Upload 32 files

8b_only_dpo/solo_dpo_model_base.log

@@ -0,0 +1,25 @@
+2025-02-06:10:04:25,579 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:10:04:37,567 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:10:04:37,568 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:10:04:37,568 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:10:04:37,569 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct'}
+2025-02-06:10:04:37,615 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:10:04:38,416 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:10:04:49,323 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:10:04:49,324 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:10:04:49,324 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 11/1319 [00:00<00:12, 101.56it/s]
  2%|▏         | 22/1319 [00:00<00:12, 105.13it/s]
  3%|▎         | 34/1319 [00:00<00:11, 107.89it/s]
  3%|▎         | 45/1319 [00:00<00:11, 107.91it/s]
  4%|▍         | 56/1319 [00:00<00:11, 105.86it/s]
  5%|▌         | 67/1319 [00:00<00:11, 106.36it/s]
  6%|▌         | 78/1319 [00:00<00:11, 107.23it/s]
  7%|▋         | 90/1319 [00:00<00:11, 108.77it/s]
  8%|▊         | 101/1319 [00:00<00:11, 109.12it/s]
  8%|▊         | 112/1319 [00:01<00:11, 109.35it/s]
  9%|▉         | 123/1319 [00:01<00:11, 108.67it/s]
 10%|█         | 134/1319 [00:01<00:10, 108.03it/s]
 11%|█         | 146/1319 [00:01<00:10, 109.08it/s]
 12%|█▏        | 157/1319 [00:01<00:10, 108.59it/s]
 13%|█▎        | 169/1319 [00:01<00:10, 109.49it/s]
 14%|█▎        | 181/1319 [00:01<00:10, 109.65it/s]
 15%|█▍        | 192/1319 [00:01<00:10, 108.81it/s]
 15%|█▌        | 204/1319 [00:01<00:10, 109.38it/s]
 16%|█▋        | 215/1319 [00:01<00:10, 109.26it/s]
 17%|█▋        | 226/1319 [00:02<00:10, 107.95it/s]
 18%|█▊        | 238/1319 [00:02<00:09, 109.18it/s]
 19%|█▉        | 249/1319 [00:02<00:09, 108.01it/s]
 20%|█▉        | 261/1319 [00:02<00:09, 109.10it/s]
 21%|██        | 273/1319 [00:02<00:09, 110.70it/s]
 22%|██▏       | 285/1319 [00:02<00:09, 109.55it/s]
 23%|██▎       | 297/1319 [00:02<00:09, 110.61it/s]
 23%|██▎       | 309/1319 [00:02<00:09, 111.58it/s]
 24%|██▍       | 321/1319 [00:02<00:08, 112.22it/s]
 25%|██▌       | 333/1319 [00:03<00:08, 112.78it/s]
 26%|██▌       | 345/1319 [00:03<00:08, 112.25it/s]
 27%|██▋       | 357/1319 [00:03<00:08, 111.84it/s]
 28%|██▊       | 369/1319 [00:03<00:08, 112.60it/s]
 29%|██▉       | 381/1319 [00:03<00:08, 111.71it/s]
 30%|██▉       | 393/1319 [00:03<00:08, 111.34it/s]
 31%|███       | 405/1319 [00:03<00:08, 111.16it/s]
 32%|███▏      | 417/1319 [00:03<00:08, 110.91it/s]
 33%|███▎      | 429/1319 [00:03<00:08, 110.58it/s]
 33%|███▎      | 441/1319 [00:04<00:08, 108.53it/s]
 34%|███▍      | 452/1319 [00:04<00:08, 107.09it/s]
 35%|███▌      | 464/1319 [00:04<00:07, 108.43it/s]
 36%|███▌      | 475/1319 [00:04<00:07, 106.92it/s]
 37%|███▋      | 486/1319 [00:04<00:07, 107.17it/s]
 38%|███▊      | 498/1319 [00:04<00:07, 108.15it/s]
 39%|███▊      | 509/1319 [00:04<00:07, 106.85it/s]
 39%|███▉      | 520/1319 [00:04<00:07, 104.99it/s]
 40%|████      | 531/1319 [00:04<00:07, 104.22it/s]
 41%|████      | 542/1319 [00:04<00:07, 102.34it/s]
 42%|████▏     | 553/1319 [00:05<00:07, 103.54it/s]
 43%|████▎     | 565/1319 [00:05<00:07, 106.11it/s]
 44%|████▎     | 577/1319 [00:05<00:06, 107.83it/s]
 45%|████▍     | 588/1319 [00:05<00:06, 108.35it/s]
 45%|████▌     | 599/1319 [00:05<00:06, 107.84it/s]
 46%|████▌     | 610/1319 [00:05<00:06, 108.15it/s]
 47%|████▋     | 621/1319 [00:05<00:06, 107.97it/s]
 48%|████▊     | 633/1319 [00:05<00:06, 109.01it/s]
 49%|████▉     | 644/1319 [00:05<00:06, 107.80it/s]
 50%|████���     | 655/1319 [00:06<00:06, 105.41it/s]
 50%|█████     | 666/1319 [00:06<00:06, 104.82it/s]
 51%|█████▏    | 677/1319 [00:06<00:06, 105.97it/s]
 52%|█████▏    | 689/1319 [00:06<00:05, 107.85it/s]
 53%|█████▎    | 701/1319 [00:06<00:05, 108.93it/s]
 54%|█████▍    | 713/1319 [00:06<00:05, 110.29it/s]
 55%|█████▍    | 725/1319 [00:06<00:05, 110.17it/s]
 56%|█████▌    | 737/1319 [00:06<00:05, 109.73it/s]
 57%|█████▋    | 748/1319 [00:06<00:05, 107.37it/s]
 58%|█████▊    | 760/1319 [00:07<00:05, 108.99it/s]
 59%|█████▊    | 772/1319 [00:07<00:04, 109.87it/s]
 59%|█████▉    | 784/1319 [00:07<00:04, 110.89it/s]
 60%|██████    | 796/1319 [00:07<00:04, 111.05it/s]
 61%|██████▏   | 808/1319 [00:07<00:04, 110.49it/s]
 62%|██████▏   | 820/1319 [00:07<00:04, 110.45it/s]
 63%|██████▎   | 832/1319 [00:07<00:04, 111.13it/s]
 64%|██████▍   | 844/1319 [00:07<00:04, 110.81it/s]
 65%|██████▍   | 856/1319 [00:07<00:04, 110.86it/s]
 66%|██████▌   | 868/1319 [00:07<00:04, 111.29it/s]
 67%|██████▋   | 880/1319 [00:08<00:03, 110.84it/s]
 68%|██████▊   | 892/1319 [00:08<00:03, 110.99it/s]
 69%|██████▊   | 904/1319 [00:08<00:03, 109.76it/s]
 69%|██████▉   | 916/1319 [00:08<00:03, 109.64it/s]
 70%|███████   | 928/1319 [00:08<00:03, 110.55it/s]
 71%|███████▏  | 940/1319 [00:08<00:03, 109.29it/s]
 72%|███████▏  | 951/1319 [00:08<00:03, 99.06it/s] 
 73%|███████▎  | 962/1319 [00:08<00:03, 94.72it/s]
 74%|███████▎  | 972/1319 [00:08<00:03, 95.53it/s]
 74%|███████▍  | 982/1319 [00:09<00:03, 96.26it/s]
 75%|███████▌  | 993/1319 [00:09<00:03, 99.10it/s]
 76%|███████▌  | 1004/1319 [00:09<00:03, 100.76it/s]
 77%|███████▋  | 1015/1319 [00:09<00:02, 102.22it/s]
 78%|███████▊  | 1026/1319 [00:09<00:02, 102.31it/s]
 79%|███████▊  | 1037/1319 [00:09<00:02, 104.03it/s]
 79%|███████▉  | 1048/1319 [00:09<00:02, 105.32it/s]
 80%|████████  | 1059/1319 [00:09<00:02, 105.17it/s]
 81%|████████  | 1070/1319 [00:09<00:02, 106.03it/s]
 82%|████████▏ | 1081/1319 [00:10<00:02, 106.55it/s]
 83%|████████▎ | 1092/1319 [00:10<00:02, 106.38it/s]
 84%|████████▎ | 1103/1319 [00:10<00:02, 106.93it/s]
 84%|████████▍ | 1114/1319 [00:10<00:01, 106.41it/s]
 85%|████████▌ | 1125/1319 [00:10<00:01, 105.43it/s]
 86%|████████▌ | 1137/1319 [00:10<00:01, 108.66it/s]
 87%|████████▋ | 1149/1319 [00:10<00:01, 111.16it/s]
 88%|████████▊ | 1161/1319 [00:10<00:01, 112.41it/s]
 89%|████████▉ | 1173/1319 [00:10<00:01, 110.76it/s]
 90%|████████▉ | 1185/1319 [00:10<00:01, 110.30it/s]
 91%|█████████ | 1197/1319 [00:11<00:01, 112.25it/s]
 92%|█████████▏| 1209/1319 [00:11<00:00, 113.86it/s]
 93%|█████████▎| 1221/1319 [00:11<00:00, 115.45it/s]
 94%|█████████▎| 1234/1319 [00:11<00:00, 116.88it/s]
 95%|█████████▍| 1247/1319 [00:11<00:00, 117.97it/s]
 96%|█████████▌| 1260/1319 [00:11<00:00, 119.12it/s]
 97%|█████████▋| 1273/1319 [00:11<00:00, 120.01it/s]
 97%|█████████▋| 1286/1319 [00:11<00:00, 120.07it/s]
 98%|█████████▊| 1299/1319 [00:11<00:00, 120.35it/s]
 99%|█████████▉| 1312/1319 [00:12<00:00, 120.74it/s]
+2025-02-06:10:05:01,456 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:10:46:25,129 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6793|±  |0.0129|
+|               |       |strict-match    |     8|exact_match|↑  |0.6717|±  |0.0129|
+

8b_only_dpo/solo_dpo_model_ckpt_100.log

@@ -0,0 +1,25 @@
+2025-02-06:11:19:17,726 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:11:19:29,750 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:11:19:29,751 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:11:19:29,752 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:11:19:29,752 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-100'}
+2025-02-06:11:19:29,793 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:11:19:30,520 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:11:20:14,907 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:11:20:14,908 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:11:20:14,908 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 10/1319 [00:00<00:13, 96.88it/s]
  2%|▏         | 21/1319 [00:00<00:12, 99.85it/s]
  2%|▏         | 32/1319 [00:00<00:12, 102.66it/s]
  3%|▎         | 43/1319 [00:00<00:12, 105.05it/s]
  4%|▍         | 54/1319 [00:00<00:11, 106.69it/s]
  5%|▍         | 65/1319 [00:00<00:11, 107.59it/s]
  6%|▌         | 76/1319 [00:00<00:11, 106.61it/s]
  7%|▋         | 87/1319 [00:00<00:11, 103.24it/s]
  7%|▋         | 98/1319 [00:00<00:11, 104.98it/s]
  8%|▊         | 109/1319 [00:01<00:11, 105.13it/s]
  9%|▉         | 120/1319 [00:01<00:11, 105.56it/s]
 10%|▉         | 131/1319 [00:01<00:11, 106.50it/s]
 11%|█         | 142/1319 [00:01<00:10, 107.43it/s]
 12%|█▏        | 154/1319 [00:01<00:10, 108.44it/s]
 13%|█▎        | 166/1319 [00:01<00:10, 109.42it/s]
 13%|█▎        | 178/1319 [00:01<00:10, 109.85it/s]
 14%|█▍        | 189/1319 [00:01<00:10, 108.39it/s]
 15%|█▌        | 200/1319 [00:01<00:10, 108.29it/s]
 16%|█▌        | 212/1319 [00:01<00:10, 108.92it/s]
 17%|█▋        | 223/1319 [00:02<00:10, 108.71it/s]
 18%|█▊        | 234/1319 [00:02<00:10, 108.21it/s]
 19%|█▊        | 246/1319 [00:02<00:09, 108.65it/s]
 19%|█▉        | 257/1319 [00:02<00:09, 108.55it/s]
 20%|██        | 268/1319 [00:02<00:09, 107.94it/s]
 21%|██        | 279/1319 [00:02<00:09, 106.57it/s]
 22%|██▏       | 291/1319 [00:02<00:09, 107.46it/s]
 23%|██▎       | 302/1319 [00:02<00:09, 103.72it/s]
 24%|██▎       | 313/1319 [00:02<00:09, 103.88it/s]
 25%|██▍       | 325/1319 [00:03<00:09, 105.88it/s]
 25%|██▌       | 336/1319 [00:03<00:09, 105.56it/s]
 26%|██▋       | 347/1319 [00:03<00:09, 106.10it/s]
 27%|██▋       | 358/1319 [00:03<00:09, 106.68it/s]
 28%|██▊       | 369/1319 [00:03<00:08, 106.69it/s]
 29%|██▉       | 380/1319 [00:03<00:08, 106.92it/s]
 30%|██▉       | 392/1319 [00:03<00:08, 107.79it/s]
 31%|███       | 403/1319 [00:03<00:08, 106.96it/s]
 31%|███▏      | 414/1319 [00:03<00:08, 106.44it/s]
 32%|███▏      | 426/1319 [00:03<00:08, 108.06it/s]
 33%|███▎      | 437/1319 [00:04<00:08, 106.93it/s]
 34%|███▍      | 448/1319 [00:04<00:08, 105.96it/s]
 35%|███▍      | 459/1319 [00:04<00:08, 107.07it/s]
 36%|███▌      | 470/1319 [00:04<00:08, 105.94it/s]
 36%|███▋      | 481/1319 [00:04<00:07, 105.26it/s]
 37%|███▋      | 493/1319 [00:04<00:07, 107.40it/s]
 38%|███▊      | 505/1319 [00:04<00:07, 108.33it/s]
 39%|███▉      | 517/1319 [00:04<00:07, 108.92it/s]
 40%|████      | 528/1319 [00:04<00:07, 108.25it/s]
 41%|████      | 539/1319 [00:05<00:07, 108.06it/s]
 42%|████▏     | 550/1319 [00:05<00:07, 107.45it/s]
 43%|████▎     | 561/1319 [00:05<00:07, 107.93it/s]
 43%|████▎     | 572/1319 [00:05<00:06, 107.56it/s]
 44%|████▍     | 583/1319 [00:05<00:06, 108.26it/s]
 45%|████▌     | 594/1319 [00:05<00:06, 106.48it/s]
 46%|████▌     | 605/1319 [00:05<00:06, 106.47it/s]
 47%|████▋     | 617/1319 [00:05<00:06, 107.60it/s]
 48%|████▊     | 628/1319 [00:05<00:06, 106.04it/s]
 48%|████▊     | 639/1319 [00:05<00:06, 106.90it/s]
 49%|████▉     | 651/1319 [00:06<00:06, 108.70it/s]
 50%|█████     | 662/1319 [00:06<00:06, 107.04it/s]
 51%|█████     | 673/1319 [00:06<00:06, 105.38it/s]
 52%|█████▏    | 684/1319 [00:06<00:05, 106.57it/s]
 53%|█████▎    | 695/1319 [00:06<00:05, 105.42it/s]
 54%|█████▎    | 706/1319 [00:06<00:05, 104.24it/s]
 54%|█████▍    | 717/1319 [00:06<00:05, 105.87it/s]
 55%|█████▌    | 728/1319 [00:06<00:05, 106.90it/s]
 56%|█████▌    | 739/1319 [00:06<00:05, 105.52it/s]
 57%|█████▋    | 750/1319 [00:07<00:05, 106.24it/s]
 58%|█████▊    | 762/1319 [00:07<00:05, 107.45it/s]
 59%|█████▊    | 773/1319 [00:07<00:05, 107.08it/s]
 59%|█████▉    | 784/1319 [00:07<00:04, 107.65it/s]
 60%|██████    | 796/1319 [00:07<00:04, 108.90it/s]
 61%|██████▏   | 808/1319 [00:07<00:04, 109.44it/s]
 62%|██████▏   | 819/1319 [00:07<00:04, 108.98it/s]
 63%|██████▎   | 831/1319 [00:07<00:04, 109.61it/s]
 64%|██████▍   | 843/1319 [00:07<00:04, 110.32it/s]
 65%|██████▍   | 855/1319 [00:07<00:04, 110.79it/s]
 66%|██████▌   | 867/1319 [00:08<00:04, 109.41it/s]
 67%|██████▋   | 879/1319 [00:08<00:03, 110.67it/s]
 68%|██████▊   | 891/1319 [00:08<00:03, 111.73it/s]
 68%|██████▊   | 903/1319 [00:08<00:03, 110.03it/s]
 69%|██████▉   | 915/1319 [00:08<00:03, 109.18it/s]
 70%|███████   | 926/1319 [00:08<00:03, 109.18it/s]
 71%|███████   | 938/1319 [00:08<00:03, 110.07it/s]
 72%|███████▏  | 950/1319 [00:08<00:03, 110.55it/s]
 73%|███████▎  | 962/1319 [00:08<00:03, 109.92it/s]
 74%|███████▍  | 973/1319 [00:09<00:03, 107.48it/s]
 75%|███████▍  | 985/1319 [00:09<00:03, 108.33it/s]
 76%|███████▌  | 997/1319 [00:09<00:02, 109.11it/s]
 76%|███████▋  | 1008/1319 [00:09<00:02, 106.92it/s]
 77%|███████▋  | 1019/1319 [00:09<00:02, 106.60it/s]
 78%|███████▊  | 1030/1319 [00:09<00:02, 107.19it/s]
 79%|███████▉  | 1042/1319 [00:09<00:02, 109.95it/s]
 80%|███████▉  | 1054/1319 [00:09<00:02, 112.82it/s]
 81%|████████  | 1066/1319 [00:09<00:02, 114.63it/s]
 82%|████████▏ | 1078/1319 [00:09<00:02, 115.40it/s]
 83%|████████▎ | 1090/1319 [00:10<00:01, 116.12it/s]
 84%|████████▎ | 1102/1319 [00:10<00:01, 116.95it/s]
 84%|████████▍ | 1114/1319 [00:10<00:01, 117.79it/s]
 85%|████████▌ | 1127/1319 [00:10<00:01, 118.56it/s]
 86%|████████▋ | 1140/1319 [00:10<00:01, 119.00it/s]
 87%|████████▋ | 1152/1319 [00:10<00:01, 117.55it/s]
 88%|████████▊ | 1164/1319 [00:10<00:01, 117.90it/s]
 89%|████████▉ | 1176/1319 [00:10<00:01, 118.03it/s]
 90%|█████████ | 1188/1319 [00:10<00:01, 118.31it/s]
 91%|█████████ | 1200/1319 [00:11<00:01, 118.49it/s]
 92%|█████████▏| 1212/1319 [00:11<00:00, 116.05it/s]
 93%|█████████▎| 1224/1319 [00:11<00:00, 113.52it/s]
 94%|█████████▎| 1236/1319 [00:11<00:00, 111.32it/s]
 95%|█████████▍| 1248/1319 [00:11<00:00, 111.40it/s]
 96%|█████████▌| 1260/1319 [00:11<00:00, 111.52it/s]
 96%|█████████▋| 1272/1319 [00:11<00:00, 111.00it/s]
 97%|█████████▋| 1284/1319 [00:11<00:00, 110.47it/s]
 98%|█████████▊| 1296/1319 [00:11<00:00, 109.69it/s]
 99%|█████████▉| 1307/1319 [00:12<00:00, 109.67it/s]
+2025-02-06:11:20:27,063 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:12:11:06,480 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-100), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6641|±  |0.0130|
+|               |       |strict-match    |     8|exact_match|↑  |0.6497|±  |0.0131|
+

8b_only_dpo/solo_dpo_model_ckpt_1000.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1100.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1200.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1300.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1400.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1500.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1600.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1700.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1800.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_1900.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_200.log

@@ -0,0 +1,25 @@
+2025-02-06:12:11:10,392 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:12:11:21,666 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:12:11:21,667 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:12:11:21,667 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:12:11:21,667 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-200'}
+2025-02-06:12:11:21,706 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:12:11:22,311 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:12:12:03,559 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:12:12:03,559 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:12:12:03,560 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 12/1319 [00:00<00:11, 113.87it/s]
  2%|▏         | 24/1319 [00:00<00:11, 117.36it/s]
  3%|▎         | 36/1319 [00:00<00:10, 118.54it/s]
  4%|▎         | 49/1319 [00:00<00:10, 119.44it/s]
  5%|▍         | 62/1319 [00:00<00:10, 120.00it/s]
  6%|▌         | 74/1319 [00:00<00:10, 119.38it/s]
  7%|▋         | 87/1319 [00:00<00:10, 119.70it/s]
  8%|▊         | 100/1319 [00:00<00:10, 120.18it/s]
  9%|▊         | 113/1319 [00:00<00:09, 120.70it/s]
 10%|▉         | 126/1319 [00:01<00:09, 121.16it/s]
 11%|█         | 139/1319 [00:01<00:09, 118.55it/s]
 11%|█▏        | 151/1319 [00:01<00:09, 118.41it/s]
 12%|█▏        | 163/1319 [00:01<00:09, 117.80it/s]
 13%|█▎        | 175/1319 [00:01<00:09, 117.98it/s]
 14%|█▍        | 187/1319 [00:01<00:09, 117.55it/s]
 15%|█▌        | 199/1319 [00:01<00:09, 116.39it/s]
 16%|█▌        | 211/1319 [00:01<00:09, 115.87it/s]
 17%|█▋        | 223/1319 [00:01<00:09, 115.14it/s]
 18%|█▊        | 235/1319 [00:01<00:09, 115.33it/s]
 19%|█▊        | 247/1319 [00:02<00:09, 115.93it/s]
 20%|█▉        | 259/1319 [00:02<00:09, 116.22it/s]
 21%|██        | 271/1319 [00:02<00:08, 116.48it/s]
 21%|██▏       | 283/1319 [00:02<00:08, 117.01it/s]
 22%|██▏       | 295/1319 [00:02<00:08, 117.14it/s]
 23%|██▎       | 307/1319 [00:02<00:08, 117.62it/s]
 24%|██▍       | 319/1319 [00:02<00:08, 117.66it/s]
 25%|██▌       | 331/1319 [00:02<00:08, 117.91it/s]
 26%|██▌       | 343/1319 [00:02<00:08, 117.65it/s]
 27%|██▋       | 355/1319 [00:03<00:08, 117.53it/s]
 28%|██▊       | 367/1319 [00:03<00:08, 117.70it/s]
 29%|██▊       | 379/1319 [00:03<00:07, 118.00it/s]
 30%|██▉       | 391/1319 [00:03<00:07, 117.69it/s]
 31%|███       | 403/1319 [00:03<00:07, 117.81it/s]
 31%|███▏      | 415/1319 [00:03<00:07, 117.66it/s]
 32%|███▏      | 427/1319 [00:03<00:07, 118.03it/s]
 33%|███▎      | 439/1319 [00:03<00:07, 117.72it/s]
 34%|███▍      | 451/1319 [00:03<00:07, 117.95it/s]
 35%|███▌      | 463/1319 [00:03<00:07, 116.98it/s]
 36%|███▌      | 475/1319 [00:04<00:07, 117.14it/s]
 37%|███▋      | 487/1319 [00:04<00:07, 117.13it/s]
 38%|███▊      | 499/1319 [00:04<00:06, 117.53it/s]
 39%|███▊      | 511/1319 [00:04<00:06, 117.57it/s]
 40%|███▉      | 523/1319 [00:04<00:06, 117.96it/s]
 41%|████      | 535/1319 [00:04<00:06, 117.44it/s]
 41%|████▏     | 547/1319 [00:04<00:06, 117.88it/s]
 42%|████▏     | 559/1319 [00:04<00:06, 117.18it/s]
 43%|████▎     | 571/1319 [00:04<00:06, 117.44it/s]
 44%|████▍     | 583/1319 [00:04<00:06, 117.44it/s]
 45%|████▌     | 595/1319 [00:05<00:06, 117.60it/s]
 46%|████▌     | 607/1319 [00:05<00:06, 117.46it/s]
 47%|████▋     | 619/1319 [00:05<00:05, 117.84it/s]
 48%|████▊     | 631/1319 [00:05<00:05, 117.47it/s]
 49%|████▊     | 643/1319 [00:05<00:05, 117.83it/s]
 50%|████▉     | 655/1319 [00:05<00:05, 115.96it/s]
 51%|█████     | 667/1319 [00:05<00:05, 115.40it/s]
 51%|███��█▏    | 679/1319 [00:05<00:05, 115.42it/s]
 52%|█████▏    | 691/1319 [00:05<00:05, 116.12it/s]
 53%|█████▎    | 703/1319 [00:05<00:05, 115.15it/s]
 54%|█████▍    | 715/1319 [00:06<00:05, 115.59it/s]
 55%|█████▌    | 727/1319 [00:06<00:05, 116.13it/s]
 56%|█████▌    | 739/1319 [00:06<00:05, 115.94it/s]
 57%|█████▋    | 751/1319 [00:06<00:04, 116.14it/s]
 58%|█████▊    | 763/1319 [00:06<00:04, 116.22it/s]
 59%|█████▉    | 775/1319 [00:06<00:04, 115.66it/s]
 60%|█████▉    | 787/1319 [00:06<00:04, 115.37it/s]
 61%|██████    | 799/1319 [00:06<00:04, 115.30it/s]
 61%|██████▏   | 811/1319 [00:06<00:04, 115.75it/s]
 62%|██████▏   | 823/1319 [00:07<00:04, 116.05it/s]
 63%|██████▎   | 835/1319 [00:07<00:04, 116.48it/s]
 64%|██████▍   | 847/1319 [00:07<00:04, 116.76it/s]
 65%|██████▌   | 859/1319 [00:07<00:03, 116.25it/s]
 66%|██████▌   | 871/1319 [00:07<00:03, 116.68it/s]
 67%|██████▋   | 883/1319 [00:07<00:03, 116.38it/s]
 68%|██████▊   | 895/1319 [00:07<00:03, 116.16it/s]
 69%|██████▉   | 907/1319 [00:07<00:03, 115.81it/s]
 70%|██████▉   | 919/1319 [00:07<00:03, 116.27it/s]
 71%|███████   | 931/1319 [00:07<00:03, 115.92it/s]
 71%|███████▏  | 943/1319 [00:08<00:03, 115.32it/s]
 72%|███████▏  | 955/1319 [00:08<00:03, 115.70it/s]
 73%|███████▎  | 967/1319 [00:08<00:03, 116.16it/s]
 74%|███████▍  | 979/1319 [00:08<00:02, 116.91it/s]
 75%|███████▌  | 991/1319 [00:08<00:02, 117.45it/s]
 76%|███████▌  | 1003/1319 [00:08<00:02, 117.81it/s]
 77%|███████▋  | 1015/1319 [00:08<00:02, 118.08it/s]
 78%|███████▊  | 1027/1319 [00:08<00:02, 118.18it/s]
 79%|███████▉  | 1039/1319 [00:08<00:02, 118.37it/s]
 80%|███████▉  | 1051/1319 [00:08<00:02, 118.48it/s]
 81%|████████  | 1063/1319 [00:09<00:02, 118.59it/s]
 82%|████████▏ | 1075/1319 [00:09<00:02, 118.64it/s]
 82%|████████▏ | 1087/1319 [00:09<00:01, 118.72it/s]
 83%|████████▎ | 1099/1319 [00:09<00:01, 118.74it/s]
 84%|████████▍ | 1111/1319 [00:09<00:01, 118.81it/s]
 85%|████████▌ | 1123/1319 [00:09<00:01, 118.77it/s]
 86%|████████▌ | 1135/1319 [00:09<00:01, 118.67it/s]
 87%|████████▋ | 1147/1319 [00:09<00:01, 118.66it/s]
 88%|████████▊ | 1159/1319 [00:09<00:01, 118.63it/s]
 89%|████████▉ | 1171/1319 [00:09<00:01, 118.64it/s]
 90%|████████▉ | 1183/1319 [00:10<00:01, 118.14it/s]
 91%|█████████ | 1195/1319 [00:10<00:01, 118.17it/s]
 92%|█████████▏| 1207/1319 [00:10<00:00, 118.24it/s]
 92%|█████████▏| 1219/1319 [00:10<00:00, 118.20it/s]
 93%|█████████▎| 1231/1319 [00:10<00:00, 118.37it/s]
 94%|█████████▍| 1243/1319 [00:10<00:00, 117.94it/s]
 95%|█████████▌| 1255/1319 [00:10<00:00, 117.63it/s]
 96%|█████████▌| 1267/1319 [00:10<00:00, 117.72it/s]
 97%|█████████▋| 1279/1319 [00:10<00:00, 117.84it/s]
 98%|█████████▊| 1291/1319 [00:10<00:00, 118.04it/s]
 99%|█████████▉| 1303/1319 [00:11<00:00, 118.14it/s]
+2025-02-06:12:12:14,829 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:13:02:55,825 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-200), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6679|±  |0.0130|
+|               |       |strict-match    |     8|exact_match|↑  |0.6550|±  |0.0131|
+

8b_only_dpo/solo_dpo_model_ckpt_2000.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2100.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2200.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2300.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2400.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2500.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2600.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2700.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2800.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_2900.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found

8b_only_dpo/solo_dpo_model_ckpt_300.log

@@ -0,0 +1,25 @@
+2025-02-06:13:02:59,775 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:13:03:11,079 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:13:03:11,080 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:13:03:11,080 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:13:03:11,080 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-300'}
+2025-02-06:13:03:11,118 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:13:03:11,744 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:13:04:02,898 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:13:04:02,898 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:13:04:02,899 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 11/1319 [00:00<00:12, 106.44it/s]
  2%|▏         | 23/1319 [00:00<00:11, 112.62it/s]
  3%|▎         | 35/1319 [00:00<00:11, 114.69it/s]
  4%|▎         | 47/1319 [00:00<00:10, 115.78it/s]
  4%|▍         | 59/1319 [00:00<00:10, 116.39it/s]
  5%|▌         | 71/1319 [00:00<00:10, 116.81it/s]
  6%|▋         | 83/1319 [00:00<00:10, 117.06it/s]
  7%|▋         | 95/1319 [00:00<00:10, 117.41it/s]
  8%|▊         | 107/1319 [00:00<00:10, 117.48it/s]
  9%|▉         | 119/1319 [00:01<00:10, 117.32it/s]
 10%|▉         | 131/1319 [00:01<00:10, 117.65it/s]
 11%|█         | 143/1319 [00:01<00:09, 117.83it/s]
 12%|█▏        | 155/1319 [00:01<00:09, 117.71it/s]
 13%|█▎        | 167/1319 [00:01<00:09, 117.42it/s]
 14%|█▎        | 179/1319 [00:01<00:09, 117.26it/s]
 14%|█▍        | 191/1319 [00:01<00:09, 117.47it/s]
 15%|█▌        | 203/1319 [00:01<00:09, 117.76it/s]
 16%|█▋        | 215/1319 [00:01<00:09, 117.99it/s]
 17%|█▋        | 227/1319 [00:01<00:09, 118.21it/s]
 18%|█▊        | 239/1319 [00:02<00:09, 118.41it/s]
 19%|█▉        | 251/1319 [00:02<00:09, 118.56it/s]
 20%|█▉        | 263/1319 [00:02<00:08, 118.63it/s]
 21%|██        | 275/1319 [00:02<00:08, 118.68it/s]
 22%|██▏       | 287/1319 [00:02<00:08, 118.51it/s]
 23%|██▎       | 299/1319 [00:02<00:08, 118.60it/s]
 24%|██▎       | 311/1319 [00:02<00:08, 118.75it/s]
 24%|██▍       | 323/1319 [00:02<00:08, 118.77it/s]
 25%|██▌       | 335/1319 [00:02<00:08, 118.76it/s]
 26%|██▋       | 347/1319 [00:02<00:08, 118.82it/s]
 27%|██▋       | 359/1319 [00:03<00:08, 118.65it/s]
 28%|██▊       | 371/1319 [00:03<00:07, 118.78it/s]
 29%|██▉       | 383/1319 [00:03<00:07, 118.81it/s]
 30%|██▉       | 395/1319 [00:03<00:07, 118.83it/s]
 31%|███       | 407/1319 [00:03<00:07, 118.91it/s]
 32%|███▏      | 419/1319 [00:03<00:07, 118.98it/s]
 33%|███▎      | 431/1319 [00:03<00:07, 119.03it/s]
 34%|███▎      | 443/1319 [00:03<00:07, 118.95it/s]
 34%|███▍      | 455/1319 [00:03<00:07, 118.90it/s]
 35%|███▌      | 467/1319 [00:03<00:07, 118.87it/s]
 36%|███▋      | 479/1319 [00:04<00:07, 118.85it/s]
 37%|███▋      | 491/1319 [00:04<00:06, 118.91it/s]
 38%|███▊      | 503/1319 [00:04<00:06, 118.88it/s]
 39%|███▉      | 515/1319 [00:04<00:06, 118.74it/s]
 40%|███▉      | 527/1319 [00:04<00:06, 118.72it/s]
 41%|████      | 539/1319 [00:04<00:06, 118.55it/s]
 42%|████▏     | 551/1319 [00:04<00:06, 118.63it/s]
 43%|████▎     | 563/1319 [00:04<00:06, 118.64it/s]
 44%|████▎     | 575/1319 [00:04<00:06, 118.74it/s]
 45%|████▍     | 587/1319 [00:04<00:06, 118.61it/s]
 45%|████▌     | 599/1319 [00:05<00:06, 118.66it/s]
 46%|████▋     | 611/1319 [00:05<00:05, 118.44it/s]
 47%|████▋     | 623/1319 [00:05<00:05, 117.91it/s]
 48%|████▊     | 635/1319 [00:05<00:05, 117.70it/s]
 49%|████▉     | 647/1319 [00:05<00:05, 117.69it/s]
 50%|████▉     | 659/1319 [00:05<00:05, 117.69it/s]
 51%|█████     | 671/1319 [00:05<00:05, 118.01it/s]
 52%|█████▏    | 683/1319 [00:05<00:05, 118.30it/s]
 53%|█████▎    | 695/1319 [00:05<00:05, 118.53it/s]
 54%|█████▎    | 707/1319 [00:05<00:05, 118.31it/s]
 55%|█████▍    | 719/1319 [00:06<00:05, 117.60it/s]
 55%|█████▌    | 731/1319 [00:06<00:04, 118.06it/s]
 56%|█████▋    | 743/1319 [00:06<00:04, 118.41it/s]
 57%|█████▋    | 755/1319 [00:06<00:04, 118.04it/s]
 58%|█████▊    | 767/1319 [00:06<00:04, 118.13it/s]
 59%|█████▉    | 779/1319 [00:06<00:04, 118.18it/s]
 60%|█████▉    | 791/1319 [00:06<00:04, 118.27it/s]
 61%|██████    | 803/1319 [00:06<00:04, 118.45it/s]
 62%|██████▏   | 815/1319 [00:06<00:04, 118.64it/s]
 63%|██████▎   | 827/1319 [00:07<00:04, 118.69it/s]
 64%|██████▎   | 839/1319 [00:07<00:04, 118.68it/s]
 65%|██████▍   | 851/1319 [00:07<00:03, 118.46it/s]
 65%|██████▌   | 863/1319 [00:07<00:03, 118.49it/s]
 66%|██████▋   | 875/1319 [00:07<00:03, 118.35it/s]
 67%|██████▋   | 887/1319 [00:07<00:03, 118.42it/s]
 68%|██████▊   | 899/1319 [00:07<00:03, 118.51it/s]
 69%|██████▉   | 911/1319 [00:07<00:03, 118.61it/s]
 70%|██████▉   | 923/1319 [00:07<00:03, 118.60it/s]
 71%|███████   | 935/1319 [00:07<00:03, 118.73it/s]
 72%|███████▏  | 947/1319 [00:08<00:03, 118.39it/s]
 73%|███████▎  | 959/1319 [00:08<00:03, 116.33it/s]
 74%|███████▎  | 971/1319 [00:08<00:02, 116.62it/s]
 75%|███████▍  | 983/1319 [00:08<00:02, 116.91it/s]
 75%|███████▌  | 995/1319 [00:08<00:02, 117.18it/s]
 76%|███████▋  | 1007/1319 [00:08<00:02, 117.54it/s]
 77%|███████▋  | 1019/1319 [00:08<00:02, 117.80it/s]
 78%|███████▊  | 1031/1319 [00:08<00:02, 118.10it/s]
 79%|███████▉  | 1043/1319 [00:08<00:02, 118.29it/s]
 80%|███████▉  | 1055/1319 [00:08<00:02, 118.51it/s]
 81%|████████  | 1067/1319 [00:09<00:02, 118.54it/s]
 82%|████████▏ | 1079/1319 [00:09<00:02, 118.60it/s]
 83%|████████▎ | 1091/1319 [00:09<00:01, 118.44it/s]
 84%|████████▎ | 1103/1319 [00:09<00:01, 118.48it/s]
 85%|████████▍ | 1115/1319 [00:09<00:01, 118.49it/s]
 85%|████████▌ | 1127/1319 [00:09<00:01, 118.58it/s]
 86%|████████▋ | 1139/1319 [00:09<00:01, 118.64it/s]
 87%|████████▋ | 1151/1319 [00:09<00:01, 118.73it/s]
 88%|████████▊ | 1163/1319 [00:09<00:01, 118.76it/s]
 89%|████████▉ | 1175/1319 [00:09<00:01, 118.81it/s]
 90%|████████▉ | 1187/1319 [00:10<00:01, 118.80it/s]
 91%|█████████ | 1199/1319 [00:10<00:01, 118.83it/s]
 92%|█████████▏| 1211/1319 [00:10<00:00, 118.87it/s]
 93%|█████████▎| 1223/1319 [00:10<00:00, 118.89it/s]
 94%|█████████▎| 1235/1319 [00:10<00:00, 118.84it/s]
 95%|█████████▍| 1247/1319 [00:10<00:00, 118.81it/s]
 95%|█████████▌| 1259/1319 [00:10<00:00, 118.77it/s]
 96%|█████████▋| 1271/1319 [00:10<00:00, 118.85it/s]
 97%|█████████▋| 1283/1319 [00:10<00:00, 118.63it/s]
 98%|█████████▊| 1295/1319 [00:10<00:00, 118.59it/s]
 99%|█████████▉| 1307/1319 [00:11<00:00, 118.57it/s]
+2025-02-06:13:04:14,090 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:13:54:11,075 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-300), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6793|±  |0.0129|
+|               |       |strict-match    |     8|exact_match|↑  |0.6634|±  |0.0130|
+

8b_only_dpo/solo_dpo_model_ckpt_3000.log

@@ -0,0 +1,25 @@
+2025-02-06:16:59:02,234 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:16:59:14,381 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:16:59:14,382 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:16:59:14,383 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:16:59:14,383 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-3000'}
+2025-02-06:16:59:14,429 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:16:59:15,222 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:16:59:58,900 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:16:59:58,901 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:16:59:58,901 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 11/1319 [00:00<00:11, 109.27it/s]
  2%|▏         | 23/1319 [00:00<00:11, 112.70it/s]
  3%|▎         | 35/1319 [00:00<00:11, 113.81it/s]
  4%|▎         | 47/1319 [00:00<00:11, 114.54it/s]
  4%|▍         | 59/1319 [00:00<00:10, 114.97it/s]
  5%|▌         | 71/1319 [00:00<00:10, 115.51it/s]
  6%|▋         | 83/1319 [00:00<00:10, 115.46it/s]
  7%|▋         | 95/1319 [00:00<00:10, 116.14it/s]
  8%|▊         | 107/1319 [00:00<00:10, 115.56it/s]
  9%|▉         | 119/1319 [00:01<00:10, 115.45it/s]
 10%|▉         | 131/1319 [00:01<00:10, 115.77it/s]
 11%|█         | 143/1319 [00:01<00:10, 115.98it/s]
 12%|█▏        | 155/1319 [00:01<00:10, 115.96it/s]
 13%|█▎        | 167/1319 [00:01<00:09, 116.09it/s]
 14%|█▎        | 179/1319 [00:01<00:09, 116.20it/s]
 14%|█▍        | 191/1319 [00:01<00:09, 116.46it/s]
 15%|█▌        | 203/1319 [00:01<00:09, 116.06it/s]
 16%|█▋        | 215/1319 [00:01<00:09, 116.25it/s]
 17%|█▋        | 227/1319 [00:01<00:09, 115.88it/s]
 18%|█▊        | 239/1319 [00:02<00:09, 115.60it/s]
 19%|█▉        | 251/1319 [00:02<00:09, 115.90it/s]
 20%|█▉        | 263/1319 [00:02<00:09, 115.87it/s]
 21%|██        | 275/1319 [00:02<00:08, 116.05it/s]
 22%|██▏       | 287/1319 [00:02<00:08, 116.35it/s]
 23%|██▎       | 299/1319 [00:02<00:08, 116.64it/s]
 24%|██▎       | 311/1319 [00:02<00:08, 116.58it/s]
 24%|██▍       | 323/1319 [00:02<00:08, 117.02it/s]
 25%|██▌       | 335/1319 [00:02<00:08, 117.43it/s]
 26%|██▋       | 347/1319 [00:02<00:08, 117.79it/s]
 27%|██▋       | 359/1319 [00:03<00:08, 117.79it/s]
 28%|██▊       | 371/1319 [00:03<00:08, 118.21it/s]
 29%|██▉       | 383/1319 [00:03<00:07, 118.13it/s]
 30%|██▉       | 395/1319 [00:03<00:07, 118.15it/s]
 31%|███       | 407/1319 [00:03<00:07, 118.33it/s]
 32%|███▏      | 419/1319 [00:03<00:07, 118.28it/s]
 33%|███▎      | 431/1319 [00:03<00:07, 118.39it/s]
 34%|███▎      | 443/1319 [00:03<00:07, 118.12it/s]
 34%|███▍      | 455/1319 [00:03<00:07, 118.27it/s]
 35%|███▌      | 467/1319 [00:04<00:07, 118.38it/s]
 36%|███▋      | 479/1319 [00:04<00:07, 118.25it/s]
 37%|███▋      | 491/1319 [00:04<00:06, 118.34it/s]
 38%|███▊      | 503/1319 [00:04<00:06, 117.59it/s]
 39%|███▉      | 515/1319 [00:04<00:06, 117.54it/s]
 40%|███▉      | 527/1319 [00:04<00:06, 117.64it/s]
 41%|████      | 539/1319 [00:04<00:06, 117.75it/s]
 42%|████▏     | 551/1319 [00:04<00:06, 118.00it/s]
 43%|████▎     | 563/1319 [00:04<00:06, 118.08it/s]
 44%|████▎     | 575/1319 [00:04<00:06, 118.35it/s]
 45%|████▍     | 587/1319 [00:05<00:06, 118.31it/s]
 45%|████▌     | 599/1319 [00:05<00:06, 118.49it/s]
 46%|████▋     | 611/1319 [00:05<00:05, 118.65it/s]
 47%|████▋     | 623/1319 [00:05<00:05, 118.37it/s]
 48%|████▊     | 635/1319 [00:05<00:05, 118.51it/s]
 49%|████▉     | 647/1319 [00:05<00:05, 118.57it/s]
 50%|████▉     | 659/1319 [00:05<00:05, 118.73it/s]
 51%|█████     | 671/1319 [00:05<00:05, 118.84it/s]
 52%|█████▏    | 683/1319 [00:05<00:05, 118.70it/s]
 53%|█████▎    | 695/1319 [00:05<00:05, 118.17it/s]
 54%|█████▎    | 707/1319 [00:06<00:05, 117.67it/s]
 55%|█████▍    | 719/1319 [00:06<00:05, 117.55it/s]
 55%|█████▌    | 731/1319 [00:06<00:04, 117.68it/s]
 56%|█████▋    | 743/1319 [00:06<00:04, 117.12it/s]
 57%|█████▋    | 755/1319 [00:06<00:04, 117.06it/s]
 58%|█████▊    | 767/1319 [00:06<00:04, 116.90it/s]
 59%|█████▉    | 779/1319 [00:06<00:04, 117.24it/s]
 60%|█████▉    | 791/1319 [00:06<00:04, 117.57it/s]
 61%|██████    | 803/1319 [00:06<00:04, 117.71it/s]
 62%|██████▏   | 815/1319 [00:06<00:04, 117.73it/s]
 63%|██████▎   | 827/1319 [00:07<00:04, 117.67it/s]
 64%|██████▎   | 839/1319 [00:07<00:04, 117.91it/s]
 65%|██████▍   | 851/1319 [00:07<00:03, 118.01it/s]
 65%|██████▌   | 863/1319 [00:07<00:03, 117.90it/s]
 66%|██████▋   | 875/1319 [00:07<00:03, 118.03it/s]
 67%|██████▋   | 887/1319 [00:07<00:03, 117.92it/s]
 68%|██████▊   | 899/1319 [00:07<00:03, 117.97it/s]
 69%|██████▉   | 911/1319 [00:07<00:03, 118.00it/s]
 70%|██████▉   | 923/1319 [00:07<00:03, 117.90it/s]
 71%|███████   | 935/1319 [00:07<00:03, 117.87it/s]
 72%|███████▏  | 947/1319 [00:08<00:03, 117.80it/s]
 73%|███████▎  | 959/1319 [00:08<00:03, 117.85it/s]
 74%|███████▎  | 971/1319 [00:08<00:02, 117.81it/s]
 75%|███████▍  | 983/1319 [00:08<00:02, 117.17it/s]
 75%|███████▌  | 995/1319 [00:08<00:02, 117.10it/s]
 76%|███████▋  | 1007/1319 [00:08<00:02, 117.00it/s]
 77%|███████▋  | 1019/1319 [00:08<00:02, 116.91it/s]
 78%|███████▊  | 1031/1319 [00:08<00:02, 116.78it/s]
 79%|███████▉  | 1043/1319 [00:08<00:02, 117.06it/s]
 80%|███████▉  | 1055/1319 [00:08<00:02, 117.15it/s]
 81%|████████  | 1067/1319 [00:09<00:02, 117.19it/s]
 82%|████████▏ | 1079/1319 [00:09<00:02, 117.40it/s]
 83%|████████▎ | 1091/1319 [00:09<00:01, 117.27it/s]
 84%|████████▎ | 1103/1319 [00:09<00:01, 117.47it/s]
 85%|████████▍ | 1115/1319 [00:09<00:01, 117.69it/s]
 85%|████████▌ | 1127/1319 [00:09<00:01, 117.82it/s]
 86%|████████▋ | 1139/1319 [00:09<00:01, 117.99it/s]
 87%|████████▋ | 1151/1319 [00:09<00:01, 118.01it/s]
 88%|████████▊ | 1163/1319 [00:09<00:01, 118.17it/s]
 89%|████████▉ | 1175/1319 [00:10<00:01, 118.15it/s]
 90%|████████▉ | 1187/1319 [00:10<00:01, 118.17it/s]
 91%|█████████ | 1199/1319 [00:10<00:01, 118.28it/s]
 92%|█████████▏| 1211/1319 [00:10<00:00, 118.12it/s]
 93%|█████████▎| 1223/1319 [00:10<00:00, 117.96it/s]
 94%|█████████▎| 1235/1319 [00:10<00:00, 117.48it/s]
 95%|█████████▍| 1247/1319 [00:10<00:00, 117.59it/s]
 95%|█████████▌| 1259/1319 [00:10<00:00, 117.68it/s]
 96%|█████████▋| 1271/1319 [00:10<00:00, 117.66it/s]
 97%|█████████▋| 1283/1319 [00:10<00:00, 117.38it/s]
 98%|█████████▊| 1295/1319 [00:11<00:00, 117.35it/s]
 99%|█████████▉| 1307/1319 [00:11<00:00, 117.44it/s]
+2025-02-06:17:00:10,174 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:17:52:44,527 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-3000), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6224|±  |0.0134|
+|               |       |strict-match    |     8|exact_match|↑  |0.6096|±  |0.0134|
+

8b_only_dpo/solo_dpo_model_ckpt_3100.log

@@ -0,0 +1,80 @@
+2025-02-06:17:52:48,419 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:17:52:59,779 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:17:52:59,780 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:17:52:59,780 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:17:52:59,781 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-3100'}
+2025-02-06:17:52:59,820 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:17:53:00,480 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:17:53:40,991 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:17:53:40,992 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:17:53:40,992 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 12/1319 [00:00<00:11, 114.44it/s]
  2%|▏         | 25/1319 [00:00<00:10, 118.13it/s]
  3%|▎         | 38/1319 [00:00<00:10, 119.53it/s]
  4%|▍         | 51/1319 [00:00<00:10, 120.16it/s]
  5%|▍         | 64/1319 [00:00<00:10, 120.63it/s]
  6%|▌         | 77/1319 [00:00<00:10, 121.12it/s]
  7%|▋         | 90/1319 [00:00<00:10, 121.59it/s]
  8%|▊         | 103/1319 [00:00<00:09, 121.79it/s]
  9%|▉         | 116/1319 [00:00<00:09, 121.83it/s]
 10%|▉         | 129/1319 [00:01<00:09, 122.04it/s]
 11%|█         | 142/1319 [00:01<00:09, 122.34it/s]
 12%|█▏        | 155/1319 [00:01<00:09, 122.49it/s]
 13%|█▎        | 168/1319 [00:01<00:09, 120.32it/s]
 14%|█▎        | 181/1319 [00:01<00:09, 119.21it/s]
 15%|█▍        | 193/1319 [00:01<00:09, 118.72it/s]
 16%|█▌        | 205/1319 [00:01<00:09, 118.57it/s]
 16%|█▋        | 217/1319 [00:01<00:09, 118.66it/s]
 17%|█▋        | 229/1319 [00:01<00:09, 118.72it/s]
 18%|█▊        | 241/1319 [00:02<00:09, 113.80it/s]
 19%|█▉        | 253/1319 [00:02<00:09, 114.75it/s]
 20%|██        | 265/1319 [00:02<00:09, 115.50it/s]
 21%|██        | 277/1319 [00:02<00:08, 116.33it/s]
 22%|██▏       | 289/1319 [00:02<00:08, 116.93it/s]
 23%|██▎       | 301/1319 [00:02<00:08, 117.48it/s]
 24%|██▎       | 313/1319 [00:02<00:08, 117.85it/s]
 25%|██▍       | 325/1319 [00:02<00:08, 118.22it/s]
 26%|██▌       | 337/1319 [00:02<00:08, 118.37it/s]
 26%|██▋       | 349/1319 [00:02<00:08, 118.40it/s]
 27%|██▋       | 361/1319 [00:03<00:08, 118.55it/s]
 28%|██▊       | 373/1319 [00:03<00:07, 118.68it/s]
 29%|██▉       | 385/1319 [00:03<00:07, 118.79it/s]
 30%|███       | 397/1319 [00:03<00:07, 118.92it/s]
 31%|███       | 409/1319 [00:03<00:07, 118.92it/s]
 32%|███▏      | 421/1319 [00:03<00:07, 119.00it/s]
 33%|███▎      | 433/1319 [00:03<00:07, 119.01it/s]
 34%|███▎      | 445/1319 [00:03<00:07, 119.06it/s]
 35%|███▍      | 457/1319 [00:03<00:07, 119.09it/s]
 36%|███▌      | 469/1319 [00:03<00:07, 118.85it/s]
 36%|███▋      | 481/1319 [00:04<00:07, 118.51it/s]
 37%|███▋      | 493/1319 [00:04<00:06, 118.61it/s]
 38%|███▊      | 505/1319 [00:04<00:06, 118.68it/s]
 39%|███▉      | 517/1319 [00:04<00:06, 118.82it/s]
 40%|████      | 529/1319 [00:04<00:06, 118.68it/s]
 41%|████      | 541/1319 [00:04<00:06, 118.86it/s]
 42%|████▏     | 553/1319 [00:04<00:06, 118.98it/s]
 43%|████▎     | 565/1319 [00:04<00:06, 119.05it/s]
 44%|████▎     | 577/1319 [00:04<00:06, 119.09it/s]
 45%|████▍     | 589/1319 [00:04<00:06, 119.14it/s]
 46%|████▌     | 601/1319 [00:05<00:06, 119.20it/s]
 46%|████▋     | 613/1319 [00:05<00:05, 119.28it/s]
 47%|████▋     | 625/1319 [00:05<00:05, 119.24it/s]
 48%|████▊     | 637/1319 [00:05<00:05, 119.28it/s]
 49%|████▉     | 649/1319 [00:05<00:05, 119.15it/s]
 50%|█████     | 661/1319 [00:05<00:05, 119.15it/s]
 51%|█████     | 673/1319 [00:05<00:05, 119.09it/s]
 52%|███���█▏    | 685/1319 [00:05<00:05, 118.98it/s]
 53%|█████▎    | 697/1319 [00:05<00:05, 119.00it/s]
 54%|█████▍    | 709/1319 [00:05<00:05, 118.96it/s]
 55%|█████▍    | 721/1319 [00:06<00:05, 118.73it/s]
 56%|█████▌    | 733/1319 [00:06<00:04, 118.57it/s]
 56%|█████▋    | 745/1319 [00:06<00:04, 118.52it/s]
 57%|█████▋    | 757/1319 [00:06<00:04, 118.59it/s]
 58%|█████▊    | 769/1319 [00:06<00:04, 118.57it/s]
 59%|█████▉    | 781/1319 [00:06<00:04, 118.65it/s]
 60%|██████    | 793/1319 [00:06<00:04, 118.68it/s]
 61%|██████    | 805/1319 [00:06<00:04, 118.69it/s]
 62%|██████▏   | 817/1319 [00:06<00:04, 118.60it/s]
 63%|██████▎   | 829/1319 [00:06<00:04, 118.57it/s]
 64%|██████▍   | 841/1319 [00:07<00:04, 118.64it/s]
 65%|██████▍   | 853/1319 [00:07<00:03, 118.72it/s]
 66%|██████▌   | 865/1319 [00:07<00:03, 118.78it/s]
 66%|██████▋   | 877/1319 [00:07<00:03, 118.89it/s]
 67%|██████▋   | 889/1319 [00:07<00:03, 118.88it/s]
 68%|██████▊   | 901/1319 [00:07<00:03, 118.96it/s]
 69%|██████▉   | 913/1319 [00:07<00:03, 118.91it/s]
 70%|███████   | 925/1319 [00:07<00:03, 118.87it/s]
 71%|███████   | 937/1319 [00:07<00:03, 118.88it/s]
 72%|███████▏  | 949/1319 [00:07<00:03, 118.90it/s]
 73%|███████▎  | 961/1319 [00:08<00:03, 118.88it/s]
 74%|███████▍  | 973/1319 [00:08<00:02, 118.66it/s]
 75%|███████▍  | 985/1319 [00:08<00:02, 118.73it/s]
 76%|███████▌  | 997/1319 [00:08<00:02, 118.85it/s]
 76%|███████▋  | 1009/1319 [00:08<00:02, 118.76it/s]
 77%|███████▋  | 1021/1319 [00:08<00:02, 118.86it/s]
 78%|███████▊  | 1033/1319 [00:08<00:02, 118.96it/s]
 79%|███████▉  | 1045/1319 [00:08<00:02, 119.00it/s]
 80%|████████  | 1057/1319 [00:08<00:02, 118.99it/s]
 81%|████████  | 1069/1319 [00:08<00:02, 118.58it/s]
 82%|████████▏ | 1081/1319 [00:09<00:02, 118.70it/s]
 83%|████████▎ | 1093/1319 [00:09<00:01, 118.88it/s]
 84%|████████▍ | 1105/1319 [00:09<00:01, 119.07it/s]
 85%|████████▍ | 1117/1319 [00:09<00:01, 119.10it/s]
 86%|████████▌ | 1129/1319 [00:09<00:01, 119.18it/s]
 87%|████████▋ | 1141/1319 [00:09<00:01, 119.29it/s]
 87%|████████▋ | 1153/1319 [00:09<00:01, 119.23it/s]
 88%|████████▊ | 1165/1319 [00:09<00:01, 119.20it/s]
 89%|████████▉ | 1177/1319 [00:09<00:01, 119.32it/s]
 90%|█████████ | 1189/1319 [00:09<00:01, 119.27it/s]
 91%|█████████ | 1201/1319 [00:10<00:00, 119.29it/s]
 92%|█████████▏| 1213/1319 [00:10<00:00, 119.02it/s]
 93%|█████████▎| 1225/1319 [00:10<00:00, 118.94it/s]
 94%|█████████▍| 1237/1319 [00:10<00:00, 119.16it/s]
 95%|█████████▍| 1249/1319 [00:10<00:00, 119.15it/s]
 96%|█████████▌| 1261/1319 [00:10<00:00, 119.23it/s]
 97%|█████████▋| 1273/1319 [00:10<00:00, 119.27it/s]
 97%|█████████▋| 1285/1319 [00:10<00:00, 119.24it/s]
 98%|█████████▊| 1297/1319 [00:10<00:00, 119.12it/s]
 99%|█████████▉| 1309/1319 [00:11<00:00, 119.13it/s]
+2025-02-06:17:53:52,116 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 467, in <module>
+    cli_evaluate()
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate
+    results = evaluator.simple_evaluate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper
+    return fn(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 304, in simple_evaluate
+    results = evaluate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper
+    return fn(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 524, in evaluate
+    resps = getattr(lm, reqtype)(cloned_reqs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 1354, in generate_until
+    cont = self._model_generate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 897, in _model_generate
+    return self.model.generate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1646, in generate
+    outputs = self.base_model.generate(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
+    return func(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 1996, in generate
+    result = self._sample(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 2923, in _sample
+    outputs = self(**model_inputs, return_dict=True)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 1182, in forward
+    outputs = self.model(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 994, in forward
+    layer_outputs = decoder_layer(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 743, in forward
+    hidden_states = self.mlp(hidden_states)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 309, in forward
+    down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/tuners/lora/layer.py", line 556, in forward
+    result = result + lora_B(lora_A(dropout(x))) * scaling
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 116, in forward
+    return F.linear(input, self.weight, self.bias)
+KeyboardInterrupt
+

8b_only_dpo/solo_dpo_model_ckpt_400.log

@@ -0,0 +1,25 @@
+2025-02-06:13:54:14,985 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:13:54:26,232 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:13:54:26,233 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:13:54:26,233 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:13:54:26,233 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-400'}
+2025-02-06:13:54:26,273 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:13:54:26,866 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:13:55:07,646 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:13:55:07,646 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:13:55:07,647 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 12/1319 [00:00<00:11, 114.43it/s]
  2%|▏         | 25/1319 [00:00<00:10, 118.16it/s]
  3%|▎         | 38/1319 [00:00<00:10, 119.50it/s]
  4%|▍         | 51/1319 [00:00<00:10, 120.30it/s]
  5%|▍         | 64/1319 [00:00<00:10, 120.71it/s]
  6%|▌         | 77/1319 [00:00<00:10, 121.17it/s]
  7%|▋         | 90/1319 [00:00<00:10, 121.45it/s]
  8%|▊         | 103/1319 [00:00<00:09, 121.70it/s]
  9%|▉         | 116/1319 [00:00<00:09, 121.84it/s]
 10%|▉         | 129/1319 [00:01<00:09, 121.93it/s]
 11%|█         | 142/1319 [00:01<00:09, 122.12it/s]
 12%|█▏        | 155/1319 [00:01<00:09, 122.27it/s]
 13%|█▎        | 168/1319 [00:01<00:09, 122.31it/s]
 14%|█▎        | 181/1319 [00:01<00:09, 121.23it/s]
 15%|█▍        | 194/1319 [00:01<00:09, 121.32it/s]
 16%|█▌        | 207/1319 [00:01<00:09, 121.12it/s]
 17%|█▋        | 220/1319 [00:01<00:09, 120.88it/s]
 18%|█▊        | 233/1319 [00:01<00:08, 121.24it/s]
 19%|█▊        | 246/1319 [00:02<00:08, 121.48it/s]
 20%|█▉        | 259/1319 [00:02<00:08, 121.80it/s]
 21%|██        | 272/1319 [00:02<00:08, 121.96it/s]
 22%|██▏       | 285/1319 [00:02<00:08, 122.11it/s]
 23%|██▎       | 298/1319 [00:02<00:08, 122.18it/s]
 24%|██▎       | 311/1319 [00:02<00:08, 122.45it/s]
 25%|██▍       | 324/1319 [00:02<00:08, 122.67it/s]
 26%|██▌       | 337/1319 [00:02<00:08, 122.47it/s]
 27%|██▋       | 350/1319 [00:02<00:07, 122.59it/s]
 28%|██▊       | 363/1319 [00:02<00:07, 122.53it/s]
 29%|██▊       | 376/1319 [00:03<00:07, 122.74it/s]
 29%|██▉       | 389/1319 [00:03<00:07, 122.63it/s]
 30%|███       | 402/1319 [00:03<00:07, 121.42it/s]
 31%|███▏      | 415/1319 [00:03<00:07, 120.69it/s]
 32%|███▏      | 428/1319 [00:03<00:07, 120.07it/s]
 33%|███▎      | 441/1319 [00:03<00:07, 120.17it/s]
 34%|███▍      | 454/1319 [00:03<00:07, 120.48it/s]
 35%|███▌      | 467/1319 [00:03<00:07, 120.91it/s]
 36%|███▋      | 480/1319 [00:03<00:06, 121.48it/s]
 37%|███▋      | 493/1319 [00:04<00:06, 121.84it/s]
 38%|███▊      | 506/1319 [00:04<00:06, 122.18it/s]
 39%|███▉      | 519/1319 [00:04<00:06, 122.38it/s]
 40%|████      | 532/1319 [00:04<00:06, 122.59it/s]
 41%|████▏     | 545/1319 [00:04<00:06, 122.78it/s]
 42%|████▏     | 558/1319 [00:04<00:06, 122.90it/s]
 43%|████▎     | 571/1319 [00:04<00:06, 123.08it/s]
 44%|████▍     | 584/1319 [00:04<00:05, 123.18it/s]
 45%|████▌     | 597/1319 [00:04<00:05, 123.24it/s]
 46%|████▌     | 610/1319 [00:05<00:05, 123.13it/s]
 47%|████▋     | 623/1319 [00:05<00:05, 123.13it/s]
 48%|████▊     | 636/1319 [00:05<00:05, 123.08it/s]
 49%|████▉     | 649/1319 [00:05<00:05, 123.12it/s]
 50%|█████     | 662/1319 [00:05<00:05, 123.16it/s]
 51%|█████     | 675/1319 [00:05<00:05, 122.86it/s]
 52%|█████▏    | 688/1319 [00:05<00:05, 122.02it/s]
 53%|█████▎    | 701/1319 [00:05<00:05, 122.31it/s]
 54%|█████▍    | 714/1319 [00:05<00:04, 122.38it/s]
 55%|█████▌    | 727/1319 [00:05<00:04, 122.58it/s]
 56%|█████▌    | 740/1319 [00:06<00:04, 122.71it/s]
 57%|█████▋    | 753/1319 [00:06<00:04, 122.70it/s]
 58%|█████▊    | 766/1319 [00:06<00:04, 122.84it/s]
 59%|█████▉    | 779/1319 [00:06<00:04, 122.76it/s]
 60%|██████    | 792/1319 [00:06<00:04, 122.89it/s]
 61%|██████    | 805/1319 [00:06<00:04, 122.95it/s]
 62%|██████▏   | 818/1319 [00:06<00:04, 122.50it/s]
 63%|██████▎   | 831/1319 [00:06<00:04, 121.98it/s]
 64%|██████▍   | 844/1319 [00:06<00:03, 121.90it/s]
 65%|██████▍   | 857/1319 [00:07<00:03, 121.88it/s]
 66%|██████▌   | 870/1319 [00:07<00:03, 122.12it/s]
 67%|██████▋   | 883/1319 [00:07<00:03, 122.32it/s]
 68%|██████▊   | 896/1319 [00:07<00:03, 122.54it/s]
 69%|██████▉   | 909/1319 [00:07<00:03, 122.66it/s]
 70%|██████▉   | 922/1319 [00:07<00:03, 122.67it/s]
 71%|███████   | 935/1319 [00:07<00:03, 120.92it/s]
 72%|███████▏  | 948/1319 [00:07<00:03, 120.04it/s]
 73%|███████▎  | 961/1319 [00:07<00:02, 119.89it/s]
 74%|███████▍  | 973/1319 [00:07<00:02, 119.67it/s]
 75%|███████▍  | 985/1319 [00:08<00:02, 119.63it/s]
 76%|███████▌  | 998/1319 [00:08<00:02, 119.78it/s]
 77%|███████▋  | 1011/1319 [00:08<00:02, 119.89it/s]
 78%|███████▊  | 1023/1319 [00:08<00:02, 119.83it/s]
 79%|███████▊  | 1036/1319 [00:08<00:02, 119.96it/s]
 80%|███████▉  | 1049/1319 [00:08<00:02, 120.07it/s]
 81%|████████  | 1062/1319 [00:08<00:02, 120.05it/s]
 82%|████████▏ | 1075/1319 [00:08<00:02, 120.23it/s]
 82%|████████▏ | 1088/1319 [00:08<00:01, 120.31it/s]
 83%|████████▎ | 1101/1319 [00:09<00:01, 120.16it/s]
 84%|████████▍ | 1114/1319 [00:09<00:01, 120.08it/s]
 85%|████████▌ | 1127/1319 [00:09<00:01, 120.55it/s]
 86%|████████▋ | 1140/1319 [00:09<00:01, 121.33it/s]
 87%|████████▋ | 1153/1319 [00:09<00:01, 121.78it/s]
 88%|████████▊ | 1166/1319 [00:09<00:01, 122.19it/s]
 89%|████████▉ | 1179/1319 [00:09<00:01, 121.67it/s]
 90%|█████████ | 1192/1319 [00:09<00:01, 121.59it/s]
 91%|█████████▏| 1205/1319 [00:09<00:00, 121.98it/s]
 92%|█████████▏| 1218/1319 [00:10<00:00, 122.04it/s]
 93%|█████████▎| 1231/1319 [00:10<00:00, 122.33it/s]
 94%|█████████▍| 1244/1319 [00:10<00:00, 122.55it/s]
 95%|█████████▌| 1257/1319 [00:10<00:00, 122.81it/s]
 96%|█████████▋| 1270/1319 [00:10<00:00, 122.95it/s]
 97%|█████████▋| 1283/1319 [00:10<00:00, 123.03it/s]
 98%|█████████▊| 1296/1319 [00:10<00:00, 123.09it/s]
 99%|█████████▉| 1309/1319 [00:10<00:00, 123.17it/s]
+2025-02-06:13:55:18,511 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:14:45:02,562 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-400), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6937|±  |0.0127|
+|               |       |strict-match    |     8|exact_match|↑  |0.6755|±  |0.0129|
+

8b_only_dpo/solo_dpo_model_ckpt_500.log

@@ -0,0 +1,25 @@
+2025-02-06:14:45:06,475 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:14:45:17,931 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:14:45:17,932 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:14:45:17,932 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:14:45:17,932 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-500'}
+2025-02-06:14:45:17,971 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:14:45:18,578 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:14:45:59,317 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:14:45:59,318 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:14:45:59,318 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 12/1319 [00:00<00:11, 112.59it/s]
  2%|▏         | 24/1319 [00:00<00:11, 115.25it/s]
  3%|▎         | 36/1319 [00:00<00:11, 116.47it/s]
  4%|▎         | 49/1319 [00:00<00:10, 117.97it/s]
  5%|▍         | 62/1319 [00:00<00:10, 119.16it/s]
  6%|▌         | 75/1319 [00:00<00:10, 119.94it/s]
  7%|▋         | 88/1319 [00:00<00:10, 120.64it/s]
  8%|▊         | 101/1319 [00:00<00:10, 121.14it/s]
  9%|▊         | 114/1319 [00:00<00:09, 121.51it/s]
 10%|▉         | 127/1319 [00:01<00:09, 121.80it/s]
 11%|█         | 140/1319 [00:01<00:09, 121.95it/s]
 12%|█▏        | 153/1319 [00:01<00:09, 121.71it/s]
 13%|█▎        | 166/1319 [00:01<00:09, 122.03it/s]
 14%|█▎        | 179/1319 [00:01<00:09, 122.33it/s]
 15%|█▍        | 192/1319 [00:01<00:09, 122.43it/s]
 16%|█▌        | 205/1319 [00:01<00:09, 122.57it/s]
 17%|█▋        | 218/1319 [00:01<00:08, 122.66it/s]
 18%|█▊        | 231/1319 [00:01<00:08, 122.83it/s]
 18%|█▊        | 244/1319 [00:02<00:08, 122.87it/s]
 19%|█▉        | 257/1319 [00:02<00:08, 122.47it/s]
 20%|██        | 270/1319 [00:02<00:08, 121.91it/s]
 21%|██▏       | 283/1319 [00:02<00:08, 122.16it/s]
 22%|██▏       | 296/1319 [00:02<00:08, 122.27it/s]
 23%|██▎       | 309/1319 [00:02<00:08, 122.39it/s]
 24%|██▍       | 322/1319 [00:02<00:08, 122.56it/s]
 25%|██▌       | 335/1319 [00:02<00:08, 122.69it/s]
 26%|██▋       | 348/1319 [00:02<00:07, 122.75it/s]
 27%|██▋       | 361/1319 [00:02<00:07, 122.80it/s]
 28%|██▊       | 374/1319 [00:03<00:07, 122.77it/s]
 29%|██▉       | 387/1319 [00:03<00:07, 122.72it/s]
 30%|███       | 400/1319 [00:03<00:07, 122.67it/s]
 31%|███▏      | 413/1319 [00:03<00:07, 122.63it/s]
 32%|███▏      | 426/1319 [00:03<00:07, 122.53it/s]
 33%|███▎      | 439/1319 [00:03<00:07, 122.52it/s]
 34%|███▍      | 452/1319 [00:03<00:07, 122.50it/s]
 35%|███▌      | 465/1319 [00:03<00:06, 122.48it/s]
 36%|███▌      | 478/1319 [00:03<00:06, 122.50it/s]
 37%|███▋      | 491/1319 [00:04<00:06, 122.56it/s]
 38%|███▊      | 504/1319 [00:04<00:06, 122.51it/s]
 39%|███▉      | 517/1319 [00:04<00:06, 120.71it/s]
 40%|████      | 530/1319 [00:04<00:06, 120.22it/s]
 41%|████      | 543/1319 [00:04<00:06, 120.13it/s]
 42%|████▏     | 556/1319 [00:04<00:06, 120.01it/s]
 43%|████▎     | 569/1319 [00:04<00:06, 120.51it/s]
 44%|████▍     | 582/1319 [00:04<00:06, 121.13it/s]
 45%|████▌     | 595/1319 [00:04<00:05, 121.54it/s]
 46%|████▌     | 608/1319 [00:04<00:05, 121.90it/s]
 47%|████▋     | 621/1319 [00:05<00:05, 122.21it/s]
 48%|████▊     | 634/1319 [00:05<00:05, 120.83it/s]
 49%|████▉     | 647/1319 [00:05<00:05, 119.42it/s]
 50%|████▉     | 659/1319 [00:05<00:05, 118.83it/s]
 51%|█████     | 671/1319 [00:05<00:05, 118.39it/s]
 52%|█████▏    | 683/1319 [00:05<00:05, 117.71it/s]
 53%|█████▎    | 695/1319 [00:05<00:05, 117.93it/s]
 54%|█████▎    | 707/1319 [00:05<00:05, 118.07it/s]
 55%|█████▍    | 719/1319 [00:05<00:05, 118.25it/s]
 55%|█████▌    | 731/1319 [00:06<00:04, 118.43it/s]
 56%|█████▋    | 743/1319 [00:06<00:04, 118.47it/s]
 57%|█████▋    | 755/1319 [00:06<00:04, 118.19it/s]
 58%|█████▊    | 767/1319 [00:06<00:04, 117.12it/s]
 59%|█████▉    | 779/1319 [00:06<00:04, 117.26it/s]
 60%|█████▉    | 791/1319 [00:06<00:04, 117.46it/s]
 61%|██████    | 803/1319 [00:06<00:04, 117.61it/s]
 62%|██████▏   | 815/1319 [00:06<00:04, 117.81it/s]
 63%|██████▎   | 827/1319 [00:06<00:04, 117.95it/s]
 64%|██████▎   | 839/1319 [00:06<00:04, 118.10it/s]
 65%|██████▍   | 851/1319 [00:07<00:03, 118.18it/s]
 65%|██████▌   | 863/1319 [00:07<00:03, 118.26it/s]
 66%|██████▋   | 875/1319 [00:07<00:03, 117.95it/s]
 67%|██████▋   | 888/1319 [00:07<00:03, 119.07it/s]
 68%|██████▊   | 901/1319 [00:07<00:03, 119.99it/s]
 69%|██████▉   | 914/1319 [00:07<00:03, 120.50it/s]
 70%|███████   | 927/1319 [00:07<00:03, 121.06it/s]
 71%|███████▏  | 940/1319 [00:07<00:03, 121.61it/s]
 72%|███████▏  | 953/1319 [00:07<00:03, 121.97it/s]
 73%|███████▎  | 966/1319 [00:08<00:02, 122.23it/s]
 74%|███████▍  | 979/1319 [00:08<00:02, 122.43it/s]
 75%|███████▌  | 992/1319 [00:08<00:02, 122.54it/s]
 76%|███████▌  | 1005/1319 [00:08<00:02, 122.19it/s]
 77%|███████▋  | 1018/1319 [00:08<00:02, 121.96it/s]
 78%|███████▊  | 1031/1319 [00:08<00:02, 122.06it/s]
 79%|███████▉  | 1044/1319 [00:08<00:02, 121.98it/s]
 80%|████████  | 1057/1319 [00:08<00:02, 122.22it/s]
 81%|████████  | 1070/1319 [00:08<00:02, 122.38it/s]
 82%|████████▏ | 1083/1319 [00:08<00:01, 122.61it/s]
 83%|████████▎ | 1096/1319 [00:09<00:01, 122.71it/s]
 84%|████████▍ | 1109/1319 [00:09<00:01, 122.78it/s]
 85%|████████▌ | 1122/1319 [00:09<00:01, 122.88it/s]
 86%|████████▌ | 1135/1319 [00:09<00:01, 122.87it/s]
 87%|████████▋ | 1148/1319 [00:09<00:01, 122.90it/s]
 88%|████████▊ | 1161/1319 [00:09<00:01, 122.91it/s]
 89%|████████▉ | 1174/1319 [00:09<00:01, 122.99it/s]
 90%|████████▉ | 1187/1319 [00:09<00:01, 123.08it/s]
 91%|█████████ | 1200/1319 [00:09<00:00, 123.15it/s]
 92%|█████████▏| 1213/1319 [00:10<00:00, 123.20it/s]
 93%|█████████▎| 1226/1319 [00:10<00:00, 123.24it/s]
 94%|█████████▍| 1239/1319 [00:10<00:00, 122.17it/s]
 95%|█████████▍| 1252/1319 [00:10<00:00, 121.70it/s]
 96%|█████████▌| 1265/1319 [00:10<00:00, 121.22it/s]
 97%|█████████▋| 1278/1319 [00:10<00:00, 121.21it/s]
 98%|█████████▊| 1291/1319 [00:10<00:00, 121.08it/s]
 99%|█████████▉| 1304/1319 [00:10<00:00, 121.53it/s]
+2025-02-06:14:46:10,228 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:15:35:26,295 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-500), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6990|±  |0.0126|
+|               |       |strict-match    |     8|exact_match|↑  |0.6816|±  |0.0128|
+

8b_only_dpo/solo_dpo_model_ckpt_600.log

@@ -0,0 +1,25 @@
+2025-02-06:15:35:30,206 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:15:35:41,457 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:15:35:41,458 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:15:35:41,458 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:15:35:41,458 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-600'}
+2025-02-06:15:35:41,498 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:15:35:42,109 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:15:36:23,340 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:15:36:23,340 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:15:36:23,341 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 12/1319 [00:00<00:11, 111.13it/s]
  2%|▏         | 24/1319 [00:00<00:11, 114.43it/s]
  3%|▎         | 36/1319 [00:00<00:11, 115.62it/s]
  4%|▎         | 48/1319 [00:00<00:10, 116.25it/s]
  5%|▍         | 60/1319 [00:00<00:10, 116.64it/s]
  5%|▌         | 72/1319 [00:00<00:10, 116.98it/s]
  6%|▋         | 84/1319 [00:00<00:10, 117.38it/s]
  7%|▋         | 96/1319 [00:00<00:10, 117.58it/s]
  8%|▊         | 108/1319 [00:00<00:10, 117.73it/s]
  9%|▉         | 120/1319 [00:01<00:10, 117.85it/s]
 10%|█         | 132/1319 [00:01<00:10, 117.73it/s]
 11%|█         | 144/1319 [00:01<00:09, 117.68it/s]
 12%|█▏        | 156/1319 [00:01<00:09, 117.29it/s]
 13%|█▎        | 168/1319 [00:01<00:09, 117.29it/s]
 14%|█▎        | 180/1319 [00:01<00:09, 117.65it/s]
 15%|█▍        | 192/1319 [00:01<00:09, 117.93it/s]
 15%|█▌        | 204/1319 [00:01<00:09, 118.27it/s]
 16%|█▋        | 216/1319 [00:01<00:09, 118.49it/s]
 17%|█▋        | 228/1319 [00:01<00:09, 118.61it/s]
 18%|█▊        | 240/1319 [00:02<00:09, 118.85it/s]
 19%|█▉        | 252/1319 [00:02<00:08, 118.92it/s]
 20%|██        | 264/1319 [00:02<00:08, 118.74it/s]
 21%|██        | 276/1319 [00:02<00:08, 118.53it/s]
 22%|██▏       | 288/1319 [00:02<00:08, 118.67it/s]
 23%|██▎       | 300/1319 [00:02<00:08, 118.80it/s]
 24%|██▎       | 312/1319 [00:02<00:08, 118.81it/s]
 25%|██▍       | 324/1319 [00:02<00:08, 118.81it/s]
 25%|██▌       | 336/1319 [00:02<00:08, 118.75it/s]
 26%|██▋       | 348/1319 [00:02<00:08, 118.73it/s]
 27%|██▋       | 360/1319 [00:03<00:08, 118.88it/s]
 28%|██▊       | 372/1319 [00:03<00:07, 118.77it/s]
 29%|██▉       | 384/1319 [00:03<00:07, 118.53it/s]
 30%|███       | 396/1319 [00:03<00:07, 118.58it/s]
 31%|███       | 408/1319 [00:03<00:07, 118.52it/s]
 32%|███▏      | 420/1319 [00:03<00:07, 118.64it/s]
 33%|███▎      | 432/1319 [00:03<00:07, 118.68it/s]
 34%|███▎      | 444/1319 [00:03<00:07, 118.77it/s]
 35%|███▍      | 456/1319 [00:03<00:07, 118.77it/s]
 35%|███▌      | 468/1319 [00:03<00:07, 118.79it/s]
 36%|███▋      | 480/1319 [00:04<00:07, 118.83it/s]
 37%|███▋      | 492/1319 [00:04<00:06, 118.86it/s]
 38%|███▊      | 504/1319 [00:04<00:06, 118.91it/s]
 39%|███▉      | 516/1319 [00:04<00:06, 118.82it/s]
 40%|████      | 528/1319 [00:04<00:06, 118.90it/s]
 41%|████      | 540/1319 [00:04<00:06, 119.02it/s]
 42%|████▏     | 552/1319 [00:04<00:06, 118.88it/s]
 43%|████▎     | 564/1319 [00:04<00:06, 118.91it/s]
 44%|████▎     | 576/1319 [00:04<00:06, 118.90it/s]
 45%|████▍     | 588/1319 [00:04<00:06, 118.95it/s]
 45%|████▌     | 600/1319 [00:05<00:06, 119.00it/s]
 46%|████▋     | 612/1319 [00:05<00:05, 118.99it/s]
 47%|████▋     | 624/1319 [00:05<00:05, 118.67it/s]
 48%|████▊     | 636/1319 [00:05<00:05, 118.61it/s]
 49%|████▉     | 648/1319 [00:05<00:05, 118.68it/s]
 50%|█████     | 660/1319 [00:05<00:05, 118.82it/s]
 51%|█████     | 672/1319 [00:05<00:05, 118.73it/s]
 52%|█████▏    | 684/1319 [00:05<00:05, 118.72it/s]
 53%|█████▎    | 696/1319 [00:05<00:05, 118.72it/s]
 54%|█████▎    | 708/1319 [00:05<00:05, 118.71it/s]
 55%|█████▍    | 720/1319 [00:06<00:05, 118.84it/s]
 55%|█████▌    | 732/1319 [00:06<00:04, 118.58it/s]
 56%|█████▋    | 744/1319 [00:06<00:04, 118.38it/s]
 57%|█████▋    | 756/1319 [00:06<00:04, 118.27it/s]
 58%|█████▊    | 768/1319 [00:06<00:04, 118.31it/s]
 59%|█████▉    | 780/1319 [00:06<00:04, 118.40it/s]
 60%|██████    | 792/1319 [00:06<00:04, 118.48it/s]
 61%|██████    | 804/1319 [00:06<00:04, 118.68it/s]
 62%|██████▏   | 816/1319 [00:06<00:04, 118.71it/s]
 63%|██████▎   | 828/1319 [00:06<00:04, 118.79it/s]
 64%|██████▎   | 840/1319 [00:07<00:04, 118.95it/s]
 65%|██████▍   | 852/1319 [00:07<00:03, 119.03it/s]
 66%|██████▌   | 864/1319 [00:07<00:03, 118.74it/s]
 66%|██████▋   | 876/1319 [00:07<00:03, 118.49it/s]
 67%|██████▋   | 888/1319 [00:07<00:03, 118.53it/s]
 68%|██████▊   | 900/1319 [00:07<00:03, 118.60it/s]
 69%|██████▉   | 912/1319 [00:07<00:03, 118.69it/s]
 70%|███████   | 924/1319 [00:07<00:03, 118.81it/s]
 71%|███████   | 936/1319 [00:07<00:03, 118.89it/s]
 72%|███████▏  | 948/1319 [00:08<00:03, 118.88it/s]
 73%|███████▎  | 960/1319 [00:08<00:03, 118.92it/s]
 74%|███████▎  | 972/1319 [00:08<00:02, 118.97it/s]
 75%|███████▍  | 984/1319 [00:08<00:02, 119.01it/s]
 76%|███████▌  | 996/1319 [00:08<00:02, 118.90it/s]
 76%|███████▋  | 1008/1319 [00:08<00:02, 118.90it/s]
 77%|███████▋  | 1020/1319 [00:08<00:02, 118.95it/s]
 78%|███████▊  | 1032/1319 [00:08<00:02, 118.95it/s]
 79%|███████▉  | 1044/1319 [00:08<00:02, 118.97it/s]
 80%|████████  | 1056/1319 [00:08<00:02, 118.93it/s]
 81%|████████  | 1068/1319 [00:09<00:02, 118.93it/s]
 82%|████████▏ | 1080/1319 [00:09<00:02, 119.15it/s]
 83%|████████▎ | 1093/1319 [00:09<00:01, 119.45it/s]
 84%|████████▍ | 1105/1319 [00:09<00:01, 119.40it/s]
 85%|████████▍ | 1117/1319 [00:09<00:01, 118.80it/s]
 86%|████████▌ | 1129/1319 [00:09<00:01, 118.90it/s]
 87%|████████▋ | 1141/1319 [00:09<00:01, 118.97it/s]
 87%|████████▋ | 1153/1319 [00:09<00:01, 119.08it/s]
 88%|████████▊ | 1165/1319 [00:09<00:01, 119.17it/s]
 89%|████████▉ | 1177/1319 [00:09<00:01, 119.23it/s]
 90%|█████████ | 1189/1319 [00:10<00:01, 119.16it/s]
 91%|█████████ | 1201/1319 [00:10<00:00, 119.19it/s]
 92%|█████████▏| 1213/1319 [00:10<00:00, 119.00it/s]
 93%|█████████▎| 1225/1319 [00:10<00:00, 119.07it/s]
 94%|█████████▍| 1237/1319 [00:10<00:00, 119.01it/s]
 95%|█████████▍| 1249/1319 [00:10<00:00, 119.03it/s]
 96%|█████████▌| 1261/1319 [00:10<00:00, 119.01it/s]
 97%|█████████▋| 1273/1319 [00:10<00:00, 119.07it/s]
 97%|█████████▋| 1285/1319 [00:10<00:00, 119.01it/s]
 98%|█████████▊| 1297/1319 [00:10<00:00, 118.97it/s]
 99%|█████████▉| 1309/1319 [00:11<00:00, 118.94it/s]
+2025-02-06:15:36:34,497 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:16:26:33,957 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-600), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6793|±  |0.0129|
+|               |       |strict-match    |     8|exact_match|↑  |0.6641|±  |0.0130|
+

8b_only_dpo/solo_dpo_model_ckpt_700.log

@@ -0,0 +1,25 @@
+2025-02-06:17:56:45,715 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:17:56:57,796 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:17:56:57,798 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:17:56:57,798 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:17:56:57,798 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-700'}
+2025-02-06:17:56:57,844 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:17:56:58,659 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:17:57:12,302 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:17:57:12,303 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:17:57:12,303 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 10/1319 [00:00<00:14, 91.97it/s]
  2%|▏         | 22/1319 [00:00<00:12, 102.99it/s]
  3%|▎         | 34/1319 [00:00<00:11, 108.22it/s]
  3%|▎         | 46/1319 [00:00<00:11, 110.86it/s]
  4%|▍         | 58/1319 [00:00<00:11, 112.41it/s]
  5%|▌         | 70/1319 [00:00<00:11, 109.77it/s]
  6%|▌         | 81/1319 [00:00<00:11, 107.47it/s]
  7%|▋         | 92/1319 [00:00<00:11, 106.89it/s]
  8%|▊         | 103/1319 [00:00<00:11, 107.01it/s]
  9%|▊         | 114/1319 [00:01<00:11, 107.55it/s]
  9%|▉         | 125/1319 [00:01<00:11, 106.93it/s]
 10%|█         | 136/1319 [00:01<00:11, 106.19it/s]
 11%|█         | 147/1319 [00:01<00:11, 106.21it/s]
 12%|█▏        | 158/1319 [00:01<00:10, 106.11it/s]
 13%|█▎        | 169/1319 [00:01<00:10, 105.51it/s]
 14%|█▎        | 181/1319 [00:01<00:10, 107.06it/s]
 15%|█▍        | 193/1319 [00:01<00:10, 108.41it/s]
 16%|█▌        | 205/1319 [00:01<00:10, 109.62it/s]
 16%|█▋        | 217/1319 [00:02<00:09, 110.99it/s]
 17%|█▋        | 229/1319 [00:02<00:09, 112.23it/s]
 18%|█▊        | 241/1319 [00:02<00:09, 113.21it/s]
 19%|█▉        | 253/1319 [00:02<00:09, 114.15it/s]
 20%|██        | 265/1319 [00:02<00:09, 114.69it/s]
 21%|██        | 277/1319 [00:02<00:09, 115.43it/s]
 22%|██▏       | 289/1319 [00:02<00:09, 114.18it/s]
 23%|██▎       | 301/1319 [00:02<00:08, 114.16it/s]
 24%|██▎       | 313/1319 [00:02<00:08, 114.40it/s]
 25%|██▍       | 325/1319 [00:02<00:08, 115.03it/s]
 26%|██▌       | 337/1319 [00:03<00:08, 115.37it/s]
 26%|██▋       | 349/1319 [00:03<00:08, 115.79it/s]
 27%|██▋       | 361/1319 [00:03<00:08, 115.93it/s]
 28%|██▊       | 373/1319 [00:03<00:08, 116.25it/s]
 29%|██▉       | 385/1319 [00:03<00:08, 116.04it/s]
 30%|███       | 397/1319 [00:03<00:07, 116.34it/s]
 31%|███       | 409/1319 [00:03<00:07, 116.03it/s]
 32%|███▏      | 421/1319 [00:03<00:07, 116.13it/s]
 33%|███▎      | 433/1319 [00:03<00:07, 116.21it/s]
 34%|███▎      | 445/1319 [00:03<00:07, 116.42it/s]
 35%|███▍      | 457/1319 [00:04<00:07, 115.93it/s]
 36%|███▌      | 469/1319 [00:04<00:07, 116.25it/s]
 36%|███▋      | 481/1319 [00:04<00:07, 116.63it/s]
 37%|███▋      | 493/1319 [00:04<00:07, 116.78it/s]
 38%|███▊      | 505/1319 [00:04<00:06, 116.89it/s]
 39%|███▉      | 517/1319 [00:04<00:06, 116.87it/s]
 40%|████      | 529/1319 [00:04<00:06, 116.90it/s]
 41%|████      | 541/1319 [00:04<00:06, 116.99it/s]
 42%|████▏     | 553/1319 [00:04<00:06, 116.32it/s]
 43%|████▎     | 565/1319 [00:05<00:06, 116.26it/s]
 44%|████▎     | 577/1319 [00:05<00:06, 116.66it/s]
 45%|████▍     | 589/1319 [00:05<00:06, 116.57it/s]
 46%|████▌     | 601/1319 [00:05<00:06, 116.89it/s]
 46%|████▋     | 613/1319 [00:05<00:06, 116.84it/s]
 47%|████▋     | 625/1319 [00:05<00:05, 116.90it/s]
 48%|████▊     | 637/1319 [00:05<00:05, 116.98it/s]
 49%|████▉     | 649/1319 [00:05<00:05, 117.13it/s]
 50%|█████     | 661/1319 [00:05<00:05, 117.13it/s]
 51%|█████     | 673/1319 [00:05<00:05, 117.19it/s]
 52%|█████▏    | 685/1319 [00:06<00:05, 117.16it/s]
 53%|█████▎    | 697/1319 [00:06<00:05, 116.60it/s]
 54%|█████▍    | 709/1319 [00:06<00:05, 116.31it/s]
 55%|█████▍    | 721/1319 [00:06<00:05, 116.38it/s]
 56%|█████▌    | 733/1319 [00:06<00:05, 116.22it/s]
 56%|█████▋    | 745/1319 [00:06<00:04, 116.25it/s]
 57%|█████▋    | 757/1319 [00:06<00:04, 116.34it/s]
 58%|█████▊    | 769/1319 [00:06<00:04, 116.29it/s]
 59%|█████▉    | 781/1319 [00:06<00:04, 116.68it/s]
 60%|██████    | 793/1319 [00:06<00:04, 116.90it/s]
 61%|██████    | 805/1319 [00:07<00:04, 117.16it/s]
 62%|██████▏   | 817/1319 [00:07<00:04, 117.26it/s]
 63%|██████▎   | 829/1319 [00:07<00:04, 117.39it/s]
 64%|██████▍   | 841/1319 [00:07<00:04, 117.48it/s]
 65%|██████▍   | 853/1319 [00:07<00:03, 117.52it/s]
 66%|██████▌   | 865/1319 [00:07<00:03, 117.52it/s]
 66%|██████▋   | 877/1319 [00:07<00:03, 115.01it/s]
 67%|██████▋   | 889/1319 [00:07<00:03, 115.37it/s]
 68%|██████▊   | 901/1319 [00:07<00:03, 115.68it/s]
 69%|██████▉   | 913/1319 [00:07<00:03, 115.97it/s]
 70%|███████   | 925/1319 [00:08<00:03, 116.02it/s]
 71%|███████   | 937/1319 [00:08<00:03, 115.85it/s]
 72%|███████▏  | 949/1319 [00:08<00:03, 115.68it/s]
 73%|███████▎  | 961/1319 [00:08<00:03, 115.79it/s]
 74%|███████▍  | 973/1319 [00:08<00:02, 116.11it/s]
 75%|███████▍  | 985/1319 [00:08<00:02, 116.35it/s]
 76%|███████▌  | 997/1319 [00:08<00:02, 116.31it/s]
 76%|███████▋  | 1009/1319 [00:08<00:02, 116.70it/s]
 77%|███████▋  | 1021/1319 [00:08<00:02, 116.97it/s]
 78%|███████▊  | 1033/1319 [00:09<00:02, 117.26it/s]
 79%|███████▉  | 1045/1319 [00:09<00:02, 117.37it/s]
 80%|████████  | 1057/1319 [00:09<00:02, 117.52it/s]
 81%|████████  | 1069/1319 [00:09<00:02, 117.59it/s]
 82%|████████▏ | 1081/1319 [00:09<00:02, 117.56it/s]
 83%|████████▎ | 1093/1319 [00:09<00:01, 117.57it/s]
 84%|████████▍ | 1105/1319 [00:09<00:01, 117.54it/s]
 85%|████████▍ | 1117/1319 [00:09<00:01, 117.52it/s]
 86%|████████▌ | 1129/1319 [00:09<00:01, 117.40it/s]
 87%|████████▋ | 1141/1319 [00:09<00:01, 117.49it/s]
 87%|████████▋ | 1153/1319 [00:10<00:01, 117.48it/s]
 88%|████████▊ | 1165/1319 [00:10<00:01, 117.51it/s]
 89%|████████▉ | 1177/1319 [00:10<00:01, 116.79it/s]
 90%|█████████ | 1189/1319 [00:10<00:01, 116.81it/s]
 91%|█████████ | 1201/1319 [00:10<00:01, 116.50it/s]
 92%|█████████▏| 1213/1319 [00:10<00:00, 116.71it/s]
 93%|█████████▎| 1225/1319 [00:10<00:00, 116.61it/s]
 94%|█████████▍| 1237/1319 [00:10<00:00, 116.89it/s]
 95%|█████████▍| 1249/1319 [00:10<00:00, 116.94it/s]
 96%|█████████▌| 1261/1319 [00:10<00:00, 117.11it/s]
 97%|█████████▋| 1273/1319 [00:11<00:00, 117.25it/s]
 97%|█████████▋| 1285/1319 [00:11<00:00, 117.38it/s]
 98%|█████████▊| 1297/1319 [00:11<00:00, 117.49it/s]
 99%|█████████▉| 1309/1319 [00:11<00:00, 117.54it/s]
+2025-02-06:17:57:23,800 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+2025-02-06:18:46:30,931 INFO     [evaluation_tracker.py:269] Output path not provided, skipping saving results aggregated
+hf (pretrained=meta-llama/Meta-Llama-3-8B-Instruct,peft=JaqueMate/solo_dpo_checkpoint-700), gen_kwargs: (do_sample=False,temperature=0.01,top_k=0,top_p=1), limit: None, num_fewshot: 8, batch_size: 6
+|     Tasks     |Version|     Filter     |n-shot|  Metric   |   |Value |   |Stderr|
+|---------------|------:|----------------|-----:|-----------|---|-----:|---|-----:|
+|gsm8k_cot_llama|      3|flexible-extract|     8|exact_match|↑  |0.6937|±  |0.0127|
+|               |       |strict-match    |     8|exact_match|↑  |0.6785|±  |0.0129|
+

8b_only_dpo/solo_dpo_model_ckpt_800.log

@@ -0,0 +1,80 @@
+2025-02-06:18:46:34,826 INFO     [__main__.py:284] Verbosity set to INFO
+2025-02-06:18:46:45,992 INFO     [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama']
+2025-02-06:18:46:45,993 INFO     [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234
+2025-02-06:18:46:45,993 WARNING  [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding!
+2025-02-06:18:46:45,993 INFO     [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/solo_dpo_checkpoint-800'}
+2025-02-06:18:46:46,033 INFO     [huggingface.py:135] Using device 'cuda:0'
+2025-02-06:18:46:46,668 INFO     [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'}
+
+2025-02-06:18:47:29,137 WARNING  [evaluator.py:271] Overwriting default num_fewshot of gsm8k_cot_llama from 8 to 8
+2025-02-06:18:47:29,138 WARNING  [evaluator.py:417] Chat template formatting change affects loglikelihood and multiple-choice tasks. See docs/chat-template-readme.md for details.
+2025-02-06:18:47:29,138 INFO     [task.py:420] Building contexts for gsm8k_cot_llama on rank 0...
+
  0%|          | 0/1319 [00:00<?, ?it/s]
  1%|          | 11/1319 [00:00<00:11, 109.53it/s]
  2%|▏         | 23/1319 [00:00<00:11, 113.16it/s]
  3%|▎         | 35/1319 [00:00<00:11, 112.65it/s]
  4%|▎         | 47/1319 [00:00<00:11, 112.79it/s]
  4%|▍         | 59/1319 [00:00<00:11, 113.29it/s]
  5%|▌         | 71/1319 [00:00<00:10, 114.75it/s]
  6%|▋         | 83/1319 [00:00<00:10, 115.74it/s]
  7%|▋         | 95/1319 [00:00<00:10, 116.47it/s]
  8%|▊         | 107/1319 [00:00<00:10, 116.73it/s]
  9%|▉         | 119/1319 [00:01<00:10, 116.52it/s]
 10%|▉         | 131/1319 [00:01<00:10, 116.75it/s]
 11%|█         | 143/1319 [00:01<00:10, 117.05it/s]
 12%|█▏        | 155/1319 [00:01<00:09, 117.40it/s]
 13%|█▎        | 167/1319 [00:01<00:09, 117.70it/s]
 14%|█▎        | 179/1319 [00:01<00:09, 117.78it/s]
 14%|█▍        | 191/1319 [00:01<00:09, 117.93it/s]
 15%|█▌        | 203/1319 [00:01<00:09, 118.06it/s]
 16%|█▋        | 215/1319 [00:01<00:09, 117.93it/s]
 17%|█▋        | 227/1319 [00:01<00:09, 117.41it/s]
 18%|█▊        | 239/1319 [00:02<00:09, 117.42it/s]
 19%|█▉        | 251/1319 [00:02<00:09, 117.55it/s]
 20%|█▉        | 263/1319 [00:02<00:08, 117.78it/s]
 21%|██        | 275/1319 [00:02<00:08, 117.99it/s]
 22%|██▏       | 287/1319 [00:02<00:08, 118.16it/s]
 23%|██▎       | 299/1319 [00:02<00:08, 118.23it/s]
 24%|██▎       | 311/1319 [00:02<00:08, 118.35it/s]
 24%|██▍       | 323/1319 [00:02<00:08, 118.35it/s]
 25%|██▌       | 335/1319 [00:02<00:08, 118.38it/s]
 26%|██▋       | 347/1319 [00:02<00:08, 118.38it/s]
 27%|██▋       | 359/1319 [00:03<00:08, 118.39it/s]
 28%|██▊       | 371/1319 [00:03<00:07, 118.52it/s]
 29%|██▉       | 383/1319 [00:03<00:07, 118.56it/s]
 30%|██▉       | 395/1319 [00:03<00:07, 118.55it/s]
 31%|███       | 407/1319 [00:03<00:07, 118.54it/s]
 32%|███▏      | 419/1319 [00:03<00:07, 118.51it/s]
 33%|███▎      | 431/1319 [00:03<00:07, 118.57it/s]
 34%|███▎      | 443/1319 [00:03<00:07, 118.54it/s]
 34%|███▍      | 455/1319 [00:03<00:07, 118.46it/s]
 35%|███▌      | 467/1319 [00:03<00:07, 117.95it/s]
 36%|███▋      | 479/1319 [00:04<00:07, 118.04it/s]
 37%|███▋      | 491/1319 [00:04<00:07, 117.86it/s]
 38%|███▊      | 503/1319 [00:04<00:06, 118.10it/s]
 39%|███▉      | 515/1319 [00:04<00:06, 118.24it/s]
 40%|███▉      | 527/1319 [00:04<00:06, 118.27it/s]
 41%|████      | 539/1319 [00:04<00:06, 118.31it/s]
 42%|████▏     | 551/1319 [00:04<00:06, 118.37it/s]
 43%|████▎     | 563/1319 [00:04<00:06, 118.37it/s]
 44%|████▎     | 575/1319 [00:04<00:06, 118.32it/s]
 45%|████▍     | 587/1319 [00:04<00:06, 118.37it/s]
 45%|████▌     | 599/1319 [00:05<00:06, 118.39it/s]
 46%|████▋     | 611/1319 [00:05<00:05, 118.41it/s]
 47%|████▋     | 623/1319 [00:05<00:05, 118.39it/s]
 48%|████▊     | 635/1319 [00:05<00:05, 118.40it/s]
 49%|████▉     | 647/1319 [00:05<00:05, 118.24it/s]
 50%|████▉     | 659/1319 [00:05<00:05, 118.27it/s]
 51%|█████     | 671/1319 [00:05<00:05, 118.30it/s]
 52%|█████▏    | 683/1319 [00:05<00:05, 118.31it/s]
 53%|█████▎    | 695/1319 [00:05<00:05, 118.25it/s]
 54%|█████▎    | 707/1319 [00:06<00:05, 116.37it/s]
 55%|█████▍    | 719/1319 [00:06<00:05, 114.71it/s]
 55%|█████▌    | 731/1319 [00:06<00:05, 114.82it/s]
 56%|█████▋    | 743/1319 [00:06<00:04, 115.50it/s]
 57%|█████▋    | 755/1319 [00:06<00:04, 116.31it/s]
 58%|█████▊    | 767/1319 [00:06<00:04, 116.79it/s]
 59%|█████▉    | 779/1319 [00:06<00:04, 117.28it/s]
 60%|█████▉    | 791/1319 [00:06<00:04, 117.59it/s]
 61%|██████    | 803/1319 [00:06<00:04, 117.88it/s]
 62%|██████▏   | 815/1319 [00:06<00:04, 118.03it/s]
 63%|██████▎   | 827/1319 [00:07<00:04, 117.53it/s]
 64%|██████▎   | 839/1319 [00:07<00:04, 117.76it/s]
 65%|██████▍   | 851/1319 [00:07<00:03, 117.97it/s]
 65%|██████▌   | 863/1319 [00:07<00:03, 118.12it/s]
 66%|██████▋   | 875/1319 [00:07<00:03, 118.22it/s]
 67%|██████▋   | 887/1319 [00:07<00:03, 118.06it/s]
 68%|██████▊   | 899/1319 [00:07<00:03, 118.18it/s]
 69%|██████▉   | 911/1319 [00:07<00:03, 117.42it/s]
 70%|██████▉   | 923/1319 [00:07<00:03, 117.54it/s]
 71%|███████   | 935/1319 [00:07<00:03, 117.68it/s]
 72%|███████▏  | 947/1319 [00:08<00:03, 116.82it/s]
 73%|███████▎  | 959/1319 [00:08<00:03, 116.37it/s]
 74%|███████▎  | 971/1319 [00:08<00:02, 116.97it/s]
 75%|███████▍  | 983/1319 [00:08<00:02, 117.37it/s]
 75%|███████▌  | 995/1319 [00:08<00:02, 117.68it/s]
 76%|███████▋  | 1007/1319 [00:08<00:02, 118.01it/s]
 77%|███████▋  | 1019/1319 [00:08<00:02, 118.19it/s]
 78%|███████▊  | 1031/1319 [00:08<00:02, 118.35it/s]
 79%|███████▉  | 1043/1319 [00:08<00:02, 118.47it/s]
 80%|███████▉  | 1055/1319 [00:08<00:02, 118.44it/s]
 81%|████████  | 1067/1319 [00:09<00:02, 118.53it/s]
 82%|████████▏ | 1079/1319 [00:09<00:02, 118.48it/s]
 83%|████████▎ | 1091/1319 [00:09<00:01, 118.54it/s]
 84%|████████▎ | 1103/1319 [00:09<00:01, 118.54it/s]
 85%|████████▍ | 1115/1319 [00:09<00:01, 117.84it/s]
 86%|████████▌ | 1128/1319 [00:09<00:01, 119.49it/s]
 87%|████████▋ | 1141/1319 [00:09<00:01, 120.74it/s]
 87%|████████▋ | 1154/1319 [00:09<00:01, 121.33it/s]
 88%|████████▊ | 1167/1319 [00:09<00:01, 122.01it/s]
 89%|████████▉ | 1180/1319 [00:10<00:01, 122.26it/s]
 90%|█████████ | 1193/1319 [00:10<00:01, 121.57it/s]
 91%|█████████▏| 1206/1319 [00:10<00:00, 121.98it/s]
 92%|█████████▏| 1219/1319 [00:10<00:00, 122.40it/s]
 93%|█████████▎| 1232/1319 [00:10<00:00, 122.78it/s]
 94%|█████████▍| 1245/1319 [00:10<00:00, 122.55it/s]
 95%|█████████▌| 1258/1319 [00:10<00:00, 122.83it/s]
 96%|█████████▋| 1271/1319 [00:10<00:00, 123.10it/s]
 97%|█████████▋| 1284/1319 [00:10<00:00, 123.26it/s]
 98%|█████████▊| 1297/1319 [00:10<00:00, 122.59it/s]
 99%|█████████▉| 1310/1319 [00:11<00:00, 122.30it/s]
+2025-02-06:18:47:40,316 INFO     [evaluator.py:513] Running generate_until requests
+
+  warnings.warn(
+/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/configuration_utils.py:589: UserWarning: `do_sample` is set to `False`. However, `top_k` is set to `0` -- this flag is only used in sample-based generation modes. You should set `do_sample=True` or unset `top_k`.
+  warnings.warn(
+
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 467, in <module>
+    cli_evaluate()
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate
+    results = evaluator.simple_evaluate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper
+    return fn(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 304, in simple_evaluate
+    results = evaluate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper
+    return fn(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 524, in evaluate
+    resps = getattr(lm, reqtype)(cloned_reqs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 1354, in generate_until
+    cont = self._model_generate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 897, in _model_generate
+    return self.model.generate(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1646, in generate
+    outputs = self.base_model.generate(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
+    return func(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 1996, in generate
+    result = self._sample(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 2923, in _sample
+    outputs = self(**model_inputs, return_dict=True)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 1182, in forward
+    outputs = self.model(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 994, in forward
+    layer_outputs = decoder_layer(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 727, in forward
+    hidden_states, self_attn_weights, present_key_value = self.self_attn(
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 665, in forward
+    attn_output = self.o_proj(attn_output)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/tuners/lora/layer.py", line 556, in forward
+    result = result + lora_B(lora_A(dropout(x))) * scaling
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
+    return self._call_impl(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
+    return forward_call(*args, **kwargs)
+  File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 116, in forward
+    return F.linear(input, self.weight, self.bias)
+KeyboardInterrupt
+

8b_only_dpo/solo_dpo_model_ckpt_900.log

@@ -0,0 +1 @@
+one_last_bench.sh: line 35: --device: command not found