easy/expA__software_development__arc_easy_seed42/train.log

[2026-06-10 21:49:28,776][__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_development
topic_bins: null
output_dir: /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/easy/expA__software_development__arc_easy_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_easy/expA__software_development__arc_easy.parquet

[2026-06-10 21:49:28,787][__main__][INFO] - Target topic(s): software_development
[2026-06-10 21:49:29,055][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 21:49:34,206][unlearning.data.dolma_pool][INFO] - Faithful forget set from /storage/ice-shared/cs7634/staff/TDA/arc_rescore/forget_texts_binlevel_easy/expA__software_development__arc_easy.parquet: 200 texts
[2026-06-10 21:49:53,300][unlearning.data.dolma_pool][INFO] - Sampled (text-file): 200 forget, 9000 retain.
[2026-06-10 21:49:53,307][unlearning.data.sampling][INFO] - Sampling manifest saved to /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/easy/expA__software_development__arc_easy_seed42/sampling_manifest.json
[2026-06-10 21:49:53,307][unlearning.data.dolma_pool][INFO] - Tokenizing 200 forget + 9000 retain docs ...
[2026-06-10 21:50:36,654][__main__][INFO] - Starting fresh training in: /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/easy/expA__software_development__arc_easy_seed42
[2026-06-10 21:50:36,655][__main__][INFO] - Loading model allenai/OLMo-3-1025-7B ...
[2026-06-10 21:50:37,049][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 21:51:12,889][__main__][INFO] - Loading MMLU subset (100 questions) for safety guard ...
[2026-06-10 21:51:12,926][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 21:51:12,947][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 21:51:18,273][__main__][INFO] - Original model MMLU accuracy (n=100): 0.5600
[2026-06-10 21:51:20,045][__main__][INFO] - Base model shuffled-token PPL (forget set, mode=segments): 33.8214
[2026-06-10 21:51:25,592][__main__][INFO] - Starting NGDiff unlearning for topic(s) 'software_development'...
[2026-06-10 21:51:25,592][__main__][INFO] - MMLU safety baseline: 0.5600  (threshold: 0.5040)
[2026-06-10 21:53:21,911][unlearning.trainer.ngdiff][INFO] - step=50  opt_step=12  f_loss=2.0291  r_loss=1.8194  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:55:16,652][unlearning.trainer.ngdiff][INFO] - step=100  opt_step=25  f_loss=2.2483  r_loss=2.1146  mmlu=0.5700  lr=1.00e-05
[2026-06-10 21:57:08,366][unlearning.trainer.ngdiff][INFO] - step=150  opt_step=37  f_loss=1.9827  r_loss=2.0650  mmlu=0.5600  lr=1.00e-05
[2026-06-10 21:59:04,448][unlearning.trainer.ngdiff][INFO] - step=200  opt_step=50  f_loss=2.6244  r_loss=2.8605  mmlu=0.5600  lr=4.48e-06
[2026-06-10 21:59:12,374][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=50: forget=12.5553  retain=10.2408
[2026-06-10 22:01:07,827][unlearning.trainer.ngdiff][INFO] - step=250  opt_step=62  f_loss=2.4206  r_loss=2.6542  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:02:57,642][unlearning.trainer.ngdiff][INFO] - step=300  opt_step=75  f_loss=1.7719  r_loss=2.4080  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:04:53,101][unlearning.trainer.ngdiff][INFO] - step=350  opt_step=87  f_loss=2.2525  r_loss=2.4838  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:06:47,751][unlearning.trainer.ngdiff][INFO] - step=400  opt_step=100  f_loss=2.2076  r_loss=2.5417  mmlu=0.5600  lr=1.00e-07
[2026-06-10 22:06:53,306][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=100: forget=12.5573  retain=10.2362
[2026-06-10 22:08:48,301][unlearning.trainer.ngdiff][INFO] - step=450  opt_step=112  f_loss=2.4910  r_loss=2.4378  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:10:43,386][unlearning.trainer.ngdiff][INFO] - step=500  opt_step=125  f_loss=2.1322  r_loss=1.4288  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:12:34,162][unlearning.trainer.ngdiff][INFO] - step=550  opt_step=137  f_loss=2.9121  r_loss=2.3423  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:14:24,985][unlearning.trainer.ngdiff][INFO] - step=600  opt_step=150  f_loss=2.4398  r_loss=1.9222  mmlu=0.5600  lr=1.75e-06
[2026-06-10 22:14:30,542][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=150: forget=12.5758  retain=10.2159
[2026-06-10 22:16:23,667][unlearning.trainer.ngdiff][INFO] - step=650  opt_step=162  f_loss=2.1149  r_loss=2.1235  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:18:16,298][unlearning.trainer.ngdiff][INFO] - step=700  opt_step=175  f_loss=1.9814  r_loss=2.1792  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:20:08,033][unlearning.trainer.ngdiff][INFO] - step=750  opt_step=187  f_loss=2.3429  r_loss=2.4267  mmlu=0.5700  lr=1.00e-05
[2026-06-10 22:21:59,796][unlearning.trainer.ngdiff][INFO] - step=800  opt_step=200  f_loss=2.1582  r_loss=2.2365  mmlu=0.5600  lr=1.00e-07
[2026-06-10 22:22:05,361][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=200: forget=12.5982  retain=10.1897
[2026-06-10 22:23:56,161][unlearning.trainer.ngdiff][INFO] - step=850  opt_step=212  f_loss=2.3720  r_loss=2.3127  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:25:45,907][unlearning.trainer.ngdiff][INFO] - step=900  opt_step=225  f_loss=2.8848  r_loss=2.3591  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:27:38,017][unlearning.trainer.ngdiff][INFO] - step=950  opt_step=237  f_loss=2.8027  r_loss=2.2957  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:29:31,268][unlearning.trainer.ngdiff][INFO] - step=1000  opt_step=250  f_loss=2.4497  r_loss=1.9654  mmlu=0.5600  lr=3.22e-06
[2026-06-10 22:29:36,840][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=250: forget=12.7150  retain=10.0380
[2026-06-10 22:31:25,733][unlearning.trainer.ngdiff][INFO] - step=1050  opt_step=262  f_loss=2.1467  r_loss=2.4195  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:33:20,474][unlearning.trainer.ngdiff][INFO] - step=1100  opt_step=275  f_loss=2.4301  r_loss=2.1729  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:35:09,692][unlearning.trainer.ngdiff][INFO] - step=1150  opt_step=287  f_loss=2.7081  r_loss=2.7323  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:37:00,615][unlearning.trainer.ngdiff][INFO] - step=1200  opt_step=300  f_loss=2.6104  r_loss=1.9943  mmlu=0.5600  lr=1.00e-07
[2026-06-10 22:37:06,181][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=300: forget=12.7857  retain=10.0048
[2026-06-10 22:38:56,602][unlearning.trainer.ngdiff][INFO] - step=1250  opt_step=312  f_loss=2.5576  r_loss=2.7110  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:40:48,073][unlearning.trainer.ngdiff][INFO] - step=1300  opt_step=325  f_loss=2.2909  r_loss=2.1514  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:42:36,849][unlearning.trainer.ngdiff][INFO] - step=1350  opt_step=337  f_loss=2.5066  r_loss=1.3628  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:44:31,850][unlearning.trainer.ngdiff][INFO] - step=1400  opt_step=350  f_loss=2.4187  r_loss=2.1937  mmlu=0.5600  lr=8.35e-07
[2026-06-10 22:44:37,412][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=350: forget=12.8928  retain=9.9515
[2026-06-10 22:46:30,778][unlearning.trainer.ngdiff][INFO] - step=1450  opt_step=362  f_loss=1.7121  r_loss=1.6711  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:48:23,428][unlearning.trainer.ngdiff][INFO] - step=1500  opt_step=375  f_loss=2.2098  r_loss=2.6186  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:50:16,894][unlearning.trainer.ngdiff][INFO] - step=1550  opt_step=387  f_loss=2.4498  r_loss=2.3920  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:52:07,773][unlearning.trainer.ngdiff][INFO] - step=1600  opt_step=400  f_loss=2.2411  r_loss=2.0268  mmlu=0.5600  lr=4.25e-06
[2026-06-10 22:52:13,352][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=400: forget=13.0451  retain=9.9025
[2026-06-10 22:54:06,212][unlearning.trainer.ngdiff][INFO] - step=1650  opt_step=412  f_loss=1.9173  r_loss=2.1593  mmlu=0.5500  lr=1.00e-05
[2026-06-10 22:55:58,916][unlearning.trainer.ngdiff][INFO] - step=1700  opt_step=425  f_loss=2.5522  r_loss=2.1458  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:57:51,490][unlearning.trainer.ngdiff][INFO] - step=1750  opt_step=437  f_loss=2.6644  r_loss=1.5414  mmlu=0.5600  lr=1.00e-05
[2026-06-10 22:59:43,821][unlearning.trainer.ngdiff][INFO] - step=1800  opt_step=450  f_loss=2.4904  r_loss=1.8954  mmlu=0.5600  lr=1.00e-07
[2026-06-10 22:59:49,623][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=450: forget=13.2222  retain=9.8755
[2026-06-10 23:01:44,431][unlearning.trainer.ngdiff][INFO] - step=1850  opt_step=462  f_loss=2.4581  r_loss=2.6100  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:03:39,478][unlearning.trainer.ngdiff][INFO] - step=1900  opt_step=475  f_loss=2.3042  r_loss=2.4935  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:05:27,943][unlearning.trainer.ngdiff][INFO] - step=1950  opt_step=487  f_loss=2.4611  r_loss=1.9995  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:07:24,528][unlearning.trainer.ngdiff][INFO] - step=2000  opt_step=500  f_loss=2.2633  r_loss=2.8824  mmlu=0.5600  lr=1.05e-06
[2026-06-10 23:07:30,078][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=500: forget=13.4967  retain=9.8199
[2026-06-10 23:09:22,831][unlearning.trainer.ngdiff][INFO] - step=2050  opt_step=512  f_loss=2.3045  r_loss=1.9331  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:11:17,011][unlearning.trainer.ngdiff][INFO] - step=2100  opt_step=525  f_loss=2.4426  r_loss=2.9838  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:13:09,629][unlearning.trainer.ngdiff][INFO] - step=2150  opt_step=537  f_loss=2.4529  r_loss=2.3730  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:15:04,430][unlearning.trainer.ngdiff][INFO] - step=2200  opt_step=550  f_loss=2.7421  r_loss=1.5400  mmlu=0.5600  lr=1.81e-04
[2026-06-10 23:15:09,983][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=550: forget=13.8892  retain=9.7921
[2026-06-10 23:17:01,493][unlearning.trainer.ngdiff][INFO] - step=2250  opt_step=562  f_loss=2.2991  r_loss=2.0590  mmlu=0.5500  lr=1.00e-05
[2026-06-10 23:18:49,453][unlearning.trainer.ngdiff][INFO] - step=2300  opt_step=575  f_loss=2.1607  r_loss=2.3206  mmlu=0.5500  lr=1.00e-05
[2026-06-10 23:20:40,915][unlearning.trainer.ngdiff][INFO] - step=2350  opt_step=587  f_loss=2.4073  r_loss=2.2408  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:22:33,305][unlearning.trainer.ngdiff][INFO] - step=2400  opt_step=600  f_loss=2.2339  r_loss=2.0834  mmlu=0.5500  lr=3.42e-06
[2026-06-10 23:22:38,816][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=600: forget=14.9246  retain=9.7944
[2026-06-10 23:24:30,347][unlearning.trainer.ngdiff][INFO] - step=2450  opt_step=612  f_loss=3.1077  r_loss=2.3034  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:26:23,860][unlearning.trainer.ngdiff][INFO] - step=2500  opt_step=625  f_loss=2.2964  r_loss=2.4214  mmlu=0.5500  lr=1.00e-05
[2026-06-10 23:28:18,776][unlearning.trainer.ngdiff][INFO] - step=2550  opt_step=637  f_loss=2.4927  r_loss=2.4955  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:30:10,168][unlearning.trainer.ngdiff][INFO] - step=2600  opt_step=650  f_loss=3.0673  r_loss=2.3851  mmlu=0.5500  lr=3.98e-07
[2026-06-10 23:30:15,670][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=650: forget=16.1369  retain=9.7929
[2026-06-10 23:32:10,047][unlearning.trainer.ngdiff][INFO] - step=2650  opt_step=662  f_loss=2.0229  r_loss=2.2803  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:34:01,256][unlearning.trainer.ngdiff][INFO] - step=2700  opt_step=675  f_loss=2.5794  r_loss=2.4859  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:35:55,986][unlearning.trainer.ngdiff][INFO] - step=2750  opt_step=687  f_loss=2.5921  r_loss=2.2119  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:37:47,241][unlearning.trainer.ngdiff][INFO] - step=2800  opt_step=700  f_loss=2.0688  r_loss=2.4626  mmlu=0.5600  lr=1.00e-07
[2026-06-10 23:37:52,770][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=700: forget=18.3015  retain=9.8034
[2026-06-10 23:39:48,253][unlearning.trainer.ngdiff][INFO] - step=2850  opt_step=712  f_loss=2.8494  r_loss=1.7329  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:41:41,942][unlearning.trainer.ngdiff][INFO] - step=2900  opt_step=725  f_loss=2.2539  r_loss=2.3459  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:43:34,759][unlearning.trainer.ngdiff][INFO] - step=2950  opt_step=737  f_loss=2.4940  r_loss=2.3956  mmlu=0.5700  lr=1.00e-05
[2026-06-10 23:45:29,139][unlearning.trainer.ngdiff][INFO] - step=3000  opt_step=750  f_loss=3.3676  r_loss=2.5618  mmlu=0.5700  lr=1.61e-05
[2026-06-10 23:45:34,647][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=750: forget=23.1137  retain=10.0353
[2026-06-10 23:47:27,265][unlearning.trainer.ngdiff][INFO] - step=3050  opt_step=762  f_loss=2.9568  r_loss=2.2681  mmlu=0.5600  lr=1.00e-05
[2026-06-10 23:49:17,539][unlearning.trainer.ngdiff][INFO] - step=3100  opt_step=775  f_loss=2.1878  r_loss=1.6994  mmlu=0.5500  lr=1.00e-05
[2026-06-10 23:51:07,658][unlearning.trainer.ngdiff][INFO] - step=3150  opt_step=787  f_loss=3.9047  r_loss=1.4128  mmlu=0.5500  lr=1.00e-05
[2026-06-10 23:52:58,833][unlearning.trainer.ngdiff][INFO] - step=3200  opt_step=800  f_loss=2.7733  r_loss=2.1079  mmlu=0.5300  lr=1.00e-07
[2026-06-10 23:53:04,582][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=800: forget=33.6301  retain=10.4721
[2026-06-10 23:54:57,710][unlearning.trainer.ngdiff][INFO] - step=3250  opt_step=812  f_loss=3.1650  r_loss=2.3099  mmlu=0.5300  lr=1.00e-05
[2026-06-10 23:56:49,416][unlearning.trainer.ngdiff][INFO] - step=3300  opt_step=825  f_loss=2.3828  r_loss=2.4434  mmlu=0.5300  lr=1.00e-05
[2026-06-10 23:58:40,759][unlearning.trainer.ngdiff][INFO] - step=3350  opt_step=837  f_loss=3.0493  r_loss=1.6804  mmlu=0.5300  lr=1.00e-05
[2026-06-11 00:00:35,476][unlearning.trainer.ngdiff][INFO] - step=3400  opt_step=850  f_loss=4.1680  r_loss=1.6838  mmlu=0.5300  lr=6.19e-06
[2026-06-11 00:00:40,992][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=850: forget=66.3555  retain=11.1794
[2026-06-11 00:00:41,001][unlearning.trainer.ngdiff][INFO] - PPL stopping triggered at optimizer_step=850: PPL=66.3555 >= threshold=33.8214
[2026-06-11 00:00:41,497][__main__][INFO] - LoRA adapter saved to /storage/ice-shared/cs7634/staff/TDA/arc_rescore/binlevel_redo/easy/expA__software_development__arc_easy_seed42/adapter (step 884)
[2026-06-11 00:00:41,498][__main__][INFO] - Done.