text stringlengths 0 613 |
|---|
[2026-06-06 03:44:55] |
=== GGUF Model Evaluation === |
[2026-06-06 03:44:55] Config : /kaggle/working/MoQ/config/quantization_pipeline.yaml.example |
[2026-06-06 03:44:55] Timestamp : 20260606_034454 |
[2026-06-06 03:44:55] Eval dir : /tmp/llama_eval/evaluation |
[2026-06-06 03:44:55] Temp dir : /kaggle/working/quants_for_model |
[2026-06-06 03:44:55] DAT dir : /tmp/llama_dat |
[2026-06-06 03:44:55] Binary dir: /workspace/llama_bins/cuda-12.8 |
[2026-06-06 03:44:55] Binary : /workspace/llama_bins/cuda-12.8/llama-perplexity |
[2026-06-06 03:44:55] Target models to evaluate: 12 |
[2026-06-06 03:44:55] ℹ️ Dataset repo_id or split is empty — skipping dataset download. |
[2026-06-06 03:44:55] ⚠️ Default test file not found: /tmp/wiki_test.raw |
[2026-06-06 03:44:55] Test files: 0 |
[2026-06-06 03:44:55] ❌ No test files available. Exiting. |
[2026-06-06 03:45:58] |
=== GGUF Model Evaluation === |
[2026-06-06 03:45:58] Config : /kaggle/working/MoQ/config/quantization_pipeline.yaml.example |
[2026-06-06 03:45:58] Timestamp : 20260606_034558 |
[2026-06-06 03:45:58] Eval dir : /tmp/llama_eval/evaluation |
[2026-06-06 03:45:58] Temp dir : /kaggle/working/quants_for_model |
[2026-06-06 03:45:58] DAT dir : /tmp/llama_dat |
[2026-06-06 03:45:58] Binary dir: /workspace/llama_bins/cuda-12.8 |
[2026-06-06 03:45:58] Binary : /workspace/llama_bins/cuda-12.8/llama-perplexity |
[2026-06-06 03:45:58] Target models to evaluate: 12 |
[2026-06-06 03:45:58] ℹ️ Dataset repo_id or split is empty — skipping dataset download. |
[2026-06-06 03:45:58] Using default test file: /tmp/wiki_test.raw |
[2026-06-06 03:45:58] Test files: 1 |
[2026-06-06 03:45:58] |
📥 Downloading base model: w-ahmad/experimental / fp16/LFM2.5-8B-A1B-F16.gguf |
[2026-06-06 03:46:53] ✅ Base model downloaded: /kaggle/working/quants_for_model/fp16/LFM2.5-8B-A1B-F16.gguf |
[2026-06-06 03:46:53] |
============================================================ |
[2026-06-06 03:46:53] PHASE 1: Generating base logits for all files |
[2026-06-06 03:46:53] ============================================================ |
[2026-06-06 03:46:53] [1/1] Base model on /tmp/wiki_test.raw → /tmp/llama_dat/base_logits_0.dat |
[2026-06-06 03:46:53] Running: /workspace/llama_bins/cuda-12.8/llama-perplexity -m /kaggle/working/quants_for_model/fp16/LFM2.5-8B-A1B-F16.gguf -f /tmp/wiki_test.raw -c 512 -b 512 -np 1 --kl-divergence-base /tmp/llama_dat/base_logits_0.dat --seed 1337 -ngl 99 --chunks 50 |
[2026-06-06 03:46:54] 0.00.325.209 I common_init_result: fitting params to device memory ... |
[2026-06-06 03:46:54] 0.00.325.237 I common_init_result: (for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on) |
[2026-06-06 03:47:19] 0.24.446.307 W llama_context: n_ctx_seq (512) < n_ctx_train (128000) -- the full capacity of the model will not be utilized |
[2026-06-06 03:47:19] 0.24.479.876 I common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable) |
[2026-06-06 03:48:07] 1.12.434.418 I |
[2026-06-06 03:48:07] 1.12.434.621 I system_info: n_threads = 2 (n_threads_batch = 2) / 4 | CUDA : ARCHS = 750,800,860,890,900,1000,1200 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 | |
[2026-06-06 03:48:07] 1.12.434.776 I perplexity: saving all logits to /tmp/llama_dat/base_logits_0.dat |
[2026-06-06 03:48:07] 1.12.434.792 I perplexity: tokenizing the input .. |
[2026-06-06 03:48:07] 1.12.729.314 I perplexity: tokenization took 294.511 ms |
[2026-06-06 03:48:07] 1.12.729.635 I perplexity: calculating perplexity over 50 chunks, n_ctx=512, batch_size=512, n_seq=1 |
[2026-06-06 03:48:22] 1.27.435.930 I perplexity: 14.61 seconds per pass - ETA 12.17 minutes |
[2026-06-06 03:48:40] [1]27.5563,[2]40.7209,[3]36.4856,[4]36.9008,[5]35.2910,[6]34.0793,[7]34.7933,[8]36.2789,[9]38.5441,[10]39.6389,[11]38.7928,[12]38.7725,[13]40.9197,[14]39.1060,[15]37.5449,[16]38.4677,[17]36.7323,[18]38.1161,[19]38.3669,[20]37.4322,[21]36.0011,[22]35.6876,[23]34.7522,[24]33.5442,[25]32.5985,[26]31.... |
[2026-06-06 03:48:41] 1.45.990.861 I Final estimate: PPL = 32.5951 +/- 1.05967 |
[2026-06-06 03:48:41] |
[2026-06-06 03:48:42] |
Phase 1 complete. Valid files: 1 / 1 |
[2026-06-06 03:48:44] 🗑️ Deleted /kaggle/working/quants_for_model/fp16/LFM2.5-8B-A1B-F16.gguf to free disk space |
[2026-06-06 03:48:44] ============================================================ |
[2026-06-06 03:48:44] PHASE 2: Running KL divergence for each target model |
[2026-06-06 03:48:44] ============================================================ |
[2026-06-06 03:48:44] |
============================================================ |
[2026-06-06 03:48:44] TARGET MODEL 1/12: w-ahmad/salt / MoQ-Quants/MoQ-2.9.gguf |
[2026-06-06 03:48:44] ============================================================ |
[2026-06-06 03:48:59] |
[1/1] MoQ-Quants/MoQ-2.9.gguf vs /tmp/wiki_test.raw (base: /tmp/llama_dat/base_logits_0.dat) |
[2026-06-06 03:48:59] Running: /workspace/llama_bins/cuda-12.8/llama-perplexity -m /kaggle/working/quants_for_model/MoQ-Quants/MoQ-2.9.gguf -f /tmp/wiki_test.raw -c 512 -b 512 -np 1 --kl-divergence-base /tmp/llama_dat/base_logits_0.dat --kl-divergence --seed 1337 -ngl 99 --chunks 50 |
[2026-06-06 03:48:59] 0.00.269.249 I common_init_result: fitting params to device memory ... |
[2026-06-06 03:48:59] 0.00.269.258 I common_init_result: (for bugs during this step try to reproduce them with -fit off, or provide --verbose logs if the bug only occurs with -fit on) |
[2026-06-06 03:49:02] 0.02.968.775 W llama_context: n_ctx_seq (512) < n_ctx_train (128000) -- the full capacity of the model will not be utilized |
[2026-06-06 03:49:02] 0.03.002.469 I common_init_from_params: warming up the model with an empty run - please wait ... (--no-warmup to disable) |
[2026-06-06 03:49:16] 0.17.823.407 I |
[2026-06-06 03:49:17] 0.17.823.573 I system_info: n_threads = 2 (n_threads_batch = 2) / 4 | CUDA : ARCHS = 750,800,860,890,900,1000,1200 | USE_GRAPHS = 1 | PEER_MAX_BATCH_SIZE = 128 | CPU : SSE3 = 1 | SSSE3 = 1 | AVX = 1 | AVX2 = 1 | F16C = 1 | FMA = 1 | BMI2 = 1 | LLAMAFILE = 1 | OPENMP = 1 | REPACK = 1 | |
[2026-06-06 03:49:17] 0.17.870.363 I kl_divergence: computing over 50 chunks, n_ctx=512, batch_size=512, n_seq=1 |
[2026-06-06 03:50:50] 1.51.552.103 I kl_divergence: 93.68 seconds per pass - ETA 1 hours 18.07 minutes |
[2026-06-06 03:50:51] |
[2026-06-06 03:50:51] chunk PPL ln(PPL(Q)/PPL(base)) KL Divergence Δp RMS Same top p |
[2026-06-06 03:50:51] 1 22.6911 ± 5.0670 -0.18615 ± 0.08778 0.65099 ± 0.04867 18.398 ± 1.582 % 62.745 ± 3.034 % |
[2026-06-06 03:50:51] 2 36.2592 ± 5.9332 -0.11067 ± 0.06403 0.70396 ± 0.03747 16.687 ± 1.128 % 62.549 ± 2.145 % |
[2026-06-06 03:50:51] 3 37.8546 ± 5.0140 0.04604 ± 0.05941 0.79061 ± 0.03484 18.506 ± 0.883 % 60.523 ± 1.768 % |
[2026-06-06 03:50:51] 4 38.3030 ± 4.3404 0.04419 ± 0.05231 0.80473 ± 0.03276 18.192 ± 0.765 % 61.765 ± 1.522 % |
[2026-06-06 03:50:52] 5 37.9258 ± 3.7417 0.07752 ± 0.04527 0.77541 ± 0.02731 18.104 ± 0.699 % 60.471 ± 1.370 % |
[2026-06-06 03:50:52] 6 39.4557 ± 3.5563 0.15108 ± 0.04122 0.76915 ± 0.02403 18.238 ± 0.649 % 60.392 ± 1.251 % |
[2026-06-06 03:50:52] 7 39.6144 ± 3.2746 0.13370 ± 0.03820 0.76821 ± 0.02245 18.245 ± 0.593 % 60.224 ± 1.159 % |
[2026-06-06 03:50:53] 8 40.9910 ± 3.1885 0.12556 ± 0.03571 0.76307 ± 0.02035 18.153 ± 0.553 % 59.118 ± 1.089 % |
[2026-06-06 03:50:53] 9 43.8834 ± 3.2524 0.13279 ± 0.03309 0.74757 ± 0.01849 17.836 ± 0.517 % 58.998 ± 1.027 % |
[2026-06-06 03:50:53] 10 43.8738 ± 3.0520 0.10425 ± 0.03118 0.74091 ± 0.01713 17.444 ± 0.489 % 58.431 ± 0.976 % |
[2026-06-06 03:50:54] 11 43.9305 ± 2.9009 0.12687 ± 0.02958 0.74212 ± 0.01637 17.630 ± 0.478 % 58.360 ± 0.931 % |
[2026-06-06 03:50:54] 12 44.5681 ± 2.8333 0.14159 ± 0.02830 0.74113 ± 0.01545 17.621 ± 0.457 % 58.203 ± 0.892 % |
[2026-06-06 03:50:54] 13 47.1566 ± 2.9113 0.14422 ± 0.02701 0.73586 ± 0.01465 17.524 ± 0.445 % 58.130 ± 0.857 % |
[2026-06-06 03:50:55] 14 45.1908 ± 2.6763 0.14680 ± 0.02573 0.72153 ± 0.01384 17.487 ± 0.424 % 58.796 ± 0.824 % |
[2026-06-06 03:50:55] 15 43.9570 ± 2.5004 0.15971 ± 0.02475 0.71639 ± 0.01327 17.398 ± 0.410 % 59.242 ± 0.795 % |
[2026-06-06 03:50:55] 16 44.7163 ± 2.4723 0.15243 ± 0.02399 0.71561 ± 0.01292 17.362 ± 0.396 % 59.289 ± 0.769 % |
[2026-06-06 03:50:56] 17 42.0254 ± 2.2450 0.13718 ± 0.02316 0.70454 ± 0.01242 17.309 ± 0.383 % 59.839 ± 0.745 % |
[2026-06-06 03:50:56] 18 43.3685 ± 2.2502 0.13151 ± 0.02236 0.70953 ± 0.01200 17.391 ± 0.374 % 59.477 ± 0.725 % |
[2026-06-06 03:50:56] 19 44.2930 ± 2.2532 0.14689 ± 0.02195 0.72184 ± 0.01177 17.602 ± 0.366 % 59.340 ± 0.706 % |
[2026-06-06 03:50:57] 20 43.8727 ± 2.1726 0.16186 ± 0.02132 0.71530 ± 0.01138 17.585 ± 0.353 % 59.667 ± 0.687 % |
[2026-06-06 03:50:57] 21 42.5424 ± 2.0521 0.16990 ± 0.02097 0.71728 ± 0.01111 17.713 ± 0.343 % 59.645 ± 0.670 % |
[2026-06-06 03:50:58] 22 42.4394 ± 1.9981 0.17609 ± 0.02050 0.71784 ± 0.01086 17.744 ± 0.333 % 59.608 ± 0.655 % |
[2026-06-06 03:50:58] 23 41.1753 ± 1.8937 0.17313 ± 0.02008 0.72036 ± 0.01064 17.790 ± 0.326 % 59.761 ± 0.640 % |
[2026-06-06 03:50:58] 24 40.1667 ± 1.8030 0.18356 ± 0.01953 0.71822 ± 0.01035 17.908 ± 0.319 % 59.967 ± 0.626 % |
[2026-06-06 03:50:59] 25 38.9064 ± 1.7059 0.18015 ± 0.01895 0.71155 ± 0.01002 17.810 ± 0.311 % 59.937 ± 0.614 % |
[2026-06-06 03:50:59] 26 37.3231 ± 1.5983 0.17846 ± 0.01854 0.70554 ± 0.00978 17.720 ± 0.302 % 60.407 ± 0.601 % |
End of preview. Expand in Data Studio
No dataset card yet
- Downloads last month
- 9