[pair] 80 eval problems | BASE=/workspace/RSI/newbase ADAPTER=/workspace/RSI/outputs/af/teach_adapter_128
[transformers] `torch_dtype` is deprecated! Use `dtype` instead!

Loading weights:   0%|          | 0/771 [00:00<?, ?it/s]
Loading weights:   0%|          | 1/771 [00:00<04:25,  2.90it/s]
Loading weights:   0%|          | 2/771 [00:00<04:22,  2.93it/s]
Loading weights:   1%|          | 4/771 [00:00<02:10,  5.90it/s]
Loading weights:   1%|          | 9/771 [00:00<00:50, 15.06it/s]
Loading weights:   2%|▏         | 17/771 [00:01<00:26, 28.57it/s]
Loading weights:   4%|▎         | 28/771 [00:01<00:16, 45.16it/s]
Loading weights:   4%|▍         | 34/771 [00:01<00:15, 47.14it/s]
Loading weights:   5%|▌         | 41/771 [00:01<00:14, 50.91it/s]
Loading weights:   7%|▋         | 52/771 [00:01<00:11, 62.00it/s]
Loading weights:   8%|▊         | 59/771 [00:01<00:11, 61.99it/s]
Loading weights:   9%|▊         | 66/771 [00:01<00:12, 55.19it/s]
Loading weights:  10%|▉         | 77/771 [00:01<00:10, 64.65it/s]
Loading weights:  11%|█▏        | 88/771 [00:02<00:09, 71.02it/s]
Loading weights:  12%|█▏        | 96/771 [00:02<00:09, 69.21it/s]
Loading weights:  13%|█▎        | 104/771 [00:02<00:10, 62.15it/s]
Loading weights:  15%|█▍        | 113/771 [00:02<00:10, 63.72it/s]
Loading weights:  16%|█▌        | 124/771 [00:02<00:09, 68.80it/s]
Loading weights:  17%|█▋        | 131/771 [00:02<00:09, 67.85it/s]
Loading weights:  18%|█▊        | 138/771 [00:02<00:10, 58.80it/s]
Loading weights:  19%|█▉        | 149/771 [00:03<00:09, 65.66it/s]
Loading weights:  21%|██        | 160/771 [00:03<00:08, 70.88it/s]
Loading weights:  22%|██▏       | 168/771 [00:03<00:08, 68.25it/s]
Loading weights:  23%|██▎       | 175/771 [00:03<00:09, 59.82it/s]
Loading weights:  24%|██▍       | 185/771 [00:03<00:09, 64.44it/s]
Loading weights:  25%|██▌       | 196/771 [00:03<00:08, 70.03it/s]
Loading weights:  26%|██▋       | 204/771 [00:03<00:08, 69.28it/s]
Loading weights:  27%|██▋       | 212/771 [00:03<00:08, 64.02it/s]
Loading weights:  29%|██▊       | 221/771 [00:04<00:08, 66.63it/s]
Loading weights:  30%|███       | 232/771 [00:04<00:07, 73.03it/s]
Loading weights:  31%|███       | 240/771 [00:04<00:07, 71.34it/s]
Loading weights:  32%|███▏      | 248/771 [00:04<00:08, 64.98it/s]
Loading weights:  33%|███▎      | 257/771 [00:04<00:07, 67.76it/s]
Loading weights:  35%|███▍      | 268/771 [00:04<00:06, 74.23it/s]
Loading weights:  36%|███▌      | 276/771 [00:04<00:06, 72.51it/s]
Loading weights:  37%|███▋      | 284/771 [00:04<00:07, 66.54it/s]
Loading weights:  38%|███▊      | 293/771 [00:05<00:06, 68.69it/s]
Loading weights:  39%|███▉      | 304/771 [00:05<00:06, 75.11it/s]
Loading weights:  40%|████      | 312/771 [00:05<00:06, 73.01it/s]
Loading weights:  42%|████▏     | 320/771 [00:05<00:06, 66.84it/s]
Loading weights:  43%|████▎     | 329/771 [00:05<00:06, 69.10it/s]
Loading weights:  44%|████▍     | 340/771 [00:05<00:05, 76.28it/s]
Loading weights:  45%|████▌     | 348/771 [00:05<00:05, 74.97it/s]
Loading weights:  46%|████▌     | 356/771 [00:05<00:05, 69.19it/s]
Loading weights:  47%|████▋     | 365/771 [00:06<00:05, 72.08it/s]
Loading weights:  49%|████▉     | 376/771 [00:06<00:05, 78.68it/s]
Loading weights:  50%|████▉     | 384/771 [00:06<00:05, 76.66it/s]
Loading weights:  51%|█████     | 392/771 [00:06<00:05, 70.17it/s]
Loading weights:  52%|█████▏    | 401/771 [00:06<00:05, 72.62it/s]
Loading weights:  53%|█████▎    | 412/771 [00:06<00:04, 79.07it/s]
Loading weights:  54%|█████▍    | 420/771 [00:06<00:04, 77.39it/s]
Loading weights:  56%|█████▌    | 428/771 [00:06<00:04, 71.55it/s]
Loading weights:  57%|█████▋    | 437/771 [00:07<00:04, 72.16it/s]
Loading weights:  58%|█████▊    | 448/771 [00:07<00:04, 77.80it/s]
Loading weights:  59%|█████▉    | 456/771 [00:07<00:04, 72.22it/s]
Loading weights:  60%|██████    | 464/771 [00:07<00:04, 65.20it/s]
Loading weights:  61%|██████▏   | 473/771 [00:07<00:04, 67.69it/s]
Loading weights:  63%|██████▎   | 484/771 [00:07<00:04, 70.81it/s]
Loading weights:  64%|██████▍   | 492/771 [00:07<00:04, 69.55it/s]
Loading weights:  65%|██████▍   | 499/771 [00:08<00:04, 57.93it/s]
Loading weights:  66%|██████▌   | 509/771 [00:08<00:04, 59.47it/s]
Loading weights:  67%|██████▋   | 520/771 [00:08<00:03, 64.02it/s]
Loading weights:  68%|██████▊   | 527/771 [00:08<00:03, 61.47it/s]
Loading weights:  69%|██████▉   | 534/771 [00:08<00:04, 55.58it/s]
Loading weights:  71%|███████   | 545/771 [00:08<00:03, 64.25it/s]
Loading weights:  72%|███████▏  | 552/771 [00:08<00:03, 64.06it/s]
Loading weights:  73%|███████▎  | 559/771 [00:09<00:03, 58.20it/s]
Loading weights:  74%|███████▍  | 569/771 [00:09<00:03, 59.93it/s]
Loading weights:  75%|███████▌  | 580/771 [00:09<00:02, 64.89it/s]
Loading weights:  76%|███████▌  | 587/771 [00:09<00:02, 61.71it/s]
Loading weights:  77%|███████▋  | 594/771 [00:09<00:03, 51.28it/s]
Loading weights:  78%|███████▊  | 605/771 [00:09<00:02, 56.88it/s]
Loading weights:  80%|███████▉  | 616/771 [00:09<00:02, 60.47it/s]
Loading weights:  81%|████████  | 623/771 [00:10<00:02, 57.38it/s]
Loading weights:  82%|████████▏ | 629/771 [00:10<00:02, 53.33it/s]
Loading weights:  83%|████████▎ | 640/771 [00:10<00:02, 60.13it/s]
Loading weights:  84%|████████▍ | 647/771 [00:10<00:02, 55.59it/s]
Loading weights:  85%|████████▍ | 653/771 [00:10<00:02, 51.68it/s]
Loading weights:  86%|████████▌ | 664/771 [00:10<00:01, 57.02it/s]
Loading weights:  87%|████████▋ | 670/771 [00:10<00:01, 52.60it/s]
Loading weights:  88%|████████▊ | 677/771 [00:11<00:01, 51.59it/s]
Loading weights:  89%|████████▉ | 688/771 [00:11<00:01, 57.33it/s]
Loading weights:  90%|█████████ | 694/771 [00:11<00:01, 52.42it/s]
Loading weights:  91%|█████████ | 701/771 [00:11<00:01, 51.08it/s]
Loading weights:  92%|█████████▏| 712/771 [00:11<00:01, 53.94it/s]
Loading weights:  93%|█████████▎| 718/771 [00:11<00:01, 50.60it/s]
Loading weights:  94%|█████████▍| 725/771 [00:12<00:00, 50.28it/s]
Loading weights:  95%|█████████▌| 736/771 [00:12<00:00, 55.64it/s]
Loading weights:  96%|█████████▌| 742/771 [00:12<00:00, 51.28it/s]
Loading weights:  97%|█████████▋| 749/771 [00:12<00:00, 49.50it/s]
Loading weights:  98%|█████████▊| 758/771 [00:12<00:00, 58.28it/s]
Loading weights:  99%|█████████▉| 765/771 [00:12<00:00, 49.05it/s]
Loading weights: 100%|██████████| 771/771 [00:12<00:00, 59.98it/s]
[pair] loaded 4bit HF base + PEFT adapter OK
=== BASE arm (adapter DISABLED) ===
[pair]   8/80
[pair]   16/80
[pair]   24/80
[pair]   32/80
[pair]   40/80
[pair]   48/80
[pair]   56/80
[pair]   64/80
[pair]   72/80
[pair]   80/80
HOLDOUT[base] = 7/80 mean_partial=0.2718
=== TEACH arm (adapter ENABLED) ===
[pair]   8/80
[pair]   16/80
[pair]   24/80
[pair]   32/80
[pair]   40/80
[pair]   48/80
[pair]   56/80
[pair]   64/80
[pair]   72/80
[pair]   80/80
HOLDOUT[teach] = 6/80 mean_partial=0.3971
TEACHER VERDICT: DEAD | base=7/80 teach=6/80 D=-1 gained=3 lost=4 McNemar_z=-0.38 | partial base_mp=0.2718 teach_mp=0.3971 Dmp=+0.1253 improved=42 worsened=16 sign_z=3.41 | backend=hf-4bit SAME both arms
PAIR_DONE