challenge/expA__software__arc_challenge_seed42/train.log

[2026-06-10 21:41:13,768][__main__][INFO] - Config:
model:
  model_id: allenai/OLMo-3-1025-7B
  torch_dtype: bfloat16
  attn_implementation: sdpa
  lora:
    r: 8
    target_modules:
    - q_proj
    - k_proj
    - v_proj
    - o_proj
    lora_alpha: 16
    lora_dropout: 0.05
    bias: none
    task_type: CAUSAL_LM
trainer:
  max_steps: 5000
  per_device_train_batch_size: 4
  gradient_accumulation_steps: 4
  learning_rate: 1.0e-05
  lr_scheduler_type: constant
  warmup_steps: 0
  weight_decay: 0.0
  adam_beta1: 0.9
  adam_beta2: 0.999
  adam_epsilon: 1.0e-08
  max_grad_norm: 1.0
  auto_lr: true
  lr_delta: 1.0e-05
  max_walltime_minutes: 340
  logging_steps: 50
  save_steps: 200
  eval_steps: 1000
  report_to:
  - wandb
  bf16: true
  fp16: false
  gradient_checkpointing: true
  length_normalized_loss: true
  ppl_shuffle_mode: segments
  ppl_n_factor: 10
  cooldown_enabled: false
  cooldown_every_k: 200
  cooldown_steps: 50
data:
  max_length: 2048
  batch_size: 4
  num_workers: 0
  max_forget_docs: 2000
  max_retain_docs: 9000
  docs_per_retain_bin: null
  min_tokens: 0
  resample_interval: 0
  seed: 42
  forget_manifest_path: null
topic_bin: software
topic_bins: null
output_dir: /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/challenge/expA__software__arc_challenge_seed42
retain_topics: null
seed: 42
wandb_project: data-attribution-unlearn-multiseed
wandb_entity: hcai-lab
forget_texts_file: /storage/ice-shared/cs7634/staff/TDA/arc_rescore/forget_texts_binlevel_challenge/expA__software__arc_challenge.parquet

[2026-06-10 21:41:13,776][__main__][INFO] - Target topic(s): software
[2026-06-10 21:41:14,018][unlearning.data.dolma_pool][INFO] - Loading 6T-filtered Arrow cache from /storage/ice-shared/cs7634/staff/TDA/arc_rescore/dolma_cache/dolma3_6t_filtered ...
[2026-06-10 21:41:16,509][unlearning.data.dolma_pool][INFO] - Faithful forget set from /storage/ice-shared/cs7634/staff/TDA/arc_rescore/forget_texts_binlevel_challenge/expA__software__arc_challenge.parquet: 200 texts
[2026-06-10 21:41:38,123][unlearning.data.dolma_pool][INFO] - Sampled (text-file): 200 forget, 9000 retain.
[2026-06-10 21:41:38,132][unlearning.data.sampling][INFO] - Sampling manifest saved to /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/challenge/expA__software__arc_challenge_seed42/sampling_manifest.json
[2026-06-10 21:41:38,132][unlearning.data.dolma_pool][INFO] - Tokenizing 200 forget + 9000 retain docs ...
[2026-06-10 21:42:18,236][__main__][INFO] - Starting fresh training in: /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/challenge/expA__software__arc_challenge_seed42
[2026-06-10 21:42:18,237][__main__][INFO] - Loading model allenai/OLMo-3-1025-7B ...
[2026-06-10 21:42:18,450][accelerate.utils.modeling][INFO] - We will use 90% of the memory on device 0 for storing the model, and 10% for the buffer to avoid OOM. You can set `max_memory` in to a higher value to use more memory (at your own risk).
[2026-06-10 21:42:25,938][__main__][INFO] - Loading MMLU subset (100 questions) for safety guard ...
[2026-06-10 21:42:25,976][datasets.load][WARNING] - Using the latest cached version of the dataset since cais/mmlu couldn't be found on the Hugging Face Hub (offline mode is enabled).
[2026-06-10 21:42:25,995][datasets.packaged_modules.cache.cache][WARNING] - Found the latest cached dataset configuration 'all' at /home/hice1/gmatlin3/scratch/hf_cache/datasets/cais___mmlu/all/0.0.0/c30699e8356da336a370243923dbaf21066bb9fe (last modified on Mon May 25 20:02:16 2026).
[2026-06-10 21:42:30,807][__main__][INFO] - Original model MMLU accuracy (n=100): 0.5600
[2026-06-10 21:42:32,729][__main__][INFO] - Base model shuffled-token PPL (forget set, mode=segments): 17.1667
[2026-06-10 21:42:36,853][__main__][INFO] - Starting NGDiff unlearning for topic(s) 'software'...
[2026-06-10 21:42:36,853][__main__][INFO] - MMLU safety baseline: 0.5600  (threshold: 0.5040)
[2026-06-10 21:44:33,868][unlearning.trainer.ngdiff][INFO] - step=50  opt_step=12  f_loss=2.8350  r_loss=2.4487  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:46:31,419][unlearning.trainer.ngdiff][INFO] - step=100  opt_step=25  f_loss=2.4062  r_loss=2.8899  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:48:25,962][unlearning.trainer.ngdiff][INFO] - step=150  opt_step=37  f_loss=1.8634  r_loss=2.3035  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:50:23,041][unlearning.trainer.ngdiff][INFO] - step=200  opt_step=50  f_loss=1.9621  r_loss=2.4058  mmlu=0.5700  lr=6.61e-07
[2026-06-10 21:50:32,061][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=50: forget=10.5665  retain=10.3768
[2026-06-10 21:52:27,015][unlearning.trainer.ngdiff][INFO] - step=250  opt_step=62  f_loss=1.7709  r_loss=1.8533  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:54:18,188][unlearning.trainer.ngdiff][INFO] - step=300  opt_step=75  f_loss=2.5034  r_loss=1.7530  mmlu=0.5700  lr=1.00e-05
[2026-06-10 21:56:13,709][unlearning.trainer.ngdiff][INFO] - step=350  opt_step=87  f_loss=2.8397  r_loss=2.3206  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:58:07,853][unlearning.trainer.ngdiff][INFO] - step=400  opt_step=100  f_loss=2.6654  r_loss=2.0676  mmlu=0.5700  lr=1.00e-07
[2026-06-10 21:58:15,210][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=100: forget=10.5694  retain=10.3610
[2026-06-10 22:00:08,942][unlearning.trainer.ngdiff][INFO] - step=450  opt_step=112  f_loss=2.0187  r_loss=1.7854  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:02:00,934][unlearning.trainer.ngdiff][INFO] - step=500  opt_step=125  f_loss=2.0793  r_loss=2.1325  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:03:55,083][unlearning.trainer.ngdiff][INFO] - step=550  opt_step=137  f_loss=1.8930  r_loss=1.9440  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:05:46,382][unlearning.trainer.ngdiff][INFO] - step=600  opt_step=150  f_loss=3.1260  r_loss=1.7160  mmlu=0.5600  lr=8.31e-07
[2026-06-10 22:05:53,732][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=150: forget=10.5788  retain=10.3399
[2026-06-10 22:07:45,477][unlearning.trainer.ngdiff][INFO] - step=650  opt_step=162  f_loss=1.8033  r_loss=2.0780  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:09:40,842][unlearning.trainer.ngdiff][INFO] - step=700  opt_step=175  f_loss=1.9402  r_loss=1.4776  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:11:35,647][unlearning.trainer.ngdiff][INFO] - step=750  opt_step=187  f_loss=1.7841  r_loss=3.0648  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:13:30,847][unlearning.trainer.ngdiff][INFO] - step=800  opt_step=200  f_loss=2.2184  r_loss=2.9455  mmlu=0.5600  lr=2.40e-06
[2026-06-10 22:13:38,184][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=200: forget=10.5934  retain=10.2901
[2026-06-10 22:15:25,237][unlearning.trainer.ngdiff][INFO] - step=850  opt_step=212  f_loss=1.9232  r_loss=3.0240  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:17:19,419][unlearning.trainer.ngdiff][INFO] - step=900  opt_step=225  f_loss=2.0606  r_loss=1.6135  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:19:11,794][unlearning.trainer.ngdiff][INFO] - step=950  opt_step=237  f_loss=2.3066  r_loss=2.0351  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:21:04,896][unlearning.trainer.ngdiff][INFO] - step=1000  opt_step=250  f_loss=2.0229  r_loss=1.9680  mmlu=0.5600  lr=1.00e-07
[2026-06-10 22:21:12,238][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=250: forget=10.6086  retain=10.2667
[2026-06-10 22:23:02,430][unlearning.trainer.ngdiff][INFO] - step=1050  opt_step=262  f_loss=1.9293  r_loss=2.4107  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:24:52,659][unlearning.trainer.ngdiff][INFO] - step=1100  opt_step=275  f_loss=2.8326  r_loss=2.8399  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:26:46,132][unlearning.trainer.ngdiff][INFO] - step=1150  opt_step=287  f_loss=2.4464  r_loss=1.3978  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:28:36,916][unlearning.trainer.ngdiff][INFO] - step=1200  opt_step=300  f_loss=2.2062  r_loss=2.3671  mmlu=0.5700  lr=8.49e-07
[2026-06-10 22:28:44,288][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=300: forget=10.6369  retain=10.2227
[2026-06-10 22:30:34,902][unlearning.trainer.ngdiff][INFO] - step=1250  opt_step=312  f_loss=2.0137  r_loss=2.4926  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:32:26,968][unlearning.trainer.ngdiff][INFO] - step=1300  opt_step=325  f_loss=3.1024  r_loss=2.4593  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:34:17,704][unlearning.trainer.ngdiff][INFO] - step=1350  opt_step=337  f_loss=2.1423  r_loss=2.8252  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:36:08,511][unlearning.trainer.ngdiff][INFO] - step=1400  opt_step=350  f_loss=1.8490  r_loss=2.9485  mmlu=0.5700  lr=2.22e-05
[2026-06-10 22:36:15,881][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=350: forget=10.6727  retain=10.1638
[2026-06-10 22:38:11,298][unlearning.trainer.ngdiff][INFO] - step=1450  opt_step=362  f_loss=1.8993  r_loss=2.0823  mmlu=0.5800  lr=1.00e-05
[2026-06-10 22:40:03,940][unlearning.trainer.ngdiff][INFO] - step=1500  opt_step=375  f_loss=2.8945  r_loss=2.1040  mmlu=0.5800  lr=1.00e-05
[2026-06-10 22:41:57,129][unlearning.trainer.ngdiff][INFO] - step=1550  opt_step=387  f_loss=1.5971  r_loss=2.4232  mmlu=0.5800  lr=1.00e-05
[2026-06-10 22:43:53,337][unlearning.trainer.ngdiff][INFO] - step=1600  opt_step=400  f_loss=2.1898  r_loss=2.2270  mmlu=0.5800  lr=1.65e-06
[2026-06-10 22:44:00,677][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=400: forget=10.7145  retain=10.1031
[2026-06-10 22:45:50,123][unlearning.trainer.ngdiff][INFO] - step=1650  opt_step=412  f_loss=2.3398  r_loss=2.3097  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:47:43,968][unlearning.trainer.ngdiff][INFO] - step=1700  opt_step=425  f_loss=2.6453  r_loss=2.7049  mmlu=0.5800  lr=1.00e-05
[2026-06-10 22:49:39,271][unlearning.trainer.ngdiff][INFO] - step=1750  opt_step=437  f_loss=2.2078  r_loss=2.4800  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:51:32,538][unlearning.trainer.ngdiff][INFO] - step=1800  opt_step=450  f_loss=2.0859  r_loss=2.1367  mmlu=0.5600  lr=1.57e-05
[2026-06-10 22:51:39,874][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=450: forget=10.7430  retain=10.0637
[2026-06-10 22:53:29,646][unlearning.trainer.ngdiff][INFO] - step=1850  opt_step=462  f_loss=3.1991  r_loss=2.5035  mmlu=0.5800  lr=1.00e-05
[2026-06-10 22:55:24,770][unlearning.trainer.ngdiff][INFO] - step=1900  opt_step=475  f_loss=1.3560  r_loss=1.9779  mmlu=0.5800  lr=1.00e-05
[2026-06-10 22:57:16,701][unlearning.trainer.ngdiff][INFO] - step=1950  opt_step=487  f_loss=1.4043  r_loss=2.3833  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:59:11,122][unlearning.trainer.ngdiff][INFO] - step=2000  opt_step=500  f_loss=2.3197  r_loss=2.1667  mmlu=0.5800  lr=1.24e-05
[2026-06-10 22:59:18,458][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=500: forget=10.7941  retain=9.9906
[2026-06-10 23:01:13,008][unlearning.trainer.ngdiff][INFO] - step=2050  opt_step=512  f_loss=2.4756  r_loss=2.8182  mmlu=0.5800  lr=1.00e-05
[2026-06-10 23:03:05,304][unlearning.trainer.ngdiff][INFO] - step=2100  opt_step=525  f_loss=2.8422  r_loss=2.2571  mmlu=0.5800  lr=1.00e-05
[2026-06-10 23:04:54,395][unlearning.trainer.ngdiff][INFO] - step=2150  opt_step=537  f_loss=2.0178  r_loss=2.2254  mmlu=0.5800  lr=1.00e-05
[2026-06-10 23:06:47,406][unlearning.trainer.ngdiff][INFO] - step=2200  opt_step=550  f_loss=3.6266  r_loss=2.0480  mmlu=0.5700  lr=6.28e-06
[2026-06-10 23:06:54,750][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=550: forget=10.8549  retain=9.9275
[2026-06-10 23:08:48,232][unlearning.trainer.ngdiff][INFO] - step=2250  opt_step=562  f_loss=1.9581  r_loss=1.6419  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:10:41,426][unlearning.trainer.ngdiff][INFO] - step=2300  opt_step=575  f_loss=1.9069  r_loss=1.9823  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:12:35,475][unlearning.trainer.ngdiff][INFO] - step=2350  opt_step=587  f_loss=1.6573  r_loss=2.4402  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:14:30,841][unlearning.trainer.ngdiff][INFO] - step=2400  opt_step=600  f_loss=2.4334  r_loss=2.4960  mmlu=0.5700  lr=7.24e-06
[2026-06-10 23:14:38,220][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=600: forget=11.0901  retain=9.7188
[2026-06-10 23:16:33,724][unlearning.trainer.ngdiff][INFO] - step=2450  opt_step=612  f_loss=1.9601  r_loss=2.0159  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:18:24,005][unlearning.trainer.ngdiff][INFO] - step=2500  opt_step=625  f_loss=2.8232  r_loss=2.0223  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:20:15,438][unlearning.trainer.ngdiff][INFO] - step=2550  opt_step=637  f_loss=2.0888  r_loss=2.0583  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:22:08,634][unlearning.trainer.ngdiff][INFO] - step=2600  opt_step=650  f_loss=2.1284  r_loss=1.9487  mmlu=0.5700  lr=1.00e-07
[2026-06-10 23:22:15,972][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=650: forget=11.2199  retain=9.6792
[2026-06-10 23:24:07,785][unlearning.trainer.ngdiff][INFO] - step=2650  opt_step=662  f_loss=2.0990  r_loss=2.1894  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:26:01,917][unlearning.trainer.ngdiff][INFO] - step=2700  opt_step=675  f_loss=2.8875  r_loss=2.3604  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:27:50,942][unlearning.trainer.ngdiff][INFO] - step=2750  opt_step=687  f_loss=2.9319  r_loss=2.5523  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:29:45,613][unlearning.trainer.ngdiff][INFO] - step=2800  opt_step=700  f_loss=2.4056  r_loss=1.9821  mmlu=0.5700  lr=2.00e-05
[2026-06-10 23:29:52,978][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=700: forget=11.3608  retain=9.5955
[2026-06-10 23:31:44,426][unlearning.trainer.ngdiff][INFO] - step=2850  opt_step=712  f_loss=4.1777  r_loss=2.1118  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:33:38,377][unlearning.trainer.ngdiff][INFO] - step=2900  opt_step=725  f_loss=2.8050  r_loss=2.4858  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:35:28,873][unlearning.trainer.ngdiff][INFO] - step=2950  opt_step=737  f_loss=2.7406  r_loss=1.9421  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:37:21,646][unlearning.trainer.ngdiff][INFO] - step=3000  opt_step=750  f_loss=2.5060  r_loss=1.9172  mmlu=0.5700  lr=1.18e-05
[2026-06-10 23:37:28,994][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=750: forget=11.5559  retain=9.5572
[2026-06-10 23:39:21,071][unlearning.trainer.ngdiff][INFO] - step=3050  opt_step=762  f_loss=2.0184  r_loss=2.0786  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:41:17,413][unlearning.trainer.ngdiff][INFO] - step=3100  opt_step=775  f_loss=2.1433  r_loss=2.5987  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:43:09,335][unlearning.trainer.ngdiff][INFO] - step=3150  opt_step=787  f_loss=2.0926  r_loss=1.7270  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:45:02,191][unlearning.trainer.ngdiff][INFO] - step=3200  opt_step=800  f_loss=2.6213  r_loss=2.3864  mmlu=0.5700  lr=1.00e-07
[2026-06-10 23:45:09,545][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=800: forget=11.7929  retain=9.5667
[2026-06-10 23:47:05,012][unlearning.trainer.ngdiff][INFO] - step=3250  opt_step=812  f_loss=2.3128  r_loss=2.5702  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:48:59,726][unlearning.trainer.ngdiff][INFO] - step=3300  opt_step=825  f_loss=2.2819  r_loss=2.4973  mmlu=0.5500  lr=1.00e-05
[2026-06-10 23:50:52,106][unlearning.trainer.ngdiff][INFO] - step=3350  opt_step=837  f_loss=2.8337  r_loss=2.1182  mmlu=0.5400  lr=1.00e-05
[2026-06-10 23:52:45,662][unlearning.trainer.ngdiff][INFO] - step=3400  opt_step=850  f_loss=2.6097  r_loss=2.4609  mmlu=0.5600  lr=6.57e-06
[2026-06-10 23:52:53,009][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=850: forget=12.2323  retain=9.5790
[2026-06-10 23:54:44,778][unlearning.trainer.ngdiff][INFO] - step=3450  opt_step=862  f_loss=3.1438  r_loss=2.9216  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:56:38,511][unlearning.trainer.ngdiff][INFO] - step=3500  opt_step=875  f_loss=2.8583  r_loss=2.1403  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:58:31,479][unlearning.trainer.ngdiff][INFO] - step=3550  opt_step=887  f_loss=2.4954  r_loss=2.0877  mmlu=0.5400  lr=1.00e-05
[2026-06-11 00:00:27,539][unlearning.trainer.ngdiff][INFO] - step=3600  opt_step=900  f_loss=2.4988  r_loss=2.5268  mmlu=0.5300  lr=5.89e-06
[2026-06-11 00:00:34,876][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=900: forget=12.9248  retain=9.7815
[2026-06-11 00:02:28,377][unlearning.trainer.ngdiff][INFO] - step=3650  opt_step=912  f_loss=2.2893  r_loss=2.4403  mmlu=0.5400  lr=1.00e-05
[2026-06-11 00:04:21,248][unlearning.trainer.ngdiff][INFO] - step=3700  opt_step=925  f_loss=2.8310  r_loss=2.5009  mmlu=0.5300  lr=1.00e-05
[2026-06-11 00:06:12,612][unlearning.trainer.ngdiff][INFO] - step=3750  opt_step=937  f_loss=2.6548  r_loss=3.1903  mmlu=0.5500  lr=1.00e-05
[2026-06-11 00:08:03,792][unlearning.trainer.ngdiff][INFO] - step=3800  opt_step=950  f_loss=4.5938  r_loss=2.1990  mmlu=0.5300  lr=1.00e-07
[2026-06-11 00:08:11,123][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=950: forget=14.2806  retain=9.9733
[2026-06-11 00:10:07,974][unlearning.trainer.ngdiff][INFO] - step=3850  opt_step=962  f_loss=2.8973  r_loss=2.4327  mmlu=0.5400  lr=1.00e-05
[2026-06-11 00:12:00,327][unlearning.trainer.ngdiff][INFO] - step=3900  opt_step=975  f_loss=4.4914  r_loss=2.0327  mmlu=0.5400  lr=1.00e-05
[2026-06-11 00:13:54,802][unlearning.trainer.ngdiff][INFO] - step=3950  opt_step=987  f_loss=3.8155  r_loss=2.3548  mmlu=0.5300  lr=1.00e-05
[2026-06-11 00:15:47,423][unlearning.trainer.ngdiff][INFO] - step=4000  opt_step=1000  f_loss=4.1207  r_loss=2.3076  mmlu=0.5400  lr=7.89e-06
[2026-06-11 00:15:54,753][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1000: forget=17.0459  retain=10.1286
[2026-06-11 00:17:47,414][unlearning.trainer.ngdiff][INFO] - step=4050  opt_step=1012  f_loss=7.3800  r_loss=2.6278  mmlu=0.5300  lr=1.00e-05
[2026-06-11 00:19:41,220][unlearning.trainer.ngdiff][INFO] - step=4100  opt_step=1025  f_loss=5.5899  r_loss=2.4218  mmlu=0.5300  lr=1.00e-05
[2026-06-11 00:21:32,377][unlearning.trainer.ngdiff][INFO] - step=4150  opt_step=1037  f_loss=9.3798  r_loss=2.8340  mmlu=0.5300  lr=1.00e-05
[2026-06-11 00:23:23,441][unlearning.trainer.ngdiff][INFO] - step=4200  opt_step=1050  f_loss=11.1346  r_loss=2.6783  mmlu=0.5400  lr=8.00e-06
[2026-06-11 00:23:31,017][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1050: forget=25.5435  retain=10.0320
[2026-06-11 00:23:31,023][unlearning.trainer.ngdiff][INFO] - PPL stopping triggered at optimizer_step=1050: PPL=25.5435 >= threshold=17.1667
[2026-06-11 00:23:31,460][__main__][INFO] - LoRA adapter saved to /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/challenge/expA__software__arc_challenge_seed42/adapter (step 1092)
[2026-06-11 00:23:31,460][__main__][INFO] - Done.