Skip to content

Eval bug: cmd_child_to_router:error:G:\xlam2\atomic-llama-cpp-turboquant\ggml\src\ggml-cuda\rope.cu:531: GGML_ASSERT(src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16) failed #75

Description

@HyperRamzey

Name and Version

.\llama-server.exe --version
version: b10269-1.5.1 (build 10696, commit cd56093)
built with Clang 22.1.8 for Windows AMD64

Operating systems

Windows

GGML backends

CUDA

Hardware

Ryzen 7 5700x3D + RTX 5070 + 56 gb ram

Models

Ling-3.0-flash-AD-IQ3_S-00001-of-00002.gguf

Problem description & steps to reproduce

At some point during generation it crashes with cmd_child_to_router:error:G:\xlam2\atomic-llama-cpp-turboquant\ggml\src\ggml-cuda\rope.cu:531: GGML_ASSERT(src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16) failed randomly.
Running with:
cat '.!start - ling.py'
import os
import subprocess
import sys

env = os.environ.copy()

cmd = [
r".\llama-server.exe",
"-m", r"D:\Ling-3.0-flash-AD-IQ3_S-00001-of-00002.gguf",
"--reasoning-preserve",
"--reasoning-budget", "32000",
"--reasoning-budget-message", "I've thought enough. Answering now with what I have.",
"--fit", "on",
"--n-cpu-moe", "40", "-bs",
"-lm", "mlock",
"-c", "120000",
"--cache-type-k", "bf16",
"--cache-type-v", "bf16",
"-np", "1",
"-fa", "on",
"-t", "8",
"-tb", "8",
"-b", "3008",
"-ub", "3008",
"--jinja",
"-kvu",
"--temp", "0.6",
"--top-p", "0.95",
"--top-k", "20",
"--samplers", "top_k;top_p;temperature",
"--alias", "ling-flash-3.0",
"--cache-reuse", "256",
"--cache-ram", "1024",
"--host", "127.0.0.1",
"--port", "8080",
"--verbosity", "4",
"--chat-template-file", r"G:\xlam3\chat_template_ling.jinja",
]

if name == "main":
try:
print("Starting llama-server...")
subprocess.run(cmd, env=env, check=True)
except KeyboardInterrupt:
print("\nServer stopped by user.")
except Exception as e:
print(f"\nError running server: {e}")

#removing --cache-reuse avoids it

First Bad Commit

No response

Relevant log output

Logs
'.\!start - ling.py'
Starting llama-server...
0.00.244.436 I cmn  common_param: common_params_print_info: build 10696 (cd5609390) with Clang 22.1.8 for Windows AMD64
0.00.244.439 I cmn  common_param: common_params_print_info: verbosity = 4 (adjust with the `-lv N` CLI arg)
0.00.244.441 I cmn  common_param: device_info:
0.00.311.086 I cmn  common_param:   - CUDA0   : NVIDIA GeForce RTX 5070 (12226 MiB, 11035 MiB free)
0.00.311.094 I cmn  common_param:   - CPU     : AMD Ryzen 7 5700X3D 8-Core Processor            (57244 MiB, 51295 MiB free)
0.00.325.146 I cmn  common_param: system_info: n_threads = 8 (n_threads_batch = 8) / 16 | CUDA : ARCHS = 1200 | FORCE_MMQ = 1 | USE_GRAPHS = 1 | FA_ALL_QUANTS = 1 | BLACKWELL_NATIVE_FP4 = 1 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 |
0.00.343.834 I srv          init: using 15 threads for HTTP server
0.00.347.088 W srv  llama_server: -----------------
0.00.347.091 W srv  llama_server: CORS is set to allow all origins ('*') and no API key is set
0.00.347.092 W srv  llama_server: this can be a security risk (cross-origin attacks)
0.00.347.092 W srv  llama_server: more info: https://github.com/ggml-org/llama.cpp/pull/25655
0.00.347.093 W srv  llama_server: -----------------
0.00.348.990 I srv         start: binding port with default address family
0.00.351.847 I srv    load_model: loading model 'D:\Ling-3.0-flash-AD-IQ3_S-00001-of-00002.gguf'
0.00.351.850 I srv    load_model: local path 'D:\Ling-3.0-flash-AD-IQ3_S-00001-of-00002.gguf'
0.00.352.132 I cmn  common_init_: fitting params to device memory ...
0.00.352.135 I cmn  common_init_: (for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on)
0.00.352.645 I common_params_fit_impl: getting device memory data for initial parameters:
0.00.776.765 I common_memory_breakdown_print: | memory breakdown [MiB] | total    free     self   model   context   compute    unaccounted |
0.00.776.768 I common_memory_breakdown_print: |   - CUDA0 (RTX 5070)   | 12226 = 10959 + ( 9581 =  6831 +     998 +    1752) +       -8314 |
0.00.776.769 I common_memory_breakdown_print: |   - Host               |                  49057 = 48310 +       0 +     747                |
0.00.820.808 I common_params_fit_impl: projected to use 9581 MiB of device memory vs. 10959 MiB of free device memory
0.00.820.811 I common_params_fit_impl: will leave 1377 >= 1024 MiB of free device memory, no changes needed
0.00.820.937 I common_fit_params: successfully fit params to free device memory
0.00.820.939 I common_fit_params: fitting params to free memory took 0.47 seconds
0.00.867.519 I llama_model_loader: additional 1 GGUFs metadata loaded.
0.00.867.523 I llama_model_loader: loaded meta data with 58 key-value pairs and 917 tensors from D:\Ling-3.0-flash-AD-IQ3_S-00001-of-00002.gguf (version GGUF V3 (latest))
0.00.867.547 I llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
0.00.867.549 I llama_model_loader: - kv   0:                       general.architecture str              = bailingmoe3
0.00.867.550 I llama_model_loader: - kv   1:                               general.type str              = model
0.00.867.551 I llama_model_loader: - kv   2:                               general.name str              = Ling 3.0 Flash
0.00.867.551 I llama_model_loader: - kv   3:                         general.size_label str              = 512x3.9B
0.00.867.552 I llama_model_loader: - kv   4:                            general.license str              = mit
0.00.867.560 I llama_model_loader: - kv   5:                               general.tags arr[str,1]       = ["text-generation"]
0.00.867.561 I llama_model_loader: - kv   6:                    bailingmoe3.block_count u32              = 42
0.00.867.562 I llama_model_loader: - kv   7:                 bailingmoe3.context_length u32              = 131072
0.00.867.563 I llama_model_loader: - kv   8:               bailingmoe3.embedding_length u32              = 2560
0.00.867.563 I llama_model_loader: - kv   9:            bailingmoe3.feed_forward_length u32              = 6144
0.00.867.564 I llama_model_loader: - kv  10:           bailingmoe3.attention.head_count u32              = 32
0.00.867.570 I llama_model_loader: - kv  11:        bailingmoe3.attention.head_count_kv arr[i32,42]      = [0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, ...
0.00.867.573 I llama_model_loader: - kv  12:                 bailingmoe3.rope.freq_base f32              = 6000000.000000
0.00.867.575 I llama_model_loader: - kv  13: bailingmoe3.attention.layer_norm_rms_epsilon f32              = 0.000001
0.00.867.575 I llama_model_loader: - kv  14:                   bailingmoe3.expert_count u32              = 512
0.00.867.576 I llama_model_loader: - kv  15:              bailingmoe3.expert_used_count u32              = 8
0.00.867.576 I llama_model_loader: - kv  16:             bailingmoe3.expert_group_count u32              = 8
0.00.867.576 I llama_model_loader: - kv  17:        bailingmoe3.expert_group_used_count u32              = 4
0.00.867.577 I llama_model_loader: - kv  18:             bailingmoe3.expert_gating_func u32              = 2
0.00.867.577 I llama_model_loader: - kv  19:           bailingmoe3.attention.key_length u32              = 576
0.00.867.578 I llama_model_loader: - kv  20:         bailingmoe3.attention.value_length u32              = 512
0.00.867.578 I llama_model_loader: - kv  21:                     bailingmoe3.vocab_size u32              = 157184
0.00.867.579 I llama_model_loader: - kv  22:                bailingmoe3.ssm.conv_kernel u32              = 4
0.00.867.579 I llama_model_loader: - kv  23:                   bailingmoe3.kda.head_dim u32              = 128
0.00.867.580 I llama_model_loader: - kv  24:         bailingmoe3.attention.kv_lora_rank u32              = 512
0.00.867.580 I llama_model_loader: - kv  25:       bailingmoe3.attention.key_length_mla u32              = 192
0.00.867.581 I llama_model_loader: - kv  26:     bailingmoe3.attention.value_length_mla u32              = 128
0.00.867.581 I llama_model_loader: - kv  27:           bailingmoe3.rope.dimension_count u32              = 64
0.00.867.582 I llama_model_loader: - kv  28:      bailingmoe3.leading_dense_block_count u32              = 2
0.00.867.582 I llama_model_loader: - kv  29:     bailingmoe3.expert_feed_forward_length u32              = 768
0.00.867.583 I llama_model_loader: - kv  30: bailingmoe3.expert_shared_feed_forward_length u32              = 768
0.00.867.583 I llama_model_loader: - kv  31:            bailingmoe3.expert_shared_count u32              = 1
0.00.867.584 I llama_model_loader: - kv  32:           bailingmoe3.expert_weights_scale f32              = 2.500000
0.00.867.585 I llama_model_loader: - kv  33:            bailingmoe3.expert_weights_norm bool             = true
0.00.867.585 I llama_model_loader: - kv  34:           bailingmoe3.nextn_predict_layers u32              = 1
0.00.867.586 I llama_model_loader: - kv  35:                       tokenizer.ggml.model str              = gpt2
0.00.867.586 I llama_model_loader: - kv  36:                         tokenizer.ggml.pre str              = bailingmoe2
0.00.895.333 I llama_model_loader: - kv  37:                      tokenizer.ggml.tokens arr[str,157184]  = ["!", "\"", "#", "$", "%", "&", "'", ...
0.00.904.601 I llama_model_loader: - kv  38:                  tokenizer.ggml.token_type arr[i32,157184]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
0.00.932.119 I llama_model_loader: - kv  39:                      tokenizer.ggml.merges arr[str,156635]  = ["Ġ Ġ", "Ġ t", "i n", "Ġ a", "h e...
0.00.932.124 I llama_model_loader: - kv  40:                tokenizer.ggml.bos_token_id u32              = 156891
0.00.932.126 I llama_model_loader: - kv  41:                tokenizer.ggml.eos_token_id u32              = 156895
0.00.932.127 I llama_model_loader: - kv  42:            tokenizer.ggml.padding_token_id u32              = 156892
0.00.932.128 I llama_model_loader: - kv  43:               tokenizer.ggml.add_bos_token bool             = false
0.00.932.128 I llama_model_loader: - kv  44:               tokenizer.ggml.add_eos_token bool             = false
0.00.932.132 I llama_model_loader: - kv  45:                    tokenizer.chat_template str              = {#- Bailing V3 chat template -#}\n{#- ...
0.00.932.135 I llama_model_loader: - kv  46:           bailingmoe3.kda.gate_lower_bound f32              = -5.000000
0.00.932.135 I llama_model_loader: - kv  47:               general.quantization_version u32              = 2
0.00.932.136 I llama_model_loader: - kv  48:                          general.file_type u32              = 26
0.00.932.137 I llama_model_loader: - kv  49:                      quantize.imatrix.file str              = /root/work/ling3-imatrix-bf16.gguf
0.00.932.137 I llama_model_loader: - kv  50:                   quantize.imatrix.dataset str              = /root/work/calib/calib_train.txt
0.00.932.138 I llama_model_loader: - kv  51:             quantize.imatrix.entries_count u32              = 573
0.00.932.138 I llama_model_loader: - kv  52:              quantize.imatrix.chunks_count u32              = 522
0.00.932.139 I llama_model_loader: - kv  53:                                   split.no u16              = 0
0.00.932.140 I llama_model_loader: - kv  54:                        split.tensors.count i32              = 917
0.00.932.140 I llama_model_loader: - kv  55:                                split.count u16              = 2
0.00.932.158 I llama_model_loader: - kv  56:               bailingmoe3.swiglu_clamp_exp arr[f32,42]      = [0.000000, 0.000000, 0.000000, 0.0000...
0.00.932.173 I llama_model_loader: - kv  57:             bailingmoe3.swiglu_clamp_shexp arr[f32,42]      = [0.000000, 0.000000, 0.000000, 0.0000...
0.00.932.174 I llama_model_loader: - type  f32:  382 tensors
0.00.932.174 I llama_model_loader: - type  f16:    1 tensors
0.00.932.175 I llama_model_loader: - type q8_0:  414 tensors
0.00.932.175 I llama_model_loader: - type iq3_s:  105 tensors
0.00.932.176 I llama_model_loader: - type iq4_xs:   15 tensors
0.00.932.176 I print_info: file format = GGUF V3 (latest)
0.00.932.177 I print_info: file type   = IQ3_S - 3.4375 bpw
0.00.932.178 I print_info: file size   = 53.85 GiB (3.72 BPW)
0.00.932.210 I llama_prepare_model_devices: using device CUDA0 (NVIDIA GeForce RTX 5070) (0000:06:00.0) - 11035 MiB free
0.00.997.223 I load: 0 unused tokens
0.01.004.737 W load: special_eos_id is not in special_eog_ids - the tokenizer config may be incorrect
0.01.004.739 I load: printing all EOG tokens:
0.01.004.739 I load:   - 156892 ('<|endoftext|>')
0.01.004.740 I load:   - 156895 ('<|role_end|>')
0.01.004.882 I load: special tokens cache size = 262
0.01.031.211 I load: token to piece cache size = 1.0009 MB
0.01.031.220 I print_info: arch                  = bailingmoe3
0.01.031.221 I print_info: vocab_only            = 0
0.01.031.221 I print_info: no_alloc              = 0
0.01.031.222 I print_info: n_ctx_train           = 131072
0.01.031.222 I print_info: n_embd_inp            = 2560
0.01.031.223 I print_info: n_embd                = 2560
0.01.031.223 I print_info: n_embd_out            = 2560
0.01.031.224 I print_info: n_layer               = 42
0.01.031.224 I print_info: n_layer_all           = 42
0.01.031.232 I print_info: n_head                = 32
0.01.031.243 I print_info: n_head_kv             = [0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1]
0.01.031.243 I print_info: n_rot                 = 64
0.01.031.244 I print_info: n_swa                 = 0
0.01.031.244 I print_info: is_swa_any            = 0
0.01.031.245 I print_info: n_embd_head_k         = 576
0.01.031.245 I print_info: n_embd_head_v         = 512
0.01.031.255 I print_info: n_gqa                 = [0, 0, 0, 0, 0, 32, 0, 0, 0, 0, 0, 32, 0, 0, 0, 0, 0, 32, 0, 0, 0, 0, 0, 32, 0, 0, 0, 0, 0, 32, 0, 0, 0, 0, 0, 32, 0, 0, 0, 0, 0, 32]
0.01.031.265 I print_info: n_embd_k_gqa          = [0, 0, 0, 0, 0, 576, 0, 0, 0, 0, 0, 576, 0, 0, 0, 0, 0, 576, 0, 0, 0, 0, 0, 576, 0, 0, 0, 0, 0, 576, 0, 0, 0, 0, 0, 576, 0, 0, 0, 0, 0, 576]
0.01.031.274 I print_info: n_embd_v_gqa          = [0, 0, 0, 0, 0, 512, 0, 0, 0, 0, 0, 512, 0, 0, 0, 0, 0, 512, 0, 0, 0, 0, 0, 512, 0, 0, 0, 0, 0, 512, 0, 0, 0, 0, 0, 512, 0, 0, 0, 0, 0, 512]
0.01.031.276 I print_info: f_norm_eps            = 0.0e+00
0.01.031.277 I print_info: f_norm_rms_eps        = 1.0e-06
0.01.031.277 I print_info: f_clamp_kqv           = 0.0e+00
0.01.031.278 I print_info: f_max_alibi_bias      = 0.0e+00
0.01.031.278 I print_info: f_logit_scale         = 0.0e+00
0.01.031.279 I print_info: f_attn_scale          = 0.0e+00
0.01.031.279 I print_info: f_attn_value_scale    = 0.0000
0.01.031.281 I print_info: n_ff                  = 6144
0.01.031.282 I print_info: n_expert              = 512
0.01.031.282 I print_info: n_expert_used         = 8
0.01.031.282 I print_info: n_expert_groups       = 8
0.01.031.283 I print_info: n_group_used          = 4
0.01.031.283 I print_info: causal attn           = 1
0.01.031.284 I print_info: pooling type          = -1
0.01.031.284 I print_info: rope type             = 0
0.01.031.285 I print_info: rope scaling          = linear
0.01.031.286 I print_info: freq_base_train       = 6000000.0
0.01.031.287 I print_info: freq_scale_train      = 1
0.01.031.287 I print_info: n_ctx_orig_yarn       = 131072
0.01.031.288 I print_info: rope_yarn_log_mul     = 0.0000
0.01.031.288 I print_info: rope_finetuned        = unknown
0.01.031.289 I print_info: model type            = 124B.A5B
0.01.031.290 I print_info: model params          = 124.41 B
0.01.031.299 I print_info: general.name          = Ling 3.0 Flash
0.01.031.300 I print_info: vocab type            = BPE
0.01.031.300 I print_info: n_vocab               = 157184
0.01.031.300 I print_info: n_merges              = 156635
0.01.031.301 I print_info: BOS token             = 156891 '<|startoftext|>'
0.01.031.301 I print_info: EOS token             = 156895 '<|role_end|>'
0.01.031.302 I print_info: EOT token             = 156892 '<|endoftext|>'
0.01.031.302 I print_info: PAD token             = 156892 '<|endoftext|>'
0.01.031.303 I print_info: LF token              = 198 'Ċ'
0.01.031.304 I print_info: EOG token             = 156892 '<|endoftext|>'
0.01.031.304 I print_info: EOG token             = 156895 '<|role_end|>'
0.01.031.304 I print_info: max token length      = 154
0.01.031.305 I load_tensors: loading model tensors, this can take a while... (load_mode = mlock)
0.02.804.333 I load_tensors: offloading output layer to GPU
0.02.804.338 I load_tensors: offloading 41 repeating layers to GPU
0.02.804.339 I load_tensors: offloaded 43/43 layers to GPU
0.02.804.346 I load_tensors:          CPU model buffer size =     0.00 MiB
0.02.804.347 I load_tensors:        CUDA0 model buffer size =  6831.14 MiB
0.02.804.351 I load_tensors:          CPU model buffer size = 48310.23 MiB
0.51.627.617 I cmn  common_init_: added <|endoftext|> logit bias = -inf
0.51.628.585 I cmn  common_init_: added <|role_end|> logit bias = -inf
0.51.762.166 I llama_context: constructing llama_context
0.52.029.466 I llama_context: setting backend sampler for seq_id 0 (n = 4)
0.52.029.486 I llama_context: n_seq_max     = 1
0.52.029.486 I llama_context: n_ctx         = 120064
0.52.029.487 I llama_context: n_ctx_seq     = 120064
0.52.029.487 I llama_context: n_batch       = 3008
0.52.029.487 I llama_context: n_ubatch      = 3008
0.52.029.490 I llama_context: causal_attn   = 1
0.52.033.765 I llama_context: flash_attn    = enabled
0.52.033.766 I llama_context: kv_unified    = true
0.52.033.780 I llama_context: freq_base     = 6000000.0
0.52.033.783 I llama_context: freq_scale    = 1
0.52.033.784 I llama_context: n_rs_seq      = 0
0.52.033.784 I llama_context: n_outputs_max = 1
0.52.033.785 I llama_context: n_ctx_seq (120064) < n_ctx_train (131072) -- the full capacity of the model will not be utilized
0.52.119.219 I llama_context:  CUDA_Host  output buffer size =     2.40 MiB
0.52.132.088 I llama_kv_cache:      CUDA0 KV buffer size =   923.34 MiB
0.52.137.746 I llama_kv_cache: size =  923.34 MiB (120064 cells,   7 layers,  1/1 seqs), K (bf16):  923.34 MiB, V (bf16):    0.00 MiB
0.52.137.755 I llama_kv_cache: attn_rot_k = 0, n_embd_head_k_all = 576
0.52.137.756 I llama_kv_cache: attn_rot_v = 0, n_embd_head_k_all = 0
0.52.138.682 I llama_memory_recurrent:      CUDA0 RS buffer size =    74.92 MiB
0.52.138.691 I llama_memory_recurrent: size =   74.92 MiB (     1 cells,  42 layers,  1 seqs  0 rs_seq), R (f32):    4.92 MiB, S (f32):   70.00 MiB
0.52.138.701 I sched_reserve: reserving ...
0.52.161.354 I resolve_fused_ops: resolving fused Gated Delta Net support:
0.52.262.130 I resolve_fused_ops: fused Gated Delta Net (autoregressive) enabled
0.52.263.957 I resolve_fused_ops: fused Gated Delta Net (chunked) enabled
0.52.263.960 I resolve_fused_ops: resolving fused Lightning Indexer support:
0.52.265.460 I resolve_fused_ops: Lightning Indexer enabled
0.52.265.463 I resolve_fused_ops: resolving fused DeepSeek V4 HC support:
0.52.266.898 I resolve_fused_ops: fused DeepSeek V4 HC pre enabled
0.52.268.373 I resolve_fused_ops: fused DeepSeek V4 HC comb enabled
0.52.269.889 I resolve_fused_ops: fused DeepSeek V4 HC post enabled
0.52.465.166 I sched_reserve:      CUDA0 compute buffer size =  1752.34 MiB
0.52.465.170 I sched_reserve:  CUDA_Host compute buffer size =   747.64 MiB
0.52.465.171 I sched_reserve: graph nodes  = 5669
0.52.465.187 I sched_reserve: graph splits = 116 (with bs=3008), 78 (with bs=1)
0.52.465.190 I sched_reserve: reserve took 326.49 ms, sched copies = 1
0.52.490.567 I cmn  common_init_: warming up the model with an empty run - please wait ... (--no-warmup to disable)
1.13.268.561 I cmn  common_conte: the context does not support partial sequence removal
1.13.283.826 I srv    load_model: speculative decoding will use checkpoints
1.13.283.831 I srv    load_model: initializing, n_slots = 1, n_ctx_slot = 120064, kv_unified = 'true'
1.13.286.153 I spec common_specu: no implementations specified for speculative decoding
1.13.286.161 I slot   load_model: id  0 | task -1 | new slot, n_ctx = 120064
1.13.296.273 I srv    load_model: prompt cache is enabled, size limit: 1024 MiB
1.13.296.281 I srv    load_model: use `--cache-ram 0` to disable the prompt cache
1.13.296.282 I srv    load_model: for more info see https://github.com/ggml-org/llama.cpp/pull/16391
1.13.296.283 I srv    load_model: context checkpoints enabled, max = 32, min spacing = 8192
1.13.310.516 I srv          init: idle slots will be saved to prompt cache and cleared upon starting a new task
1.13.515.846 I srv          init: init: chat template, example_format: '<role>SYSTEM</role>You are a helpful assistant
detailed thinking on<|role_end|><role>HUMAN</role>Hello<|role_end|><role>ASSISTANT</role>
<think></think>Hi there<|role_end|><role>HUMAN</role>How are you?<|role_end|><role>ASSISTANT</role>
<think>'
1.13.538.501 I srv          init: init: chat template, thinking = 1
1.13.567.405 I srv  llama_server: model loaded
1.13.567.417 I srv  llama_server: listening on http://127.0.0.1:8080
1.13.567.417 W srv  llama_server: NOTICE: server default port will be changed to :9931 in a future release
1.13.567.417 W srv  llama_server:         ref: https://github.com/ggml-org/llama.cpp/pull/26508
1.13.574.165 I srv  update_slots: all slots are idle
1.50.907.477 I srv  server_strea: conv_id= (empty=1)
1.51.531.318 I srv    operator(): chat format: peg-native
1.51.547.703 I slot get_availabl: id  0 | task -1 |  - skipping, slot is empty
1.51.547.709 I slot get_availabl: id  0 | task -1 | selected slot by LRU, t_last = -1
1.51.547.719 I srv  get_availabl: updating prompt cache
1.51.548.589 I srv          load:  - looking for better prompt, base f_keep = -1.000, f_sim = 0.000
1.51.548.594 I srv        update:  - cache state: 0 prompts, 0.000 MiB (limits: 1024.000 MiB, 120064 tokens, 1073741824 est)
1.51.548.596 I srv  get_availabl: prompt cache update took 0.87 ms
1.51.576.489 I cmn  common_reaso: activated, budget=32000 tokens
1.51.576.502 W common_sampler_init: backend sampling is not compatible with grammar, disabling
1.51.577.067 I slot launch_slot_: id  0 | task -1 | sampler chain: logits -> top-k -> top-p -> temp-ext -> dist
1.51.577.090 I slot launch_slot_: id  0 | task -1 | sampler params:
        repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
        dry_multiplier = 0.000, dry_base = 1.750, dry_allowed_length = 2, dry_penalty_last_n = 120064
        top_k = 20, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -1.000, temp = 0.600
        mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000, adaptive_target = -1.000, adaptive_decay = 0.900
1.51.577.093 I slot launch_slot_: id  0 | task 0 | processing task, is_child = 0
1.51.577.746 I slot   operator(): id  0 | task 0 | new prompt, n_ctx_slot = 120064, n_keep = 0, task.n_tokens = 75903
1.51.578.707 I slot   operator(): id  0 | task 0 | cached n_tokens = 0, memory_seq_rm [0, end)
1.51.592.963 I sched_reserve: reserving ...
1.51.769.011 I sched_reserve:      CUDA0 compute buffer size =  1752.34 MiB
1.51.769.015 I sched_reserve:  CUDA_Host compute buffer size =   747.64 MiB
1.51.769.016 I sched_reserve: graph nodes  = 5632
1.51.769.017 I sched_reserve: graph splits = 116 (with bs=3008), 78 (with bs=1)
1.51.769.020 I sched_reserve: reserve took 176.04 ms, sched copies = 1
1.57.351.622 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =   3008, progress = 0.04, t =   5.77 s / 520.97 tokens per second
1.57.351.626 I slot   operator(): id  0 | task 0 | cached n_tokens = 3008, memory_seq_rm [3008, end)
2.02.559.247 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =   6016, progress = 0.08, t =  10.98 s / 547.83 tokens per second
2.02.559.251 I slot   operator(): id  0 | task 0 | cached n_tokens = 6016, memory_seq_rm [6016, end)
2.07.244.385 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =   9024, progress = 0.12, t =  15.67 s / 576.00 tokens per second
2.07.244.389 I slot   operator(): id  0 | task 0 | cached n_tokens = 9024, memory_seq_rm [9024, end)
2.11.305.192 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  12032, progress = 0.16, t =  19.73 s / 609.91 tokens per second
2.11.305.196 I slot   operator(): id  0 | task 0 | cached n_tokens = 12032, memory_seq_rm [12032, end)
2.15.845.162 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  15040, progress = 0.20, t =  24.27 s / 619.76 tokens per second
2.15.845.166 I slot   operator(): id  0 | task 0 | cached n_tokens = 15040, memory_seq_rm [15040, end)
2.20.456.507 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  18048, progress = 0.24, t =  28.88 s / 624.96 tokens per second
2.20.456.511 I slot   operator(): id  0 | task 0 | cached n_tokens = 18048, memory_seq_rm [18048, end)
2.25.206.641 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  21056, progress = 0.28, t =  33.63 s / 626.13 tokens per second
2.25.206.644 I slot   operator(): id  0 | task 0 | cached n_tokens = 21056, memory_seq_rm [21056, end)
2.30.430.391 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  24064, progress = 0.32, t =  38.85 s / 619.37 tokens per second
2.30.430.397 I slot   operator(): id  0 | task 0 | cached n_tokens = 24064, memory_seq_rm [24064, end)
2.35.444.468 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  27072, progress = 0.36, t =  43.87 s / 617.14 tokens per second
2.35.444.475 I slot   operator(): id  0 | task 0 | cached n_tokens = 27072, memory_seq_rm [27072, end)
2.40.841.838 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  30080, progress = 0.40, t =  49.26 s / 610.59 tokens per second
2.40.841.843 I slot   operator(): id  0 | task 0 | cached n_tokens = 30080, memory_seq_rm [30080, end)
2.43.609.589 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  30445, progress = 0.40, t =  52.03 s / 585.12 tokens per second
2.43.609.593 I slot   operator(): id  0 | task 0 | cached n_tokens = 30445, memory_seq_rm [30445, end)
2.43.670.143 I slot create_check: id  0 | task 0 | created context checkpoint 1 of 32 (pos_min = 30444, pos_max = 30444, n_tokens = 30445, size = 74.923 MiB)
2.49.561.167 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  33453, progress = 0.44, t =  57.98 s / 576.94 tokens per second
2.49.561.172 I slot   operator(): id  0 | task 0 | cached n_tokens = 33453, memory_seq_rm [33453, end)
2.55.434.005 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  36461, progress = 0.48, t =  63.86 s / 570.99 tokens per second
2.55.434.009 I slot   operator(): id  0 | task 0 | cached n_tokens = 36461, memory_seq_rm [36461, end)
3.01.504.348 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  39469, progress = 0.52, t =  69.93 s / 564.43 tokens per second
3.01.504.355 I slot   operator(): id  0 | task 0 | cached n_tokens = 39469, memory_seq_rm [39469, end)
3.07.719.944 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  42477, progress = 0.56, t =  76.14 s / 557.86 tokens per second
3.07.719.952 I slot   operator(): id  0 | task 0 | cached n_tokens = 42477, memory_seq_rm [42477, end)
3.14.285.187 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  45485, progress = 0.60, t =  82.71 s / 549.95 tokens per second
3.14.285.193 I slot   operator(): id  0 | task 0 | cached n_tokens = 45485, memory_seq_rm [45485, end)
3.20.916.378 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  48493, progress = 0.64, t =  89.34 s / 542.80 tokens per second
3.20.916.382 I slot   operator(): id  0 | task 0 | cached n_tokens = 48493, memory_seq_rm [48493, end)
3.27.813.095 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  51501, progress = 0.68, t =  96.24 s / 535.16 tokens per second
3.27.813.100 I slot   operator(): id  0 | task 0 | cached n_tokens = 51501, memory_seq_rm [51501, end)
3.34.398.132 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  54509, progress = 0.72, t = 102.82 s / 530.14 tokens per second
3.34.398.140 I slot   operator(): id  0 | task 0 | cached n_tokens = 54509, memory_seq_rm [54509, end)
3.41.413.816 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  57517, progress = 0.76, t = 109.84 s / 523.66 tokens per second
3.41.413.822 I slot   operator(): id  0 | task 0 | cached n_tokens = 57517, memory_seq_rm [57517, end)
3.48.076.256 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  60525, progress = 0.80, t = 116.50 s / 519.53 tokens per second
3.48.076.262 I slot   operator(): id  0 | task 0 | cached n_tokens = 60525, memory_seq_rm [60525, end)
3.55.211.865 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  63533, progress = 0.84, t = 123.63 s / 513.88 tokens per second
3.55.211.871 I slot   operator(): id  0 | task 0 | cached n_tokens = 63533, memory_seq_rm [63533, end)
4.02.576.015 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  66541, progress = 0.88, t = 131.00 s / 507.95 tokens per second
4.02.576.032 I slot   operator(): id  0 | task 0 | cached n_tokens = 66541, memory_seq_rm [66541, end)
4.08.194.217 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  67918, progress = 0.89, t = 136.62 s / 497.14 tokens per second
4.08.194.221 I slot   operator(): id  0 | task 0 | cached n_tokens = 67918, memory_seq_rm [67918, end)
4.08.355.217 I slot create_check: id  0 | task 0 | created context checkpoint 2 of 32 (pos_min = 67917, pos_max = 67917, n_tokens = 67918, size = 74.923 MiB)
4.13.550.642 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  69558, progress = 0.92, t = 141.97 s / 489.94 tokens per second
4.13.550.645 I slot   operator(): id  0 | task 0 | cached n_tokens = 69558, memory_seq_rm [69558, end)
4.20.660.842 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  72566, progress = 0.96, t = 149.08 s / 486.75 tokens per second
4.20.660.846 I slot   operator(): id  0 | task 0 | cached n_tokens = 72566, memory_seq_rm [72566, end)
4.23.727.090 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  72895, progress = 0.96, t = 152.15 s / 479.10 tokens per second
4.23.727.094 I slot   operator(): id  0 | task 0 | cached n_tokens = 72895, memory_seq_rm [72895, end)
4.23.778.933 I slot create_check: id  0 | task 0 | created context checkpoint 3 of 32 (pos_min = 72894, pos_max = 72894, n_tokens = 72895, size = 74.923 MiB)
4.30.911.031 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  75890, progress = 1.00, t = 159.33 s / 476.30 tokens per second
4.30.911.035 I slot   operator(): id  0 | task 0 | cached n_tokens = 75890, memory_seq_rm [75890, end)
4.31.293.884 I slot create_check: id  0 | task 0 | created context checkpoint 4 of 32 (pos_min = 75889, pos_max = 75889, n_tokens = 75890, size = 74.923 MiB)
4.31.798.962 I slot print_timing: id  0 | task 0 | prompt processing, n_tokens =  75899, progress = 1.00, t = 160.22 s / 473.71 tokens per second
4.31.798.966 I slot   operator(): id  0 | task 0 | cached n_tokens = 75899, memory_seq_rm [75899, end)
4.31.819.267 I slot init_sampler: id  0 | task 0 | init sampler, took 8.09 ms, tokens: text = 75903, total = 75903
4.31.834.242 I slot create_check: id  0 | task 0 | created context checkpoint 5 of 32 (pos_min = 75898, pos_max = 75898, n_tokens = 75899, size = 74.923 MiB)
4.39.068.259 I slot print_timing: id  0 | task 0 | n_decoded =    100, tg =  14.36 t/s, tg_3s =  14.36 t/s
4.42.105.103 I slot print_timing: id  0 | task 0 | n_decoded =    141, tg =  14.10 t/s, tg_3s =  13.50 t/s
4.45.144.858 I slot print_timing: id  0 | task 0 | n_decoded =    187, tg =  14.34 t/s, tg_3s =  15.13 t/s
4.48.155.182 I slot print_timing: id  0 | task 0 | n_decoded =    232, tg =  14.45 t/s, tg_3s =  14.95 t/s
4.49.868.631 W srv          stop: cancel task, id_task = 0
4.49.888.188 I slot      release: id  0 | task 0 | stop processing: n_tokens = 76159, truncated = 0
4.49.888.207 I srv  update_slots: all slots are idle
4.58.638.030 I srv  server_strea: conv_id= (empty=1)
4.58.820.135 I srv    operator(): chat format: peg-native
4.58.828.895 I slot get_availabl: id  0 | task -1 |  - checking sim = 1.000 (75897/75918) > 0.100
4.58.828.900 I slot get_availabl: id  0 | task -1 | selected slot by LCP similarity, f_sim_best = 1.000 (> 0.100 thold), f_keep = 0.997
4.58.832.727 I cmn  common_reaso: activated, budget=32000 tokens
4.58.832.738 W common_sampler_init: backend sampling is not compatible with grammar, disabling
4.58.833.415 I slot launch_slot_: id  0 | task -1 | sampler chain: logits -> top-k -> top-p -> temp-ext -> dist
4.58.833.452 I slot launch_slot_: id  0 | task -1 | sampler params:
        repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
        dry_multiplier = 0.000, dry_base = 1.750, dry_allowed_length = 2, dry_penalty_last_n = 120064
        top_k = 20, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -1.000, temp = 0.600
        mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000, adaptive_target = -1.000, adaptive_decay = 0.900
4.58.833.455 I slot launch_slot_: id  0 | task 288 | processing task, is_child = 0
4.58.833.462 I slot   operator(): id  0 | task 288 | new prompt, n_ctx_slot = 120064, n_keep = 0, task.n_tokens = 75918
4.58.833.513 I slot   operator(): id  0 | task 288 | checking checkpoint with [75898, 75898] against 75897...
4.58.833.514 I slot   operator(): id  0 | task 288 | checking checkpoint with [75889, 75889] against 75897...
4.58.845.816 I slot   operator(): id  0 | task 288 | restored context checkpoint (pos_min = 75889, pos_max = 75889, n_tokens = 75890, n_past = 75890, size = 74.923 MiB)
4.58.845.821 I slot   operator(): id  0 | task 288 | erased invalidated context checkpoint (pos_min = 75898, pos_max = 75898, n_tokens = 75899, n_swa = 0, pos_next = 75890, size = 74.923 MiB)
4.58.847.759 I slot   operator(): id  0 | task 288 | cached n_tokens = 75890, memory_seq_rm [75890, end)
4.58.848.105 I slot create_check: id  0 | task 288 | erasing context checkpoint too close to an earlier one (pos_min = 72894, pos_max = 72894, n_tokens = 72895, size = 74.923 MiB)
4.58.854.349 I slot create_check: id  0 | task 288 | erasing context checkpoint too close to an earlier one (pos_min = 75889, pos_max = 75889, n_tokens = 75890, size = 74.923 MiB)
4.58.873.336 I slot create_check: id  0 | task 288 | created context checkpoint 3 of 32 (pos_min = 75889, pos_max = 75889, n_tokens = 75890, size = 74.923 MiB)
4.59.247.232 I slot   operator(): id  0 | task 288 | cached n_tokens = 75896, memory_seq_rm [75896, end)
4.59.267.573 I slot create_check: id  0 | task 288 | created context checkpoint 4 of 32 (pos_min = 75895, pos_max = 75895, n_tokens = 75896, size = 74.923 MiB)
5.00.345.694 I slot   operator(): id  0 | task 288 | cached n_tokens = 75914, memory_seq_rm [75914, end)
5.00.353.940 I slot init_sampler: id  0 | task 288 | init sampler, took 8.14 ms, tokens: text = 75918, total = 75918
5.00.368.968 I slot create_check: id  0 | task 288 | created context checkpoint 5 of 32 (pos_min = 75913, pos_max = 75913, n_tokens = 75914, size = 74.923 MiB)
5.07.264.198 I slot print_timing: id  0 | task 288 | n_decoded =    100, tg =  14.91 t/s, tg_3s =  14.91 t/s
5.07.459.876 I cmn  common_reaso: deactivated (natural end)
5.10.321.943 I slot print_timing: id  0 | task 288 | n_decoded =    144, tg =  14.75 t/s, tg_3s =  14.39 t/s
5.13.371.614 I slot print_timing: id  0 | task 288 | n_decoded =    189, tg =  14.75 t/s, tg_3s =  14.76 t/s
5.16.408.689 I slot print_timing: id  0 | task 288 | n_decoded =    234, tg =  14.76 t/s, tg_3s =  14.82 t/s
5.19.414.881 I slot print_timing: id  0 | task 288 | n_decoded =    279, tg =  14.80 t/s, tg_3s =  14.97 t/s
5.22.463.704 I slot print_timing: id  0 | task 288 | n_decoded =    324, tg =  14.79 t/s, tg_3s =  14.76 t/s
5.25.481.639 I slot print_timing: id  0 | task 288 | n_decoded =    369, tg =  14.81 t/s, tg_3s =  14.91 t/s
5.28.501.634 I slot print_timing: id  0 | task 288 | n_decoded =    413, tg =  14.78 t/s, tg_3s =  14.57 t/s
5.31.508.425 I slot print_timing: id  0 | task 288 | n_decoded =    458, tg =  14.80 t/s, tg_3s =  14.97 t/s
5.34.518.491 I slot print_timing: id  0 | task 288 | n_decoded =    503, tg =  14.81 t/s, tg_3s =  14.95 t/s
5.37.548.419 I slot print_timing: id  0 | task 288 | n_decoded =    547, tg =  14.79 t/s, tg_3s =  14.52 t/s
5.40.612.162 I slot print_timing: id  0 | task 288 | n_decoded =    592, tg =  14.78 t/s, tg_3s =  14.69 t/s
5.43.612.246 I slot print_timing: id  0 | task 288 | n_decoded =    635, tg =  14.75 t/s, tg_3s =  14.33 t/s
5.46.625.991 I slot print_timing: id  0 | task 288 | n_decoded =    680, tg =  14.76 t/s, tg_3s =  14.93 t/s
5.49.687.849 I slot print_timing: id  0 | task 288 | n_decoded =    727, tg =  14.80 t/s, tg_3s =  15.35 t/s
5.52.712.670 I slot print_timing: id  0 | task 288 | n_decoded =    773, tg =  14.82 t/s, tg_3s =  15.21 t/s
5.55.740.121 I slot print_timing: id  0 | task 288 | n_decoded =    820, tg =  14.86 t/s, tg_3s =  15.52 t/s
5.58.795.374 I slot print_timing: id  0 | task 288 | n_decoded =    866, tg =  14.87 t/s, tg_3s =  15.06 t/s
6.01.842.891 I slot print_timing: id  0 | task 288 | n_decoded =    912, tg =  14.88 t/s, tg_3s =  15.09 t/s
6.04.851.576 I slot print_timing: id  0 | task 288 | n_decoded =    959, tg =  14.92 t/s, tg_3s =  15.62 t/s
6.07.913.295 I slot print_timing: id  0 | task 288 | n_decoded =   1006, tg =  14.94 t/s, tg_3s =  15.35 t/s
6.10.970.907 I slot print_timing: id  0 | task 288 | n_decoded =   1053, tg =  14.95 t/s, tg_3s =  15.37 t/s
6.13.994.481 I slot print_timing: id  0 | task 288 | n_decoded =   1099, tg =  14.97 t/s, tg_3s =  15.21 t/s
6.16.999.322 I slot print_timing: id  0 | task 288 | n_decoded =   1145, tg =  14.98 t/s, tg_3s =  15.31 t/s
6.20.000.857 I slot print_timing: id  0 | task 288 | n_decoded =   1191, tg =  14.99 t/s, tg_3s =  15.33 t/s
6.23.009.168 I slot print_timing: id  0 | task 288 | n_decoded =   1237, tg =  15.00 t/s, tg_3s =  15.29 t/s
6.26.062.595 I slot print_timing: id  0 | task 288 | n_decoded =   1284, tg =  15.02 t/s, tg_3s =  15.39 t/s
6.29.123.967 I slot print_timing: id  0 | task 288 | n_decoded =   1331, tg =  15.03 t/s, tg_3s =  15.35 t/s
6.32.177.905 I slot print_timing: id  0 | task 288 | n_decoded =   1378, tg =  15.04 t/s, tg_3s =  15.39 t/s
6.35.206.323 I slot print_timing: id  0 | task 288 | n_decoded =   1424, tg =  15.05 t/s, tg_3s =  15.19 t/s
6.38.228.636 I slot print_timing: id  0 | task 288 | n_decoded =   1470, tg =  15.05 t/s, tg_3s =  15.22 t/s
6.41.234.895 I slot print_timing: id  0 | task 288 | n_decoded =   1517, tg =  15.07 t/s, tg_3s =  15.63 t/s
6.44.241.152 I slot print_timing: id  0 | task 288 | n_decoded =   1563, tg =  15.07 t/s, tg_3s =  15.30 t/s
6.47.284.318 I slot print_timing: id  0 | task 288 | n_decoded =   1610, tg =  15.09 t/s, tg_3s =  15.44 t/s
6.50.308.636 I slot print_timing: id  0 | task 288 | n_decoded =   1656, tg =  15.09 t/s, tg_3s =  15.21 t/s
6.53.310.106 I slot print_timing: id  0 | task 288 | n_decoded =   1702, tg =  15.10 t/s, tg_3s =  15.33 t/s
6.56.348.836 I slot print_timing: id  0 | task 288 | n_decoded =   1749, tg =  15.11 t/s, tg_3s =  15.47 t/s
6.59.374.086 I slot print_timing: id  0 | task 288 | n_decoded =   1795, tg =  15.11 t/s, tg_3s =  15.21 t/s
7.02.432.256 I slot print_timing: id  0 | task 288 | n_decoded =   1842, tg =  15.11 t/s, tg_3s =  15.37 t/s
7.03.663.676 I slot print_timing: id  0 | task 288 | prompt eval time =    1725.19 ms /    28 tokens (   61.61 ms per token,    16.23 tokens per second)
7.03.663.680 I slot print_timing: id  0 | task 288 |        eval time =  123105.01 ms /  1861 tokens (   66.15 ms per token,    15.12 tokens per second)
7.03.663.682 I slot print_timing: id  0 | task 288 |       total time =  124830.20 ms /  1889 tokens
7.03.663.682 I slot print_timing: id  0 | task 288 |    graphs reused =       2107
7.03.830.453 I slot      release: id  0 | task 288 | stop processing: n_tokens = 77778, truncated = 0
7.03.831.570 I srv  update_slots: all slots are idle
7.10.150.598 I srv  server_strea: conv_id= (empty=1)
7.10.314.236 I srv    operator(): chat format: peg-native
7.10.322.786 I slot get_availabl: id  0 | task -1 |  - checking sim = 0.973 (77778/79902) > 0.100
7.10.322.792 I slot get_availabl: id  0 | task -1 | selected slot by LCP similarity, f_sim_best = 0.973 (> 0.100 thold), f_keep = 1.000
7.10.326.691 I cmn  common_reaso: activated, budget=32000 tokens
7.10.326.697 W common_sampler_init: backend sampling is not compatible with grammar, disabling
7.10.326.979 I slot launch_slot_: id  0 | task -1 | sampler chain: logits -> top-k -> top-p -> temp-ext -> dist
7.10.326.993 I slot launch_slot_: id  0 | task -1 | sampler params:
        repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
        dry_multiplier = 0.000, dry_base = 1.750, dry_allowed_length = 2, dry_penalty_last_n = 120064
        top_k = 20, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -1.000, temp = 0.600
        mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000, adaptive_target = -1.000, adaptive_decay = 0.900
7.10.326.995 I slot launch_slot_: id  0 | task 2152 | processing task, is_child = 0
7.10.327.003 I slot   operator(): id  0 | task 2152 | new prompt, n_ctx_slot = 120064, n_keep = 0, task.n_tokens = 79902
7.10.327.046 I slot   operator(): id  0 | task 2152 | cached n_tokens = 77778, memory_seq_rm [77778, end)
7.10.327.273 I slot create_check: id  0 | task 2152 | erasing context checkpoint too close to an earlier one (pos_min = 75889, pos_max = 75889, n_tokens = 75890, size = 74.923 MiB)
7.10.329.024 I slot create_check: id  0 | task 2152 | erasing context checkpoint too close to an earlier one (pos_min = 75895, pos_max = 75895, n_tokens = 75896, size = 74.923 MiB)
7.10.330.659 I slot create_check: id  0 | task 2152 | erasing context checkpoint too close to an earlier one (pos_min = 75913, pos_max = 75913, n_tokens = 75914, size = 74.923 MiB)
7.10.350.861 I slot create_check: id  0 | task 2152 | created context checkpoint 3 of 32 (pos_min = 77777, pos_max = 77777, n_tokens = 77778, size = 74.923 MiB)
7.15.798.933 I slot print_timing: id  0 | task 2152 | prompt processing, n_tokens =   1939, progress = 1.00, t =   5.47 s / 354.35 tokens per second
7.15.798.939 I slot   operator(): id  0 | task 2152 | cached n_tokens = 79717, memory_seq_rm [79717, end)
7.16.048.417 I slot create_check: id  0 | task 2152 | created context checkpoint 4 of 32 (pos_min = 79716, pos_max = 79716, n_tokens = 79717, size = 74.923 MiB)
7.18.723.977 I slot print_timing: id  0 | task 2152 | prompt processing, n_tokens =   2120, progress = 1.00, t =   8.40 s / 252.47 tokens per second
7.18.723.981 I slot   operator(): id  0 | task 2152 | cached n_tokens = 79898, memory_seq_rm [79898, end)
7.18.732.882 I slot init_sampler: id  0 | task 2152 | init sampler, took 8.74 ms, tokens: text = 79902, total = 79902
7.18.761.969 I slot create_check: id  0 | task 2152 | created context checkpoint 5 of 32 (pos_min = 79897, pos_max = 79897, n_tokens = 79898, size = 74.923 MiB)
7.22.703.046 I cmn  common_reaso: deactivated (natural end)
7.25.752.371 I slot print_timing: id  0 | task 2152 | n_decoded =    100, tg =  14.73 t/s, tg_3s =  14.73 t/s
7.26.160.562 I slot print_timing: id  0 | task 2152 | prompt eval time =    8637.02 ms /  2124 tokens (    4.07 ms per token,   245.92 tokens per second)
7.26.160.566 I slot print_timing: id  0 | task 2152 |        eval time =    7196.53 ms /   106 tokens (   67.89 ms per token,    14.73 tokens per second)
7.26.160.567 I slot print_timing: id  0 | task 2152 |       total time =   15833.55 ms /  2230 tokens
7.26.160.568 I slot print_timing: id  0 | task 2152 |    graphs reused =       2211
7.26.163.252 I slot      release: id  0 | task 2152 | stop processing: n_tokens = 80007, truncated = 0
7.26.163.614 I srv  update_slots: all slots are idle
7.26.570.292 I srv  server_strea: conv_id= (empty=1)
7.26.782.355 I srv    operator(): chat format: peg-native
7.26.793.796 I slot get_availabl: id  0 | task -1 |  - checking sim = 0.996 (79718/80077) > 0.100
7.26.793.801 I slot get_availabl: id  0 | task -1 | selected slot by LCP similarity, f_sim_best = 0.996 (> 0.100 thold), f_keep = 0.996
7.26.797.518 I cmn  common_reaso: activated, budget=32000 tokens
7.26.797.526 W common_sampler_init: backend sampling is not compatible with grammar, disabling
7.26.797.757 I slot launch_slot_: id  0 | task -1 | sampler chain: logits -> top-k -> top-p -> temp-ext -> dist
7.26.797.773 I slot launch_slot_: id  0 | task -1 | sampler params:
        repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
        dry_multiplier = 0.000, dry_base = 1.750, dry_allowed_length = 2, dry_penalty_last_n = 120064
        top_k = 20, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -1.000, temp = 0.600
        mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000, adaptive_target = -1.000, adaptive_decay = 0.900
7.26.797.776 I slot launch_slot_: id  0 | task 2261 | processing task, is_child = 0
7.26.797.786 I slot   operator(): id  0 | task 2261 | new prompt, n_ctx_slot = 120064, n_keep = 0, task.n_tokens = 80077
7.26.797.827 I slot   operator(): id  0 | task 2261 | checking checkpoint with [79897, 79897] against 79718...
7.26.797.829 I slot   operator(): id  0 | task 2261 | checking checkpoint with [79716, 79716] against 79718...
7.26.809.492 I slot   operator(): id  0 | task 2261 | restored context checkpoint (pos_min = 79716, pos_max = 79716, n_tokens = 79717, n_past = 79717, size = 74.923 MiB)
7.26.809.497 I slot   operator(): id  0 | task 2261 | erased invalidated context checkpoint (pos_min = 79897, pos_max = 79897, n_tokens = 79898, n_swa = 0, pos_next = 79717, size = 74.923 MiB)
7.26.811.350 I slot   operator(): id  0 | task 2261 | cached n_tokens = 79717, memory_seq_rm [79717, end)
7.26.811.617 I slot create_check: id  0 | task 2261 | erasing context checkpoint too close to an earlier one (pos_min = 79716, pos_max = 79716, n_tokens = 79717, size = 74.923 MiB)
7.26.830.780 I slot create_check: id  0 | task 2261 | created context checkpoint 4 of 32 (pos_min = 79716, pos_max = 79716, n_tokens = 79717, size = 74.923 MiB)
7.30.056.837 I slot print_timing: id  0 | task 2261 | prompt processing, n_tokens =    258, progress = 1.00, t =   3.26 s / 79.16 tokens per second
7.30.056.842 I slot   operator(): id  0 | task 2261 | cached n_tokens = 79975, memory_seq_rm [79975, end)
7.30.105.160 I slot create_check: id  0 | task 2261 | created context checkpoint 5 of 32 (pos_min = 79974, pos_max = 79974, n_tokens = 79975, size = 74.923 MiB)
7.32.323.165 I slot print_timing: id  0 | task 2261 | prompt processing, n_tokens =    356, progress = 1.00, t =   5.53 s / 64.43 tokens per second
7.32.323.169 I slot   operator(): id  0 | task 2261 | cached n_tokens = 80073, memory_seq_rm [80073, end)
7.32.334.016 I slot init_sampler: id  0 | task 2261 | init sampler, took 10.75 ms, tokens: text = 80077, total = 80077
7.32.353.908 I slot create_check: id  0 | task 2261 | created context checkpoint 6 of 32 (pos_min = 80072, pos_max = 80072, n_tokens = 80073, size = 74.923 MiB)
7.36.423.851 I cmn  common_reaso: deactivated (natural end)
7.39.084.058 I slot print_timing: id  0 | task 2261 | n_decoded =    100, tg =  15.30 t/s, tg_3s =  15.30 t/s
7.42.106.177 I slot print_timing: id  0 | task 2261 | n_decoded =    146, tg =  15.27 t/s, tg_3s =  15.22 t/s
7.45.121.760 I slot print_timing: id  0 | task 2261 | n_decoded =    190, tg =  15.11 t/s, tg_3s =  14.59 t/s
7.48.156.417 I slot print_timing: id  0 | task 2261 | n_decoded =    237, tg =  15.18 t/s, tg_3s =  15.49 t/s
7.51.187.141 I slot print_timing: id  0 | task 2261 | n_decoded =    284, tg =  15.24 t/s, tg_3s =  15.51 t/s
7.54.217.409 I slot print_timing: id  0 | task 2261 | n_decoded =    328, tg =  15.14 t/s, tg_3s =  14.52 t/s
7.57.241.288 I slot print_timing: id  0 | task 2261 | n_decoded =    373, tg =  15.10 t/s, tg_3s =  14.88 t/s
8.00.299.139 I slot print_timing: id  0 | task 2261 | n_decoded =    419, tg =  15.10 t/s, tg_3s =  15.04 t/s
8.03.334.834 I slot print_timing: id  0 | task 2261 | n_decoded =    465, tg =  15.10 t/s, tg_3s =  15.15 t/s
8.06.343.865 I slot print_timing: id  0 | task 2261 | n_decoded =    509, tg =  15.06 t/s, tg_3s =  14.62 t/s
8.09.355.377 I slot print_timing: id  0 | task 2261 | prompt eval time =    5749.32 ms /   360 tokens (   15.97 ms per token,    62.62 tokens per second)
8.09.355.382 I slot print_timing: id  0 | task 2261 |        eval time =   36808.26 ms /   553 tokens (   66.56 ms per token,    15.02 tokens per second)
8.09.355.384 I slot print_timing: id  0 | task 2261 |       total time =   42557.58 ms /   913 tokens
8.09.355.385 I slot print_timing: id  0 | task 2261 |    graphs reused =       2760
8.09.359.280 I slot      release: id  0 | task 2261 | stop processing: n_tokens = 80629, truncated = 0
8.09.359.724 I srv  update_slots: all slots are idle
8.12.918.586 I srv  server_strea: conv_id= (empty=1)
8.13.079.883 I srv    operator(): chat format: peg-native
8.13.088.640 I slot get_availabl: id  0 | task -1 |  - checking sim = 0.991 (79976/80687) > 0.100
8.13.088.645 I slot get_availabl: id  0 | task -1 | selected slot by LCP similarity, f_sim_best = 0.991 (> 0.100 thold), f_keep = 0.992
8.13.092.811 I cmn  common_reaso: activated, budget=32000 tokens
8.13.092.818 W common_sampler_init: backend sampling is not compatible with grammar, disabling
8.13.093.118 I slot launch_slot_: id  0 | task -1 | sampler chain: logits -> top-k -> top-p -> temp-ext -> dist
8.13.093.130 I slot launch_slot_: id  0 | task -1 | sampler params:
        repeat_last_n = 64, repeat_penalty = 1.000, frequency_penalty = 0.000, presence_penalty = 0.000
        dry_multiplier = 0.000, dry_base = 1.750, dry_allowed_length = 2, dry_penalty_last_n = 120064
        top_k = 20, top_p = 0.950, min_p = 0.050, xtc_probability = 0.000, xtc_threshold = 0.100, typical_p = 1.000, top_n_sigma = -1.000, temp = 0.600
        mirostat = 0, mirostat_lr = 0.100, mirostat_ent = 5.000, adaptive_target = -1.000, adaptive_decay = 0.900
8.13.093.158 I slot launch_slot_: id  0 | task 2817 | processing task, is_child = 0
8.13.093.167 I slot   operator(): id  0 | task 2817 | new prompt, n_ctx_slot = 120064, n_keep = 0, task.n_tokens = 80687
8.13.093.199 I slot   operator(): id  0 | task 2817 | reusing chunk with size 558, shifting KV cache [80071, 80629) -> [79976, 80534)
8.13.093.915 I slot   operator(): id  0 | task 2817 | cached n_tokens = 80534, memory_seq_rm [80534, end)
8.13.094.006 I slot create_check: id  0 | task 2817 | erasing context checkpoint too close to an earlier one (pos_min = 79716, pos_max = 79716, n_tokens = 79717, size = 74.923 MiB)
8.13.096.051 I slot create_check: id  0 | task 2817 | erasing context checkpoint too close to an earlier one (pos_min = 79974, pos_max = 79974, n_tokens = 79975, size = 74.923 MiB)
8.13.098.036 I slot create_check: id  0 | task 2817 | erasing context checkpoint too close to an earlier one (pos_min = 80072, pos_max = 80072, n_tokens = 80073, size = 74.923 MiB)
8.13.124.148 I slot create_check: id  0 | task 2817 | created context checkpoint 4 of 32 (pos_min = 80533, pos_max = 80533, n_tokens = 80534, size = 74.923 MiB)
cmd_child_to_router:error:G:\xlam2\atomic-llama-cpp-turboquant\ggml\src\ggml-cuda\rope.cu:531: GGML_ASSERT(src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16) failed

Error running server: Command '['.\\llama-server.exe', '-m', 'D:\\Ling-3.0-flash-AD-IQ3_S-00001-of-00002.gguf', '--reasoning-preserve', '--reasoning-budget', '32000', '--reasoning-budget-message', "I've thought enough. Answering now with what I have.", '--fit', 'on', '--n-cpu-moe', '40', '-bs', '-lm', 'mlock', '-c', '120000', '--cache-type-k', 'bf16', '--cache-type-v', 'bf16', '-np', '1', '-fa', 'on', '-t', '8', '-tb', '8', '-b', '3008', '-ub', '3008', '--jinja', '-kvu', '--temp', '0.6', '--top-p', '0.95', '--top-k', '20', '--samplers', 'top_k;top_p;temperature', '--alias', 'laguna-s-2.1', '--cache-reuse', '256', '--cache-ram', '1024', '--host', '127.0.0.1', '--port', '8080', '--verbosity', '4', '--chat-template-file', 'G:\\xlam3\\chat_template_ling.jinja']' returned non-zero exit status 3221226505.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions