Buckets:

glennmatlin's picture
download
raw
46.6 kB
[2026-06-25 22:11:41,741][__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: 900
logging_steps: 50
save_steps: 200
save_total_limit: 1
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: 1000
max_retain_docs: 9000
docs_per_retain_bin: null
min_tokens: 0
resample_interval: 0
seed: 42
forget_manifest_path: null
forget_texts_path: null
topic_bin: socialiqa
topic_bins: null
output_dir: /storage/ice-shared/cs7634/staff/TDA/runs/unlearn_32b/contrastive_7b_f_ctrl_seed42
retain_topics: null
seed: 42
wandb_project: social-data-attribution
wandb_entity: hcai-lab
forget_texts_file: /home/hice1/gmatlin3/scratch/forget_texts_32b/f_ctrl.parquet
[2026-06-25 22:11:41,750][__main__][INFO] - Target topic(s): socialiqa
[2026-06-25 22:11:42,090][unlearning.data.dolma_pool][INFO] - Loading 6T-filtered Arrow cache from /home/hice1/gmatlin3/scratch/dolma3_6t_filtered_w ...
[2026-06-25 22:11:42,252][unlearning.data.dolma_pool][INFO] - Loading forget set from texts: 1000/1000 rows kept from /home/hice1/gmatlin3/scratch/forget_texts_32b/f_ctrl.parquet
[2026-06-25 22:11:44,936][unlearning.data.sampling][INFO] - Sampling manifest saved to /storage/ice-shared/cs7634/staff/TDA/runs/unlearn_32b/contrastive_7b_f_ctrl_seed42/sampling_manifest.json
[2026-06-25 22:11:44,937][unlearning.data.dolma_pool][INFO] - Tokenizing 1000 forget + 9000 retain docs ...
[2026-06-25 22:12:24,896][__main__][INFO] - Starting fresh training in: /storage/ice-shared/cs7634/staff/TDA/runs/unlearn_32b/contrastive_7b_f_ctrl_seed42
[2026-06-25 22:12:24,897][__main__][INFO] - Loading model allenai/OLMo-3-1025-7B ...
[2026-06-25 22:12:25,331][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).
[2026-06-25 22:12:34,514][__main__][INFO] - Loading MMLU subset (100 questions) for safety guard ...
[2026-06-25 22:12:34,556][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).
[2026-06-25 22:12:34,576][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).
[2026-06-25 22:12:40,240][__main__][INFO] - Original model MMLU accuracy (n=100): 0.5600
[2026-06-25 22:12:42,053][__main__][INFO] - Base model shuffled-token PPL (forget set, mode=segments): 146.9135
[2026-06-25 22:12:47,254][__main__][INFO] - Starting NGDiff unlearning for topic(s) 'socialiqa'...
[2026-06-25 22:12:47,255][__main__][INFO] - MMLU safety baseline: 0.5600 (threshold: 0.5040)
[2026-06-25 22:14:34,058][unlearning.trainer.ngdiff][INFO] - step=50 opt_step=12 f_loss=2.0038 r_loss=2.5359 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:16:19,783][unlearning.trainer.ngdiff][INFO] - step=100 opt_step=25 f_loss=3.6561 r_loss=2.2031 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:18:05,766][unlearning.trainer.ngdiff][INFO] - step=150 opt_step=37 f_loss=2.0070 r_loss=2.0605 mmlu=0.5700 lr=1.00e-05
[2026-06-25 22:19:50,887][unlearning.trainer.ngdiff][INFO] - step=200 opt_step=50 f_loss=2.0671 r_loss=2.4595 mmlu=0.5600 lr=1.00e-07
[2026-06-25 22:19:58,246][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=50: forget=10.9427 retain=9.1651
[2026-06-25 22:21:40,824][unlearning.trainer.ngdiff][INFO] - step=250 opt_step=62 f_loss=2.6878 r_loss=2.1151 mmlu=0.5700 lr=1.00e-05
[2026-06-25 22:23:26,901][unlearning.trainer.ngdiff][INFO] - step=300 opt_step=75 f_loss=2.7909 r_loss=1.7529 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:25:11,438][unlearning.trainer.ngdiff][INFO] - step=350 opt_step=87 f_loss=2.0028 r_loss=2.6462 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:26:56,002][unlearning.trainer.ngdiff][INFO] - step=400 opt_step=100 f_loss=2.9681 r_loss=1.8865 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:27:01,227][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=100: forget=10.9512 retain=9.1647
[2026-06-25 22:28:42,616][unlearning.trainer.ngdiff][INFO] - step=450 opt_step=112 f_loss=2.4645 r_loss=2.1026 mmlu=0.5700 lr=1.00e-05
[2026-06-25 22:30:20,998][unlearning.trainer.ngdiff][INFO] - step=500 opt_step=125 f_loss=2.3875 r_loss=1.8923 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:32:07,712][unlearning.trainer.ngdiff][INFO] - step=550 opt_step=137 f_loss=2.5790 r_loss=2.4790 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:33:50,461][unlearning.trainer.ngdiff][INFO] - step=600 opt_step=150 f_loss=3.0761 r_loss=2.4326 mmlu=0.5600 lr=1.00e-07
[2026-06-25 22:33:55,692][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=150: forget=10.9465 retain=9.1650
[2026-06-25 22:35:39,079][unlearning.trainer.ngdiff][INFO] - step=650 opt_step=162 f_loss=2.2392 r_loss=2.0296 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:37:21,254][unlearning.trainer.ngdiff][INFO] - step=700 opt_step=175 f_loss=3.2002 r_loss=2.4958 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:39:00,771][unlearning.trainer.ngdiff][INFO] - step=750 opt_step=187 f_loss=2.3976 r_loss=2.5532 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:40:41,000][unlearning.trainer.ngdiff][INFO] - step=800 opt_step=200 f_loss=2.0326 r_loss=1.8230 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:40:46,229][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=200: forget=10.9628 retain=9.1647
[2026-06-25 22:42:32,561][unlearning.trainer.ngdiff][INFO] - step=850 opt_step=212 f_loss=2.5323 r_loss=2.6534 mmlu=0.5700 lr=1.00e-05
[2026-06-25 22:44:22,563][unlearning.trainer.ngdiff][INFO] - step=900 opt_step=225 f_loss=2.5021 r_loss=1.4683 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:46:09,684][unlearning.trainer.ngdiff][INFO] - step=950 opt_step=237 f_loss=2.1528 r_loss=2.3741 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:47:53,541][unlearning.trainer.ngdiff][INFO] - step=1000 opt_step=250 f_loss=2.7810 r_loss=2.6689 mmlu=0.5600 lr=1.00e-07
[2026-06-25 22:47:58,940][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=250: forget=10.9687 retain=9.1636
[2026-06-25 22:49:40,527][unlearning.trainer.ngdiff][INFO] - step=1050 opt_step=262 f_loss=3.1024 r_loss=2.5575 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:51:24,099][unlearning.trainer.ngdiff][INFO] - step=1100 opt_step=275 f_loss=2.0876 r_loss=1.9840 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:53:05,976][unlearning.trainer.ngdiff][INFO] - step=1150 opt_step=287 f_loss=1.8720 r_loss=2.2095 mmlu=0.5600 lr=1.00e-05
[2026-06-25 22:54:51,544][unlearning.trainer.ngdiff][INFO] - step=1200 opt_step=300 f_loss=2.7129 r_loss=2.1397 mmlu=0.5600 lr=1.00e-07
[2026-06-25 22:54:56,954][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=300: forget=10.9966 retain=9.1621
[2026-06-25 22:56:38,815][unlearning.trainer.ngdiff][INFO] - step=1250 opt_step=312 f_loss=2.4616 r_loss=2.9207 mmlu=0.5700 lr=1.00e-05
[2026-06-25 22:58:21,595][unlearning.trainer.ngdiff][INFO] - step=1300 opt_step=325 f_loss=2.2767 r_loss=1.7449 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:00:03,338][unlearning.trainer.ngdiff][INFO] - step=1350 opt_step=337 f_loss=2.4744 r_loss=2.3787 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:01:51,128][unlearning.trainer.ngdiff][INFO] - step=1400 opt_step=350 f_loss=2.2483 r_loss=1.7497 mmlu=0.5500 lr=1.00e-05
[2026-06-25 23:01:56,764][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=350: forget=11.0281 retain=9.1600
[2026-06-25 23:03:35,326][unlearning.trainer.ngdiff][INFO] - step=1450 opt_step=362 f_loss=3.2591 r_loss=2.0019 mmlu=0.5500 lr=1.00e-05
[2026-06-25 23:05:21,052][unlearning.trainer.ngdiff][INFO] - step=1500 opt_step=375 f_loss=2.4307 r_loss=2.6577 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:07:04,150][unlearning.trainer.ngdiff][INFO] - step=1550 opt_step=387 f_loss=2.5532 r_loss=2.8095 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:08:43,431][unlearning.trainer.ngdiff][INFO] - step=1600 opt_step=400 f_loss=2.3069 r_loss=2.0406 mmlu=0.5600 lr=1.00e-07
[2026-06-25 23:08:48,757][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=400: forget=11.0522 retain=9.1599
[2026-06-25 23:10:31,886][unlearning.trainer.ngdiff][INFO] - step=1650 opt_step=412 f_loss=2.2101 r_loss=2.0549 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:12:15,403][unlearning.trainer.ngdiff][INFO] - step=1700 opt_step=425 f_loss=2.6395 r_loss=3.0595 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:13:59,595][unlearning.trainer.ngdiff][INFO] - step=1750 opt_step=437 f_loss=2.0502 r_loss=2.0905 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:15:44,104][unlearning.trainer.ngdiff][INFO] - step=1800 opt_step=450 f_loss=2.9390 r_loss=2.2694 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:15:49,419][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=450: forget=11.0763 retain=9.1594
[2026-06-25 23:17:32,799][unlearning.trainer.ngdiff][INFO] - step=1850 opt_step=462 f_loss=3.5466 r_loss=2.5437 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:19:21,191][unlearning.trainer.ngdiff][INFO] - step=1900 opt_step=475 f_loss=2.4065 r_loss=1.3162 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:21:02,363][unlearning.trainer.ngdiff][INFO] - step=1950 opt_step=487 f_loss=2.8042 r_loss=2.1886 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:22:43,261][unlearning.trainer.ngdiff][INFO] - step=2000 opt_step=500 f_loss=1.9728 r_loss=2.7994 mmlu=0.5700 lr=1.99e-06
[2026-06-25 23:22:48,610][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=500: forget=11.1089 retain=9.1566
[2026-06-25 23:24:29,300][unlearning.trainer.ngdiff][INFO] - step=2050 opt_step=512 f_loss=1.3769 r_loss=2.3116 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:26:16,328][unlearning.trainer.ngdiff][INFO] - step=2100 opt_step=525 f_loss=2.4886 r_loss=1.5559 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:27:57,883][unlearning.trainer.ngdiff][INFO] - step=2150 opt_step=537 f_loss=2.8856 r_loss=2.6026 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:29:41,205][unlearning.trainer.ngdiff][INFO] - step=2200 opt_step=550 f_loss=2.4547 r_loss=1.9813 mmlu=0.5600 lr=1.00e-07
[2026-06-25 23:29:46,570][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=550: forget=11.1628 retain=9.1533
[2026-06-25 23:31:34,785][unlearning.trainer.ngdiff][INFO] - step=2250 opt_step=562 f_loss=2.4385 r_loss=2.5585 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:33:19,900][unlearning.trainer.ngdiff][INFO] - step=2300 opt_step=575 f_loss=3.1369 r_loss=1.4157 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:35:05,515][unlearning.trainer.ngdiff][INFO] - step=2350 opt_step=587 f_loss=2.8572 r_loss=2.3638 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:36:46,497][unlearning.trainer.ngdiff][INFO] - step=2400 opt_step=600 f_loss=2.4134 r_loss=2.2989 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:36:51,850][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=600: forget=11.2140 retain=9.1530
[2026-06-25 23:38:34,173][unlearning.trainer.ngdiff][INFO] - step=2450 opt_step=612 f_loss=2.1848 r_loss=2.1308 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:40:20,456][unlearning.trainer.ngdiff][INFO] - step=2500 opt_step=625 f_loss=3.1107 r_loss=2.2280 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:42:07,095][unlearning.trainer.ngdiff][INFO] - step=2550 opt_step=637 f_loss=1.7479 r_loss=2.3505 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:43:44,816][unlearning.trainer.ngdiff][INFO] - step=2600 opt_step=650 f_loss=2.6320 r_loss=2.2535 mmlu=0.5700 lr=1.00e-07
[2026-06-25 23:43:50,186][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=650: forget=11.2600 retain=9.1521
[2026-06-25 23:45:30,239][unlearning.trainer.ngdiff][INFO] - step=2650 opt_step=662 f_loss=3.5111 r_loss=2.7106 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:47:16,807][unlearning.trainer.ngdiff][INFO] - step=2700 opt_step=675 f_loss=2.4472 r_loss=2.1740 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:49:00,879][unlearning.trainer.ngdiff][INFO] - step=2750 opt_step=687 f_loss=2.4679 r_loss=2.3526 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:50:45,116][unlearning.trainer.ngdiff][INFO] - step=2800 opt_step=700 f_loss=2.2036 r_loss=1.9861 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:50:50,477][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=700: forget=11.2873 retain=9.1514
[2026-06-25 23:52:32,145][unlearning.trainer.ngdiff][INFO] - step=2850 opt_step=712 f_loss=2.4367 r_loss=2.6494 mmlu=0.5700 lr=1.00e-05
[2026-06-25 23:54:14,163][unlearning.trainer.ngdiff][INFO] - step=2900 opt_step=725 f_loss=1.6463 r_loss=2.4180 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:55:56,059][unlearning.trainer.ngdiff][INFO] - step=2950 opt_step=737 f_loss=2.3587 r_loss=2.5348 mmlu=0.5600 lr=1.00e-05
[2026-06-25 23:57:39,497][unlearning.trainer.ngdiff][INFO] - step=3000 opt_step=750 f_loss=3.2087 r_loss=2.5959 mmlu=0.5600 lr=1.00e-07
[2026-06-25 23:57:44,858][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=750: forget=11.3448 retain=9.1486
[2026-06-25 23:59:25,344][unlearning.trainer.ngdiff][INFO] - step=3050 opt_step=762 f_loss=2.7348 r_loss=2.3007 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:01:08,789][unlearning.trainer.ngdiff][INFO] - step=3100 opt_step=775 f_loss=2.3400 r_loss=2.0981 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:02:53,318][unlearning.trainer.ngdiff][INFO] - step=3150 opt_step=787 f_loss=1.9099 r_loss=1.5350 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:04:35,297][unlearning.trainer.ngdiff][INFO] - step=3200 opt_step=800 f_loss=2.4415 r_loss=2.4875 mmlu=0.5600 lr=2.53e-06
[2026-06-26 00:04:40,653][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=800: forget=11.4492 retain=9.1532
[2026-06-26 00:06:28,878][unlearning.trainer.ngdiff][INFO] - step=3250 opt_step=812 f_loss=2.1914 r_loss=2.1120 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:08:06,179][unlearning.trainer.ngdiff][INFO] - step=3300 opt_step=825 f_loss=1.9751 r_loss=2.0018 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:09:49,901][unlearning.trainer.ngdiff][INFO] - step=3350 opt_step=837 f_loss=3.0455 r_loss=2.1235 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:11:37,814][unlearning.trainer.ngdiff][INFO] - step=3400 opt_step=850 f_loss=2.2313 r_loss=1.8802 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:11:43,150][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=850: forget=11.6012 retain=9.1631
[2026-06-26 00:13:25,687][unlearning.trainer.ngdiff][INFO] - step=3450 opt_step=862 f_loss=2.7093 r_loss=2.4395 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:15:08,289][unlearning.trainer.ngdiff][INFO] - step=3500 opt_step=875 f_loss=2.3535 r_loss=2.7879 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:16:51,024][unlearning.trainer.ngdiff][INFO] - step=3550 opt_step=887 f_loss=2.6285 r_loss=2.8364 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:18:28,657][unlearning.trainer.ngdiff][INFO] - step=3600 opt_step=900 f_loss=3.6034 r_loss=2.0934 mmlu=0.5600 lr=1.00e-07
[2026-06-26 00:18:34,018][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=900: forget=11.8252 retain=9.1783
[2026-06-26 00:20:16,839][unlearning.trainer.ngdiff][INFO] - step=3650 opt_step=912 f_loss=2.6209 r_loss=1.8749 mmlu=0.5700 lr=1.00e-05
[2026-06-26 00:21:59,298][unlearning.trainer.ngdiff][INFO] - step=3700 opt_step=925 f_loss=2.1558 r_loss=2.4601 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:23:41,666][unlearning.trainer.ngdiff][INFO] - step=3750 opt_step=937 f_loss=2.3571 r_loss=2.0940 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:25:19,743][unlearning.trainer.ngdiff][INFO] - step=3800 opt_step=950 f_loss=2.5011 r_loss=2.0952 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:25:25,109][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=950: forget=12.3167 retain=9.2140
[2026-06-26 00:27:06,131][unlearning.trainer.ngdiff][INFO] - step=3850 opt_step=962 f_loss=2.1800 r_loss=1.9761 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:28:53,705][unlearning.trainer.ngdiff][INFO] - step=3900 opt_step=975 f_loss=2.1628 r_loss=2.3239 mmlu=0.5500 lr=1.00e-05
[2026-06-26 00:30:38,136][unlearning.trainer.ngdiff][INFO] - step=3950 opt_step=987 f_loss=2.6406 r_loss=1.4373 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:32:22,375][unlearning.trainer.ngdiff][INFO] - step=4000 opt_step=1000 f_loss=1.4964 r_loss=2.2297 mmlu=0.5700 lr=1.22e-05
[2026-06-26 00:32:27,761][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1000: forget=12.4010 retain=9.2165
[2026-06-26 00:34:12,353][unlearning.trainer.ngdiff][INFO] - step=4050 opt_step=1012 f_loss=1.8368 r_loss=2.6112 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:35:53,532][unlearning.trainer.ngdiff][INFO] - step=4100 opt_step=1025 f_loss=1.9856 r_loss=2.0127 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:37:37,359][unlearning.trainer.ngdiff][INFO] - step=4150 opt_step=1037 f_loss=1.8912 r_loss=2.5170 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:39:21,863][unlearning.trainer.ngdiff][INFO] - step=4200 opt_step=1050 f_loss=2.2218 r_loss=2.4373 mmlu=0.5600 lr=1.00e-07
[2026-06-26 00:39:27,240][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1050: forget=12.5000 retain=9.2198
[2026-06-26 00:41:10,229][unlearning.trainer.ngdiff][INFO] - step=4250 opt_step=1062 f_loss=4.0102 r_loss=1.9046 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:42:55,302][unlearning.trainer.ngdiff][INFO] - step=4300 opt_step=1075 f_loss=4.1519 r_loss=2.3719 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:44:33,546][unlearning.trainer.ngdiff][INFO] - step=4350 opt_step=1087 f_loss=2.6019 r_loss=2.2079 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:46:15,757][unlearning.trainer.ngdiff][INFO] - step=4400 opt_step=1100 f_loss=2.6142 r_loss=2.2219 mmlu=0.5500 lr=1.00e-05
[2026-06-26 00:46:21,121][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1100: forget=12.7860 retain=9.2317
[2026-06-26 00:48:08,570][unlearning.trainer.ngdiff][INFO] - step=4450 opt_step=1112 f_loss=2.2408 r_loss=2.8044 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:49:54,031][unlearning.trainer.ngdiff][INFO] - step=4500 opt_step=1125 f_loss=2.0488 r_loss=2.4441 mmlu=0.5700 lr=1.00e-05
[2026-06-26 00:51:41,440][unlearning.trainer.ngdiff][INFO] - step=4550 opt_step=1137 f_loss=1.7424 r_loss=1.8877 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:53:27,134][unlearning.trainer.ngdiff][INFO] - step=4600 opt_step=1150 f_loss=2.4224 r_loss=2.5946 mmlu=0.5500 lr=7.31e-06
[2026-06-26 00:53:32,478][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1150: forget=12.8741 retain=9.2279
[2026-06-26 00:55:14,797][unlearning.trainer.ngdiff][INFO] - step=4650 opt_step=1162 f_loss=3.1731 r_loss=1.6071 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:56:55,117][unlearning.trainer.ngdiff][INFO] - step=4700 opt_step=1175 f_loss=3.9404 r_loss=2.6572 mmlu=0.5600 lr=1.00e-05
[2026-06-26 00:58:39,055][unlearning.trainer.ngdiff][INFO] - step=4750 opt_step=1187 f_loss=2.4293 r_loss=2.4180 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:00:19,811][unlearning.trainer.ngdiff][INFO] - step=4800 opt_step=1200 f_loss=1.5498 r_loss=2.2675 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:00:25,082][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1200: forget=12.8927 retain=9.2149
[2026-06-26 01:02:10,710][unlearning.trainer.ngdiff][INFO] - step=4850 opt_step=1212 f_loss=2.3701 r_loss=2.1027 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:03:58,225][unlearning.trainer.ngdiff][INFO] - step=4900 opt_step=1225 f_loss=1.8622 r_loss=2.2722 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:05:41,637][unlearning.trainer.ngdiff][INFO] - step=4950 opt_step=1237 f_loss=3.3031 r_loss=2.3731 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:07:22,709][unlearning.trainer.ngdiff][INFO] - step=5000 opt_step=1250 f_loss=2.4202 r_loss=1.9333 mmlu=0.5600 lr=1.82e-04
[2026-06-26 01:07:27,997][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1250: forget=13.3228 retain=9.2391
[2026-06-26 01:09:08,096][unlearning.trainer.ngdiff][INFO] - step=5050 opt_step=1262 f_loss=2.1033 r_loss=2.1881 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:10:44,087][unlearning.trainer.ngdiff][INFO] - step=5100 opt_step=1275 f_loss=2.2477 r_loss=1.9874 mmlu=0.5500 lr=1.00e-05
[2026-06-26 01:12:29,422][unlearning.trainer.ngdiff][INFO] - step=5150 opt_step=1287 f_loss=3.0429 r_loss=1.8162 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:14:13,015][unlearning.trainer.ngdiff][INFO] - step=5200 opt_step=1300 f_loss=2.3954 r_loss=2.5325 mmlu=0.5600 lr=1.00e-07
[2026-06-26 01:14:18,305][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1300: forget=13.6125 retain=9.2500
[2026-06-26 01:16:02,260][unlearning.trainer.ngdiff][INFO] - step=5250 opt_step=1312 f_loss=3.0864 r_loss=2.2283 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:17:47,302][unlearning.trainer.ngdiff][INFO] - step=5300 opt_step=1325 f_loss=3.8924 r_loss=2.2402 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:19:29,061][unlearning.trainer.ngdiff][INFO] - step=5350 opt_step=1337 f_loss=2.1966 r_loss=2.1442 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:21:12,689][unlearning.trainer.ngdiff][INFO] - step=5400 opt_step=1350 f_loss=2.8751 r_loss=2.0118 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:21:17,970][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1350: forget=13.6958 retain=9.2499
[2026-06-26 01:23:04,044][unlearning.trainer.ngdiff][INFO] - step=5450 opt_step=1362 f_loss=2.6951 r_loss=2.6488 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:24:51,323][unlearning.trainer.ngdiff][INFO] - step=5500 opt_step=1375 f_loss=3.1361 r_loss=2.2465 mmlu=0.5500 lr=1.00e-05
[2026-06-26 01:26:32,296][unlearning.trainer.ngdiff][INFO] - step=5550 opt_step=1387 f_loss=3.4508 r_loss=2.2310 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:28:19,492][unlearning.trainer.ngdiff][INFO] - step=5600 opt_step=1400 f_loss=2.5834 r_loss=2.4391 mmlu=0.5600 lr=5.96e-06
[2026-06-26 01:28:24,764][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1400: forget=13.9811 retain=9.2535
[2026-06-26 01:30:10,615][unlearning.trainer.ngdiff][INFO] - step=5650 opt_step=1412 f_loss=2.8376 r_loss=2.4788 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:31:51,496][unlearning.trainer.ngdiff][INFO] - step=5700 opt_step=1425 f_loss=2.5703 r_loss=1.6348 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:33:33,951][unlearning.trainer.ngdiff][INFO] - step=5750 opt_step=1437 f_loss=1.8690 r_loss=1.6707 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:35:11,614][unlearning.trainer.ngdiff][INFO] - step=5800 opt_step=1450 f_loss=2.0338 r_loss=2.5790 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:35:16,899][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1450: forget=13.9690 retain=9.2428
[2026-06-26 01:36:58,091][unlearning.trainer.ngdiff][INFO] - step=5850 opt_step=1462 f_loss=2.6983 r_loss=2.3449 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:38:44,843][unlearning.trainer.ngdiff][INFO] - step=5900 opt_step=1475 f_loss=3.5846 r_loss=2.2707 mmlu=0.5500 lr=1.00e-05
[2026-06-26 01:40:22,853][unlearning.trainer.ngdiff][INFO] - step=5950 opt_step=1487 f_loss=3.1967 r_loss=2.1750 mmlu=0.5500 lr=1.00e-05
[2026-06-26 01:42:14,734][unlearning.trainer.ngdiff][INFO] - step=6000 opt_step=1500 f_loss=2.1650 r_loss=2.3570 mmlu=0.5600 lr=1.00e-07
[2026-06-26 01:42:20,005][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1500: forget=14.2711 retain=9.2544
[2026-06-26 01:44:00,451][unlearning.trainer.ngdiff][INFO] - step=6050 opt_step=1512 f_loss=2.5955 r_loss=2.5381 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:45:46,632][unlearning.trainer.ngdiff][INFO] - step=6100 opt_step=1525 f_loss=2.0446 r_loss=2.0161 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:47:24,587][unlearning.trainer.ngdiff][INFO] - step=6150 opt_step=1537 f_loss=1.7157 r_loss=2.1269 mmlu=0.5500 lr=1.00e-05
[2026-06-26 01:49:02,872][unlearning.trainer.ngdiff][INFO] - step=6200 opt_step=1550 f_loss=2.0011 r_loss=2.3543 mmlu=0.5600 lr=6.10e-07
[2026-06-26 01:49:08,005][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1550: forget=14.4880 retain=9.2616
[2026-06-26 01:50:54,706][unlearning.trainer.ngdiff][INFO] - step=6250 opt_step=1562 f_loss=2.2839 r_loss=1.7858 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:52:38,348][unlearning.trainer.ngdiff][INFO] - step=6300 opt_step=1575 f_loss=4.7675 r_loss=2.0749 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:54:23,586][unlearning.trainer.ngdiff][INFO] - step=6350 opt_step=1587 f_loss=2.4376 r_loss=2.2513 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:56:04,696][unlearning.trainer.ngdiff][INFO] - step=6400 opt_step=1600 f_loss=5.3290 r_loss=1.7895 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:56:09,822][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1600: forget=14.7310 retain=9.2745
[2026-06-26 01:57:52,511][unlearning.trainer.ngdiff][INFO] - step=6450 opt_step=1612 f_loss=4.0532 r_loss=1.5335 mmlu=0.5600 lr=1.00e-05
[2026-06-26 01:59:38,220][unlearning.trainer.ngdiff][INFO] - step=6500 opt_step=1625 f_loss=2.6374 r_loss=2.2876 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:01:21,025][unlearning.trainer.ngdiff][INFO] - step=6550 opt_step=1637 f_loss=2.2792 r_loss=2.6491 mmlu=0.5600 lr=1.00e-05
[2026-06-26 02:03:04,478][unlearning.trainer.ngdiff][INFO] - step=6600 opt_step=1650 f_loss=2.4914 r_loss=2.5583 mmlu=0.5600 lr=3.50e-06
[2026-06-26 02:03:09,626][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1650: forget=14.9824 retain=9.2753
[2026-06-26 02:04:53,624][unlearning.trainer.ngdiff][INFO] - step=6650 opt_step=1662 f_loss=2.1775 r_loss=2.3878 mmlu=0.5600 lr=1.00e-05
[2026-06-26 02:06:34,171][unlearning.trainer.ngdiff][INFO] - step=6700 opt_step=1675 f_loss=3.7772 r_loss=1.8332 mmlu=0.5600 lr=1.00e-05
[2026-06-26 02:08:20,237][unlearning.trainer.ngdiff][INFO] - step=6750 opt_step=1687 f_loss=3.0073 r_loss=2.8743 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:10:07,398][unlearning.trainer.ngdiff][INFO] - step=6800 opt_step=1700 f_loss=2.0827 r_loss=2.2610 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:10:12,517][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1700: forget=15.0410 retain=9.2714
[2026-06-26 02:11:55,328][unlearning.trainer.ngdiff][INFO] - step=6850 opt_step=1712 f_loss=3.0244 r_loss=2.4594 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:13:39,836][unlearning.trainer.ngdiff][INFO] - step=6900 opt_step=1725 f_loss=5.1968 r_loss=2.2769 mmlu=0.5600 lr=1.00e-05
[2026-06-26 02:15:21,332][unlearning.trainer.ngdiff][INFO] - step=6950 opt_step=1737 f_loss=2.4811 r_loss=2.3246 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:17:04,974][unlearning.trainer.ngdiff][INFO] - step=7000 opt_step=1750 f_loss=3.0195 r_loss=2.4662 mmlu=0.5500 lr=8.01e-05
[2026-06-26 02:17:10,331][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1750: forget=15.3959 retain=9.2893
[2026-06-26 02:18:50,999][unlearning.trainer.ngdiff][INFO] - step=7050 opt_step=1762 f_loss=6.6175 r_loss=2.2768 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:20:30,269][unlearning.trainer.ngdiff][INFO] - step=7100 opt_step=1775 f_loss=2.3415 r_loss=2.0730 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:22:12,605][unlearning.trainer.ngdiff][INFO] - step=7150 opt_step=1787 f_loss=2.1679 r_loss=2.0299 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:23:53,164][unlearning.trainer.ngdiff][INFO] - step=7200 opt_step=1800 f_loss=4.0754 r_loss=2.1722 mmlu=0.5600 lr=1.00e-07
[2026-06-26 02:23:58,293][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1800: forget=15.8429 retain=9.3115
[2026-06-26 02:25:44,792][unlearning.trainer.ngdiff][INFO] - step=7250 opt_step=1812 f_loss=1.9113 r_loss=1.6841 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:27:24,395][unlearning.trainer.ngdiff][INFO] - step=7300 opt_step=1825 f_loss=2.5913 r_loss=1.9282 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:29:08,420][unlearning.trainer.ngdiff][INFO] - step=7350 opt_step=1837 f_loss=2.1939 r_loss=2.4710 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:30:49,515][unlearning.trainer.ngdiff][INFO] - step=7400 opt_step=1850 f_loss=1.5116 r_loss=1.8893 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:30:54,643][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1850: forget=16.1099 retain=9.3231
[2026-06-26 02:32:38,828][unlearning.trainer.ngdiff][INFO] - step=7450 opt_step=1862 f_loss=3.4492 r_loss=2.2202 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:34:21,484][unlearning.trainer.ngdiff][INFO] - step=7500 opt_step=1875 f_loss=3.1080 r_loss=1.8755 mmlu=0.5600 lr=1.00e-05
[2026-06-26 02:36:08,283][unlearning.trainer.ngdiff][INFO] - step=7550 opt_step=1887 f_loss=2.5240 r_loss=2.1883 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:37:49,457][unlearning.trainer.ngdiff][INFO] - step=7600 opt_step=1900 f_loss=2.6948 r_loss=1.8335 mmlu=0.5500 lr=2.25e-06
[2026-06-26 02:37:54,612][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1900: forget=16.5519 retain=9.3447
[2026-06-26 02:39:40,290][unlearning.trainer.ngdiff][INFO] - step=7650 opt_step=1912 f_loss=1.9206 r_loss=2.5251 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:41:21,761][unlearning.trainer.ngdiff][INFO] - step=7700 opt_step=1925 f_loss=2.5576 r_loss=1.9515 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:43:06,268][unlearning.trainer.ngdiff][INFO] - step=7750 opt_step=1937 f_loss=3.2796 r_loss=1.9040 mmlu=0.5400 lr=1.00e-05
[2026-06-26 02:44:50,730][unlearning.trainer.ngdiff][INFO] - step=7800 opt_step=1950 f_loss=3.0699 r_loss=2.6316 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:44:55,866][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=1950: forget=17.0924 retain=9.3667
[2026-06-26 02:46:37,302][unlearning.trainer.ngdiff][INFO] - step=7850 opt_step=1962 f_loss=2.6203 r_loss=2.0404 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:48:20,344][unlearning.trainer.ngdiff][INFO] - step=7900 opt_step=1975 f_loss=2.4475 r_loss=2.3714 mmlu=0.5400 lr=1.00e-05
[2026-06-26 02:50:07,512][unlearning.trainer.ngdiff][INFO] - step=7950 opt_step=1987 f_loss=5.9762 r_loss=1.9372 mmlu=0.5500 lr=1.00e-05
[2026-06-26 02:51:51,077][unlearning.trainer.ngdiff][INFO] - step=8000 opt_step=2000 f_loss=2.0905 r_loss=2.3116 mmlu=0.5500 lr=1.00e-07
[2026-06-26 02:51:56,204][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2000: forget=17.6301 retain=9.3830
[2026-06-26 02:53:38,905][unlearning.trainer.ngdiff][INFO] - step=8050 opt_step=2012 f_loss=1.7964 r_loss=2.5540 mmlu=0.5400 lr=1.00e-05
[2026-06-26 02:55:25,324][unlearning.trainer.ngdiff][INFO] - step=8100 opt_step=2025 f_loss=2.3070 r_loss=2.0700 mmlu=0.5400 lr=1.00e-05
[2026-06-26 02:57:05,205][unlearning.trainer.ngdiff][INFO] - step=8150 opt_step=2037 f_loss=2.6619 r_loss=2.4229 mmlu=0.5400 lr=1.00e-05
[2026-06-26 02:58:47,770][unlearning.trainer.ngdiff][INFO] - step=8200 opt_step=2050 f_loss=2.2229 r_loss=2.7829 mmlu=0.5400 lr=1.00e-07
[2026-06-26 02:58:52,920][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2050: forget=17.4313 retain=9.3684
[2026-06-26 03:00:36,270][unlearning.trainer.ngdiff][INFO] - step=8250 opt_step=2062 f_loss=2.3959 r_loss=2.4475 mmlu=0.5400 lr=1.00e-05
[2026-06-26 03:02:20,692][unlearning.trainer.ngdiff][INFO] - step=8300 opt_step=2075 f_loss=2.2735 r_loss=1.7880 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:04:06,455][unlearning.trainer.ngdiff][INFO] - step=8350 opt_step=2087 f_loss=2.4001 r_loss=2.1340 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:05:49,721][unlearning.trainer.ngdiff][INFO] - step=8400 opt_step=2100 f_loss=2.0904 r_loss=2.3690 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:05:54,862][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2100: forget=18.6484 retain=9.4131
[2026-06-26 03:07:38,790][unlearning.trainer.ngdiff][INFO] - step=8450 opt_step=2112 f_loss=2.5176 r_loss=2.6240 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:09:21,162][unlearning.trainer.ngdiff][INFO] - step=8500 opt_step=2125 f_loss=2.2959 r_loss=2.0926 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:11:02,442][unlearning.trainer.ngdiff][INFO] - step=8550 opt_step=2137 f_loss=2.7177 r_loss=2.4112 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:12:45,369][unlearning.trainer.ngdiff][INFO] - step=8600 opt_step=2150 f_loss=2.4851 r_loss=2.3869 mmlu=0.5500 lr=1.00e-07
[2026-06-26 03:12:50,504][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2150: forget=18.8886 retain=9.4361
[2026-06-26 03:14:34,642][unlearning.trainer.ngdiff][INFO] - step=8650 opt_step=2162 f_loss=5.8269 r_loss=2.1577 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:16:20,353][unlearning.trainer.ngdiff][INFO] - step=8700 opt_step=2175 f_loss=2.4456 r_loss=2.3534 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:18:03,107][unlearning.trainer.ngdiff][INFO] - step=8750 opt_step=2187 f_loss=4.8325 r_loss=1.8294 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:19:48,447][unlearning.trainer.ngdiff][INFO] - step=8800 opt_step=2200 f_loss=2.6696 r_loss=2.4636 mmlu=0.5300 lr=1.00e-05
[2026-06-26 03:19:53,593][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2200: forget=19.2830 retain=9.4617
[2026-06-26 03:21:41,067][unlearning.trainer.ngdiff][INFO] - step=8850 opt_step=2212 f_loss=2.3649 r_loss=1.8696 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:23:24,403][unlearning.trainer.ngdiff][INFO] - step=8900 opt_step=2225 f_loss=2.2797 r_loss=3.0293 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:25:02,426][unlearning.trainer.ngdiff][INFO] - step=8950 opt_step=2237 f_loss=4.4772 r_loss=2.7519 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:26:43,474][unlearning.trainer.ngdiff][INFO] - step=9000 opt_step=2250 f_loss=2.3559 r_loss=2.3617 mmlu=0.5500 lr=3.70e-06
[2026-06-26 03:26:48,628][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2250: forget=19.8696 retain=9.4903
[2026-06-26 03:28:31,309][unlearning.trainer.ngdiff][INFO] - step=9050 opt_step=2262 f_loss=2.3883 r_loss=1.9638 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:30:16,886][unlearning.trainer.ngdiff][INFO] - step=9100 opt_step=2275 f_loss=2.4248 r_loss=2.3774 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:31:57,841][unlearning.trainer.ngdiff][INFO] - step=9150 opt_step=2287 f_loss=2.1478 r_loss=2.7299 mmlu=0.5400 lr=1.00e-05
[2026-06-26 03:33:41,311][unlearning.trainer.ngdiff][INFO] - step=9200 opt_step=2300 f_loss=2.8197 r_loss=2.4597 mmlu=0.5500 lr=5.51e-06
[2026-06-26 03:33:46,463][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2300: forget=20.3143 retain=9.5076
[2026-06-26 03:35:26,290][unlearning.trainer.ngdiff][INFO] - step=9250 opt_step=2312 f_loss=2.4437 r_loss=1.6596 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:37:13,635][unlearning.trainer.ngdiff][INFO] - step=9300 opt_step=2325 f_loss=2.3350 r_loss=2.1215 mmlu=0.5400 lr=1.00e-05
[2026-06-26 03:38:55,074][unlearning.trainer.ngdiff][INFO] - step=9350 opt_step=2337 f_loss=2.2110 r_loss=2.1652 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:40:34,759][unlearning.trainer.ngdiff][INFO] - step=9400 opt_step=2350 f_loss=2.9532 r_loss=1.5359 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:40:39,899][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2350: forget=21.7445 retain=9.5733
[2026-06-26 03:42:20,589][unlearning.trainer.ngdiff][INFO] - step=9450 opt_step=2362 f_loss=2.3935 r_loss=2.5554 mmlu=0.5400 lr=1.00e-05
[2026-06-26 03:44:01,286][unlearning.trainer.ngdiff][INFO] - step=9500 opt_step=2375 f_loss=2.4598 r_loss=2.2072 mmlu=0.5500 lr=1.00e-05
[2026-06-26 03:45:42,445][unlearning.trainer.ngdiff][INFO] - step=9550 opt_step=2387 f_loss=4.3162 r_loss=2.4699 mmlu=0.5400 lr=1.00e-05
[2026-06-26 03:47:23,290][unlearning.trainer.ngdiff][INFO] - step=9600 opt_step=2400 f_loss=2.3823 r_loss=2.4782 mmlu=0.5400 lr=4.64e-06
[2026-06-26 03:47:28,433][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2400: forget=22.6286 retain=9.5967
[2026-06-26 03:49:10,706][unlearning.trainer.ngdiff][INFO] - step=9650 opt_step=2412 f_loss=2.4548 r_loss=2.2403 mmlu=0.5200 lr=1.00e-05
[2026-06-26 03:50:49,942][unlearning.trainer.ngdiff][INFO] - step=9700 opt_step=2425 f_loss=2.4889 r_loss=2.7058 mmlu=0.5100 lr=1.00e-05
[2026-06-26 03:52:30,233][unlearning.trainer.ngdiff][INFO] - step=9750 opt_step=2437 f_loss=2.2092 r_loss=2.1684 mmlu=0.5300 lr=1.00e-05
[2026-06-26 03:54:17,162][unlearning.trainer.ngdiff][INFO] - step=9800 opt_step=2450 f_loss=8.4708 r_loss=2.2732 mmlu=0.5300 lr=1.00e-05
[2026-06-26 03:54:22,332][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2450: forget=22.7999 retain=9.6186
[2026-06-26 03:56:03,244][unlearning.trainer.ngdiff][INFO] - step=9850 opt_step=2462 f_loss=2.6162 r_loss=1.9010 mmlu=0.5300 lr=1.00e-05
[2026-06-26 03:57:44,800][unlearning.trainer.ngdiff][INFO] - step=9900 opt_step=2475 f_loss=2.6599 r_loss=2.3078 mmlu=0.5300 lr=1.00e-05
[2026-06-26 03:59:29,355][unlearning.trainer.ngdiff][INFO] - step=9950 opt_step=2487 f_loss=1.9369 r_loss=2.1216 mmlu=0.5200 lr=1.00e-05
[2026-06-26 04:01:10,649][unlearning.trainer.ngdiff][INFO] - step=10000 opt_step=2500 f_loss=2.2654 r_loss=2.0639 mmlu=0.5200 lr=1.14e-05
[2026-06-26 04:01:15,799][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2500: forget=23.5626 retain=9.6966
[2026-06-26 04:02:56,779][unlearning.trainer.ngdiff][INFO] - step=10050 opt_step=2512 f_loss=2.8937 r_loss=2.5855 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:04:36,690][unlearning.trainer.ngdiff][INFO] - step=10100 opt_step=2525 f_loss=2.7484 r_loss=1.9355 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:06:17,835][unlearning.trainer.ngdiff][INFO] - step=10150 opt_step=2537 f_loss=3.2353 r_loss=2.2607 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:07:59,037][unlearning.trainer.ngdiff][INFO] - step=10200 opt_step=2550 f_loss=2.3559 r_loss=2.0679 mmlu=0.5300 lr=1.00e-07
[2026-06-26 04:08:04,182][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2550: forget=24.6828 retain=9.8725
[2026-06-26 04:09:49,550][unlearning.trainer.ngdiff][INFO] - step=10250 opt_step=2562 f_loss=2.7599 r_loss=2.2161 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:11:30,199][unlearning.trainer.ngdiff][INFO] - step=10300 opt_step=2575 f_loss=2.1075 r_loss=2.7141 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:13:16,197][unlearning.trainer.ngdiff][INFO] - step=10350 opt_step=2587 f_loss=3.7639 r_loss=2.1468 mmlu=0.5500 lr=1.00e-05
[2026-06-26 04:14:59,049][unlearning.trainer.ngdiff][INFO] - step=10400 opt_step=2600 f_loss=1.6072 r_loss=2.3409 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:15:04,175][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2600: forget=25.7470 retain=9.9480
[2026-06-26 04:16:49,215][unlearning.trainer.ngdiff][INFO] - step=10450 opt_step=2612 f_loss=3.0768 r_loss=2.5016 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:18:32,315][unlearning.trainer.ngdiff][INFO] - step=10500 opt_step=2625 f_loss=3.2589 r_loss=1.9396 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:20:14,950][unlearning.trainer.ngdiff][INFO] - step=10550 opt_step=2637 f_loss=8.7695 r_loss=2.7078 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:21:57,818][unlearning.trainer.ngdiff][INFO] - step=10600 opt_step=2650 f_loss=2.9010 r_loss=2.1527 mmlu=0.5300 lr=1.00e-07
[2026-06-26 04:22:02,947][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2650: forget=25.6698 retain=9.9194
[2026-06-26 04:23:44,290][unlearning.trainer.ngdiff][INFO] - step=10650 opt_step=2662 f_loss=2.5646 r_loss=2.6995 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:25:31,156][unlearning.trainer.ngdiff][INFO] - step=10700 opt_step=2675 f_loss=2.5339 r_loss=2.4849 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:27:14,201][unlearning.trainer.ngdiff][INFO] - step=10750 opt_step=2687 f_loss=2.3022 r_loss=2.6464 mmlu=0.5200 lr=1.00e-05
[2026-06-26 04:29:00,526][unlearning.trainer.ngdiff][INFO] - step=10800 opt_step=2700 f_loss=2.3631 r_loss=1.7233 mmlu=0.5200 lr=1.00e-05
[2026-06-26 04:29:05,655][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2700: forget=25.9209 retain=9.9274
[2026-06-26 04:30:47,633][unlearning.trainer.ngdiff][INFO] - step=10850 opt_step=2712 f_loss=2.5017 r_loss=3.1401 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:32:31,951][unlearning.trainer.ngdiff][INFO] - step=10900 opt_step=2725 f_loss=2.5766 r_loss=2.2919 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:34:14,403][unlearning.trainer.ngdiff][INFO] - step=10950 opt_step=2737 f_loss=2.6636 r_loss=1.7449 mmlu=0.5200 lr=1.00e-05
[2026-06-26 04:35:54,032][unlearning.trainer.ngdiff][INFO] - step=11000 opt_step=2750 f_loss=3.1945 r_loss=1.9003 mmlu=0.5400 lr=7.04e-07
[2026-06-26 04:35:59,165][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2750: forget=28.9815 retain=10.4417
[2026-06-26 04:37:43,575][unlearning.trainer.ngdiff][INFO] - step=11050 opt_step=2762 f_loss=2.4833 r_loss=2.3052 mmlu=0.5500 lr=1.00e-05
[2026-06-26 04:39:26,175][unlearning.trainer.ngdiff][INFO] - step=11100 opt_step=2775 f_loss=2.7980 r_loss=2.6702 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:41:08,606][unlearning.trainer.ngdiff][INFO] - step=11150 opt_step=2787 f_loss=4.6488 r_loss=2.6527 mmlu=0.5500 lr=1.00e-05
[2026-06-26 04:42:50,707][unlearning.trainer.ngdiff][INFO] - step=11200 opt_step=2800 f_loss=3.4869 r_loss=2.6002 mmlu=0.5300 lr=1.00e-07
[2026-06-26 04:42:55,848][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2800: forget=30.7021 retain=10.8064
[2026-06-26 04:44:39,569][unlearning.trainer.ngdiff][INFO] - step=11250 opt_step=2812 f_loss=2.9299 r_loss=2.7885 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:46:20,833][unlearning.trainer.ngdiff][INFO] - step=11300 opt_step=2825 f_loss=2.3392 r_loss=2.0989 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:48:08,711][unlearning.trainer.ngdiff][INFO] - step=11350 opt_step=2837 f_loss=4.2368 r_loss=2.2080 mmlu=0.5500 lr=1.00e-05
[2026-06-26 04:49:50,614][unlearning.trainer.ngdiff][INFO] - step=11400 opt_step=2850 f_loss=2.1150 r_loss=2.3744 mmlu=0.5400 lr=1.00e-05
[2026-06-26 04:49:55,730][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2850: forget=34.3353 retain=11.7278
[2026-06-26 04:51:35,270][unlearning.trainer.ngdiff][INFO] - step=11450 opt_step=2862 f_loss=3.0337 r_loss=2.8851 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:53:20,020][unlearning.trainer.ngdiff][INFO] - step=11500 opt_step=2875 f_loss=4.3468 r_loss=2.5653 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:54:56,056][unlearning.trainer.ngdiff][INFO] - step=11550 opt_step=2887 f_loss=3.7168 r_loss=2.1334 mmlu=0.5300 lr=1.00e-05
[2026-06-26 04:56:40,709][unlearning.trainer.ngdiff][INFO] - step=11600 opt_step=2900 f_loss=5.0271 r_loss=2.6283 mmlu=0.5300 lr=2.54e-05
[2026-06-26 04:56:45,834][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2900: forget=38.5286 retain=12.9835
[2026-06-26 04:58:27,361][unlearning.trainer.ngdiff][INFO] - step=11650 opt_step=2912 f_loss=6.3294 r_loss=3.5963 mmlu=0.5300 lr=1.00e-05
[2026-06-26 05:00:15,164][unlearning.trainer.ngdiff][INFO] - step=11700 opt_step=2925 f_loss=4.5918 r_loss=4.1398 mmlu=0.5300 lr=1.00e-05
[2026-06-26 05:01:55,409][unlearning.trainer.ngdiff][INFO] - step=11750 opt_step=2937 f_loss=7.7814 r_loss=2.6288 mmlu=0.5300 lr=1.00e-05
[2026-06-26 05:03:42,055][unlearning.trainer.ngdiff][INFO] - step=11800 opt_step=2950 f_loss=2.6864 r_loss=1.9077 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:03:47,184][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=2950: forget=71.6130 retain=29.4351
[2026-06-26 05:05:29,795][unlearning.trainer.ngdiff][INFO] - step=11850 opt_step=2962 f_loss=9.9771 r_loss=2.4881 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:07:05,229][unlearning.trainer.ngdiff][INFO] - step=11900 opt_step=2975 f_loss=3.3445 r_loss=2.0081 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:08:51,109][unlearning.trainer.ngdiff][INFO] - step=11950 opt_step=2987 f_loss=13.1929 r_loss=2.2869 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:10:32,705][unlearning.trainer.ngdiff][INFO] - step=12000 opt_step=3000 f_loss=2.5968 r_loss=2.4635 mmlu=0.5200 lr=9.17e-06
[2026-06-26 05:10:37,823][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=3000: forget=81.7004 retain=54.4689
[2026-06-26 05:12:21,451][unlearning.trainer.ngdiff][INFO] - step=12050 opt_step=3012 f_loss=7.1830 r_loss=2.2025 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:14:04,722][unlearning.trainer.ngdiff][INFO] - step=12100 opt_step=3025 f_loss=11.6667 r_loss=2.7864 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:15:46,878][unlearning.trainer.ngdiff][INFO] - step=12150 opt_step=3037 f_loss=13.7216 r_loss=1.9366 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:17:28,777][unlearning.trainer.ngdiff][INFO] - step=12200 opt_step=3050 f_loss=13.1983 r_loss=2.2513 mmlu=0.5200 lr=6.10e-06
[2026-06-26 05:17:33,908][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=3050: forget=135.8072 retain=210.3336
[2026-06-26 05:19:15,179][unlearning.trainer.ngdiff][INFO] - step=12250 opt_step=3062 f_loss=34.2696 r_loss=2.1575 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:20:57,839][unlearning.trainer.ngdiff][INFO] - step=12300 opt_step=3075 f_loss=11.8818 r_loss=15.1864 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:22:39,517][unlearning.trainer.ngdiff][INFO] - step=12350 opt_step=3087 f_loss=35.8440 r_loss=11.1272 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:24:23,557][unlearning.trainer.ngdiff][INFO] - step=12400 opt_step=3100 f_loss=27.8984 r_loss=1.9904 mmlu=0.5200 lr=1.00e-05
[2026-06-26 05:24:28,674][unlearning.trainer.ngdiff][INFO] - PPL at optimizer_step=3100: forget=1112.4339 retain=7623.0900
[2026-06-26 05:24:28,681][unlearning.trainer.ngdiff][INFO] - PPL stopping triggered at optimizer_step=3100: PPL=1112.4339 >= threshold=146.9135
[2026-06-26 05:24:29,101][__main__][INFO] - LoRA adapter saved to /storage/ice-shared/cs7634/staff/TDA/runs/unlearn_32b/contrastive_7b_f_ctrl_seed42/adapter (step 3124)
[2026-06-26 05:24:29,101][__main__][INFO] - Done.

Xet Storage Details

Size:
46.6 kB
·
Xet hash:
fb873beaff4b39711efc49db0887e3bc70326f6637684a015159901fd8be29bd

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.