 val_bpb:1.05780842 eval_time:519467ms +total_eval_time:519.5s