Add files using upload-large-folder tool
Browse files- GIT_LOG.txt +17 -0
- logs/eval_IdkDPO_lr1e-05_beta0.05_alpha2_epoch10.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.05_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.05_alpha5_epoch10.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.05_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.1_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.1_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.1_alpha2_epoch10.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.1_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.1_alpha5_epoch10.log +92 -0
- logs/eval_IdkDPO_lr1e-05_beta0.1_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.5_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.5_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.5_alpha2_epoch10.log +78 -0
- logs/eval_IdkDPO_lr1e-05_beta0.5_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.05_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.05_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.05_alpha2_epoch10.log +96 -0
- logs/eval_IdkDPO_lr2e-05_beta0.05_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.05_alpha5_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.05_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.1_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.1_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.1_alpha2_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.1_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.1_alpha5_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.1_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.5_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.5_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.5_alpha2_epoch10.log +92 -0
- logs/eval_IdkDPO_lr2e-05_beta0.5_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.5_alpha5_epoch10.log +78 -0
- logs/eval_IdkDPO_lr2e-05_beta0.5_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.05_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.05_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.05_alpha2_epoch10.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.05_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.05_alpha5_epoch10.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.05_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.1_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.1_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.1_alpha2_epoch10.log +76 -0
- logs/eval_IdkDPO_lr5e-05_beta0.1_alpha2_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.1_alpha5_epoch10.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.1_alpha5_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.5_alpha1_epoch10.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.5_alpha1_epoch5.log +78 -0
- logs/eval_IdkDPO_lr5e-05_beta0.5_alpha2_epoch10.log +78 -0
- logs/eval_sanity_full.log +82 -0
- logs/h2_train_h3rl_WD_npo_l0.log +4 -0
GIT_LOG.txt
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
61a7caa 2026-07-22 paper: add strong-attack finding (IdkDPO dominance is weak-attack artifact)
|
| 2 |
+
13427ec 2026-07-22 supplement: 3-epoch strong attack — H3 negative robust; IdkDPO's dominance is a weak-attack artifact (revives most under strong attack)
|
| 3 |
+
a8a9f0e 2026-07-22 H3: rank/alpha sweep confirms no viable operating point — trade-off cliff, not a tuning failure
|
| 4 |
+
faf3c23 2026-07-22 H3: WD-Ablate (parameter-level J-space deletion) rank sweep — moves causal UDS slightly but does not beat IdkDPO; both intervention families fail identically
|
| 5 |
+
b2710b6 2026-07-21 H3/paper: honest across-checkpoint JOcc-UDS correlation (name rho=-0.75 n=12; knowledge rho=-0.29); fix n=4 overfit claim
|
| 6 |
+
4a5b573 2026-07-21 paper+H3: integrate UDS causal validation — natural methods agree (rho=-1.0), WD-Train Goodharts JOcc while UDS flat
|
| 7 |
+
203c68f 2026-07-21 paper: AAAI-2027 draft (J-Occupancy: diagnostic + predictive + cautionary), positioned vs UDS activation-patching depth
|
| 8 |
+
938569b 2026-07-21 H3: negative result — occupancy suppression does not delete knowledge (within-author relearn falsifies); honest report
|
| 9 |
+
3e12d98 2026-07-21 H3: fixed occupancy (proper-noun concept + margin hinge), final analysis + goodhart scripts
|
| 10 |
+
a616966 2026-07-21 H3: eval + analysis scripts (4-axis pareto, paired JOcc test)
|
| 11 |
+
bd5322b 2026-07-21 H3: WD-Train method (workspace occupancy suppression), prereg, train grid
|
| 12 |
+
107eea4 2026-07-21 H2 results: checkpoint-level predictive validity holds (partial+within z=6.1); sample-level & incremental fail; report + secondary analyses
|
| 13 |
+
da3c0f2 2026-07-21 H2: jailbreak runner in pred phase
|
| 14 |
+
4b3e27e 2026-07-21 H2: jailbreak templates frozen (A2), orchestrator
|
| 15 |
+
c8d0039 2026-07-21 H2 analysis scripts committed before main grid (prereg requirement)
|
| 16 |
+
56a7686 2026-07-21 H2: attack pipeline (relearn_h2, eval_prob parity-checked), pick24 candidates
|
| 17 |
+
dcd685d 2026-07-21 H2 prereg: frozen lists, protocol, lr amendment A1 (pre-attack commit)
|
logs/eval_IdkDPO_lr1e-05_beta0.05_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.04it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.18it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.48it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.85it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.20it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 12.75it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.23it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.91it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 12.94it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 12.88it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 12.91it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 12.99it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:45:58,625[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha2_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:45:58,626[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:45:58,626[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:45:58,626[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:45:59,455[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:46:03,174[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:46:17,509[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:46:17,510[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6038612854549082[0m
|
| 15 |
+
[[36m2026-07-22 16:46:17,514[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:46:17,514[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:46:17,521[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:46:17,522[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.0004368337391825136[0m
|
| 19 |
+
[[36m2026-07-22 16:46:17,935[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:46:20,814[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.08074190139770508[0m
|
| 23 |
+
[[36m2026-07-22 16:46:21,125[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:46:22,921[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.030208501519246443[0m
|
| 27 |
+
[[36m2026-07-22 16:46:23,376[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:46:26,497[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:46:30,740[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:46:33,881[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:46:47,895[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:46:48,296[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:46:49,127[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:46:50,916[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:46:51,230[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:46:51,946[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:46:51,946[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:46:51,946[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:46:52,268[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:46:53,160[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:46:54,699[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:46:55,226[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:46:56,231[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:46:56,231[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:46:56,231[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:46:56,231[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:46:56,231[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5207360594640909[0m
|
| 68 |
+
[[36m2026-07-22 16:46:56,873[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:46:56,881[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.04it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.18it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.48it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.85it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.20it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 12.75it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.23it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.91it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 12.94it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 12.88it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 12.91it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 12.99it/s]
|
| 86 |
+
[[36m2026-07-22 16:46:58,840[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:46:58,847[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:46:58,847[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 2.464011661644601[0m
|
| 89 |
+
[[36m2026-07-22 16:46:59,252[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:47:00,343[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.09190842694187668[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.05_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 12.99it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.01it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.64it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.02it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.45it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.25it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.27it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.18it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.53it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.48it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.60it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.71it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-23 08:13:57,444[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-23 08:13:57,445[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-23 08:13:57,445[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-23 08:13:57,445[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-23 08:13:58,209[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-23 08:14:01,743[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-23 08:14:16,148[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-23 08:14:16,149[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6128740596898488[0m
|
| 15 |
+
[[36m2026-07-23 08:14:16,153[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-23 08:14:16,153[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-23 08:14:16,161[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-23 08:14:16,161[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 3.089852009867272e-06[0m
|
| 19 |
+
[[36m2026-07-23 08:14:16,672[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-23 08:14:19,547[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.1981719970703125[0m
|
| 23 |
+
[[36m2026-07-23 08:14:19,846[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-23 08:14:21,704[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.015541535635926225[0m
|
| 27 |
+
[[36m2026-07-23 08:14:22,010[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-23 08:14:25,108[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-23 08:14:28,988[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-23 08:14:32,163[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-23 08:14:46,170[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-23 08:14:46,597[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-23 08:14:47,451[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-23 08:14:49,053[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-23 08:14:49,365[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-23 08:14:50,236[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-23 08:14:50,236[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-23 08:14:50,236[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-23 08:14:50,533[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-23 08:14:51,317[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-23 08:14:52,873[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-23 08:14:53,209[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-23 08:14:55,289[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-23 08:14:55,289[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-23 08:14:55,289[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-23 08:14:55,290[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-23 08:14:55,290[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.3858647056837753[0m
|
| 68 |
+
[[36m2026-07-23 08:14:55,910[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-23 08:14:55,922[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 12.99it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.01it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.64it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.02it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.45it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.25it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.27it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.18it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.53it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.48it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.60it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.71it/s]
|
| 86 |
+
[[36m2026-07-23 08:14:57,806[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-23 08:14:57,813[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-23 08:14:57,813[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -52.150189297297054[0m
|
| 89 |
+
[[36m2026-07-23 08:14:58,123[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-23 08:14:59,213[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.11449087193379612[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.05_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.27it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.96it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.53it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.88it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.18it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.93it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.73it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.64it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.95it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.92it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.98it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 14.08it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-23 08:03:53,415[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha5_epoch10[0m
|
| 4 |
+
[[36m2026-07-23 08:03:53,416[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-23 08:03:53,416[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-23 08:03:53,416[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-23 08:03:54,127[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-23 08:03:57,614[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-23 08:04:11,972[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-23 08:04:11,973[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5804155683340142[0m
|
| 15 |
+
[[36m2026-07-23 08:04:11,977[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-23 08:04:11,977[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-23 08:04:11,984[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-23 08:04:11,985[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 6.784754222980868e-07[0m
|
| 19 |
+
[[36m2026-07-23 08:04:12,417[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-23 08:04:15,293[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.1843743133544922[0m
|
| 23 |
+
[[36m2026-07-23 08:04:15,636[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-23 08:04:17,446[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.02484487482222637[0m
|
| 27 |
+
[[36m2026-07-23 08:04:17,775[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-23 08:04:20,817[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-23 08:04:25,219[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-23 08:04:28,841[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-23 08:04:42,874[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-23 08:04:43,314[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-23 08:04:44,236[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-23 08:04:45,824[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-23 08:04:46,160[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-23 08:04:46,933[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-23 08:04:46,933[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-23 08:04:46,933[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-23 08:04:47,234[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-23 08:04:48,007[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-23 08:04:49,578[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-23 08:04:50,021[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-23 08:04:50,944[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-23 08:04:50,944[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-23 08:04:50,944[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-23 08:04:50,944[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-23 08:04:50,945[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5389347495390985[0m
|
| 68 |
+
[[36m2026-07-23 08:04:51,545[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-23 08:04:51,552[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.27it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.96it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.53it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.88it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.18it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.93it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.73it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.64it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.95it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.92it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.98it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 14.08it/s]
|
| 86 |
+
[[36m2026-07-23 08:04:53,355[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-23 08:04:53,362[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-23 08:04:53,362[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -48.96792937191507[0m
|
| 89 |
+
[[36m2026-07-23 08:04:53,671[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-23 08:04:54,779[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.11288381028605204[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.05_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.76it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.42it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.97it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.99it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.32it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.02it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.83it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.69it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.01it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.02it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.12it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.19it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:47:27,231[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:47:27,232[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:47:27,232[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:47:27,232[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.05_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:47:28,009[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:47:31,547[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:47:45,859[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:47:45,859[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5738443476129337[0m
|
| 15 |
+
[[36m2026-07-22 16:47:45,864[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:47:45,864[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:47:45,872[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:47:45,872[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.7473118094396964e-09[0m
|
| 19 |
+
[[36m2026-07-22 16:47:46,405[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:47:49,295[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.28228973388671874[0m
|
| 23 |
+
[[36m2026-07-22 16:47:49,663[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:47:51,630[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.02123507729339516[0m
|
| 27 |
+
[[36m2026-07-22 16:47:51,940[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:47:55,168[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:47:59,445[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:48:02,573[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:48:16,567[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:48:17,079[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:48:17,980[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:48:19,589[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:48:19,887[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:48:20,636[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:48:20,636[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:48:20,636[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:48:21,042[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:48:21,846[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:48:23,391[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:48:23,695[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:48:24,669[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:48:24,669[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:48:24,670[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:48:24,670[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:48:24,670[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.48522629853102317[0m
|
| 68 |
+
[[36m2026-07-22 16:48:25,303[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:48:25,309[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.76it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.42it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.97it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.99it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.32it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.02it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.83it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.69it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.01it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.02it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.12it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.19it/s]
|
| 86 |
+
[[36m2026-07-22 16:48:27,230[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:48:27,237[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:48:27,238[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -75.08452955800297[0m
|
| 89 |
+
[[36m2026-07-22 16:48:27,560[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:48:28,730[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.1446340257275569[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.1_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
8%|▊ | 1/13 [00:00<00:01, 8.51it/s]
|
| 2 |
15%|█▌ | 2/13 [00:00<00:01, 7.68it/s]
|
| 3 |
23%|██▎ | 3/13 [00:00<00:01, 6.52it/s]
|
| 4 |
31%|███ | 4/13 [00:00<00:01, 7.29it/s]
|
| 5 |
38%|███▊ | 5/13 [00:00<00:01, 7.01it/s]
|
| 6 |
46%|████▌ | 6/13 [00:00<00:00, 7.13it/s]
|
| 7 |
54%|█████▍ | 7/13 [00:00<00:00, 6.92it/s]
|
| 8 |
62%|██████▏ | 8/13 [00:01<00:00, 7.14it/s]
|
| 9 |
69%|██████▉ | 9/13 [00:01<00:00, 7.17it/s]
|
| 10 |
77%|███████▋ | 10/13 [00:01<00:00, 7.13it/s]
|
| 11 |
85%|████████▍ | 11/13 [00:01<00:00, 6.87it/s]
|
| 12 |
92%|█████████▏| 12/13 [00:01<00:00, 7.01it/s]
|
|
|
|
| 13 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 14 |
8%|▊ | 1/13 [00:00<00:01, 6.89it/s]
|
| 15 |
15%|█▌ | 2/13 [00:00<00:01, 6.77it/s]
|
| 16 |
23%|██▎ | 3/13 [00:00<00:01, 6.68it/s]
|
| 17 |
31%|███ | 4/13 [00:00<00:01, 7.38it/s]
|
| 18 |
38%|███▊ | 5/13 [00:00<00:01, 6.91it/s]
|
| 19 |
46%|████▌ | 6/13 [00:00<00:00, 7.09it/s]
|
| 20 |
54%|█████▍ | 7/13 [00:00<00:00, 7.11it/s]
|
| 21 |
62%|██████▏ | 8/13 [00:01<00:00, 6.90it/s]
|
| 22 |
69%|██████▉ | 9/13 [00:01<00:00, 6.72it/s]
|
| 23 |
77%|███████▋ | 10/13 [00:01<00:00, 7.29it/s]
|
| 24 |
85%|████████▍ | 11/13 [00:01<00:00, 7.90it/s]
|
| 25 |
92%|█████████▏| 12/13 [00:01<00:00, 7.51it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 14:31:03,320[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-20 14:31:03,321[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 14:31:03,321[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 14:31:03,322[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 14:31:04,104[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 14:31:10,431[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 14:31:32,272[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 14:31:32,273[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5261557044293154[0m
|
| 15 |
+
[[36m2026-07-20 14:31:32,282[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 14:31:32,283[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 14:31:32,298[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 14:31:32,300[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 3.1252666340221845e-13[0m
|
| 19 |
+
[[36m2026-07-20 14:31:32,652[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 14:31:36,073[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.46811767578125[0m
|
| 23 |
+
[[36m2026-07-20 14:31:36,358[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 14:31:44,288[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.09865915122700787[0m
|
| 27 |
+
[[36m2026-07-20 14:31:44,660[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-20 14:31:49,232[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-20 14:32:01,341[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-20 14:32:06,774[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-20 14:32:27,724[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-20 14:32:28,069[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-20 14:32:29,138[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-20 14:32:31,590[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-20 14:32:31,853[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-20 14:32:34,377[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-20 14:32:34,377[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-20 14:32:34,377[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-20 14:32:34,673[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-20 14:32:35,796[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 14:32:37,805[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-20 14:32:38,106[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-20 14:32:40,334[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-20 14:32:40,334[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-20 14:32:40,334[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-20 14:32:40,335[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-20 14:32:40,336[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5686667476550382[0m
|
| 68 |
+
[[36m2026-07-20 14:32:40,904[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-20 14:32:40,920[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
8%|▊ | 1/13 [00:00<00:01, 8.51it/s]
|
| 73 |
15%|█▌ | 2/13 [00:00<00:01, 7.68it/s]
|
| 74 |
23%|██▎ | 3/13 [00:00<00:01, 6.52it/s]
|
| 75 |
31%|███ | 4/13 [00:00<00:01, 7.29it/s]
|
| 76 |
38%|███▊ | 5/13 [00:00<00:01, 7.01it/s]
|
| 77 |
46%|████▌ | 6/13 [00:00<00:00, 7.13it/s]
|
| 78 |
54%|█████▍ | 7/13 [00:00<00:00, 6.92it/s]
|
| 79 |
62%|██████▏ | 8/13 [00:01<00:00, 7.14it/s]
|
| 80 |
69%|██████▉ | 9/13 [00:01<00:00, 7.17it/s]
|
| 81 |
77%|███████▋ | 10/13 [00:01<00:00, 7.13it/s]
|
| 82 |
85%|████████▍ | 11/13 [00:01<00:00, 6.87it/s]
|
| 83 |
92%|█████████▏| 12/13 [00:01<00:00, 7.01it/s]
|
| 84 |
+
|
| 85 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 86 |
8%|▊ | 1/13 [00:00<00:01, 6.89it/s]
|
| 87 |
15%|█▌ | 2/13 [00:00<00:01, 6.77it/s]
|
| 88 |
23%|██▎ | 3/13 [00:00<00:01, 6.68it/s]
|
| 89 |
31%|███ | 4/13 [00:00<00:01, 7.38it/s]
|
| 90 |
38%|███▊ | 5/13 [00:00<00:01, 6.91it/s]
|
| 91 |
46%|████▌ | 6/13 [00:00<00:00, 7.09it/s]
|
| 92 |
54%|█████▍ | 7/13 [00:00<00:00, 7.11it/s]
|
| 93 |
62%|██████▏ | 8/13 [00:01<00:00, 6.90it/s]
|
| 94 |
69%|██████▉ | 9/13 [00:01<00:00, 6.72it/s]
|
| 95 |
77%|███████▋ | 10/13 [00:01<00:00, 7.29it/s]
|
| 96 |
85%|████████▍ | 11/13 [00:01<00:00, 7.90it/s]
|
| 97 |
92%|█████████▏| 12/13 [00:01<00:00, 7.51it/s]
|
| 98 |
+
[[36m2026-07-20 14:32:44,443[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 99 |
+
[[36m2026-07-20 14:32:44,458[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 100 |
+
[[36m2026-07-20 14:32:44,458[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -94.21250833646413[0m
|
| 101 |
+
[[36m2026-07-20 14:32:44,770[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 102 |
+
|
| 103 |
+
Evaluated 400 examples
|
| 104 |
+
[[36m2026-07-20 14:32:48,003[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.21131362387257205[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.1_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.97it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.71it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.36it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.71it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.05it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.71it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.37it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.19it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.61it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.58it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.59it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:43:49,174[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:43:49,174[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:43:49,175[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:43:49,175[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:43:49,953[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:43:53,473[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:44:07,918[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:44:07,919[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5256114404377825[0m
|
| 15 |
+
[[36m2026-07-22 16:44:07,923[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:44:07,923[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:44:07,930[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:44:07,930[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.02724452984523e-14[0m
|
| 19 |
+
[[36m2026-07-22 16:44:08,339[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:44:11,223[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.53031982421875[0m
|
| 23 |
+
[[36m2026-07-22 16:44:11,543[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:44:15,114[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.07249865204598734[0m
|
| 27 |
+
[[36m2026-07-22 16:44:15,459[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:44:18,515[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:44:22,907[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:44:26,032[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:44:40,080[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:44:40,561[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:44:41,429[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:44:43,016[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:44:43,315[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:44:44,029[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:44:44,029[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:44:44,029[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:44:44,457[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:44:45,243[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:44:46,811[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:44:47,112[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:44:47,982[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:44:47,982[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:44:47,982[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:44:47,982[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:44:47,982[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5454318654251654[0m
|
| 68 |
+
[[36m2026-07-22 16:44:48,625[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:44:48,632[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.97it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.71it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.36it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.71it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.05it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.71it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.37it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.19it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.61it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.58it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.59it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
| 86 |
+
[[36m2026-07-22 16:44:50,476[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:44:50,483[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:44:50,483[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -96.33131540623494[0m
|
| 89 |
+
[[36m2026-07-22 16:44:50,843[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:44:51,928[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.24401550448673476[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.1_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.11it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.92it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.61it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 14.00it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.40it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 14.10it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.80it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.64it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 14.08it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 14.10it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 14.17it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 14.26it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:50:23,845[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha2_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:50:23,846[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:50:23,846[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:50:23,846[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:50:24,826[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:50:28,391[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:50:42,755[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:50:42,756[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5054628464657762[0m
|
| 15 |
+
[[36m2026-07-22 16:50:42,760[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:50:42,760[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:50:42,767[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:50:42,767[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.4921665744660114e-16[0m
|
| 19 |
+
[[36m2026-07-22 16:50:43,175[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:50:46,020[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.558121337890625[0m
|
| 23 |
+
[[36m2026-07-22 16:50:46,326[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:50:50,427[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.18637759942837584[0m
|
| 27 |
+
[[36m2026-07-22 16:50:50,814[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:50:53,859[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:50:58,015[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:51:01,140[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:51:15,134[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:51:15,564[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:51:16,481[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:51:18,050[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:51:18,362[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:51:19,250[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:51:19,251[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:51:19,251[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:51:19,555[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:51:20,343[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:51:21,903[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:51:22,206[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:51:23,209[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:51:23,209[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:51:23,209[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:51:23,209[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:51:23,209[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5737756660646853[0m
|
| 68 |
+
[[36m2026-07-22 16:51:23,820[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:51:23,827[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.11it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.92it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.61it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 14.00it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.40it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 14.10it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.80it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.64it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 14.08it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 14.10it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 14.17it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 14.26it/s]
|
| 86 |
+
[[36m2026-07-22 16:51:25,612[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:51:25,618[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:51:25,618[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -96.65019941339106[0m
|
| 89 |
+
[[36m2026-07-22 16:51:25,999[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:51:27,042[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.2636539930441975[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.1_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.89it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.44it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.70it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.10it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.40it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.20it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.10it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.84it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.21it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.21it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.28it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.43it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:48:53,985[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:48:53,989[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:48:53,989[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:48:53,989[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:48:54,801[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:48:58,387[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:49:12,761[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:49:12,761[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5045529615691151[0m
|
| 15 |
+
[[36m2026-07-22 16:49:12,765[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:49:12,766[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:49:12,773[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:49:12,773[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.1781277220965948e-17[0m
|
| 19 |
+
[[36m2026-07-22 16:49:13,198[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:49:16,051[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.61283203125[0m
|
| 23 |
+
[[36m2026-07-22 16:49:16,367[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:49:20,174[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.18329288333601224[0m
|
| 27 |
+
[[36m2026-07-22 16:49:20,496[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:49:23,558[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:49:27,962[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:49:31,088[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:49:45,078[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:49:45,480[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:49:46,357[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:49:47,919[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:49:48,222[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:49:49,033[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:49:49,033[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:49:49,033[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:49:49,358[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:49:50,163[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:49:51,717[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:49:52,056[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:49:53,056[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:49:53,056[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:49:53,056[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:49:53,056[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:49:53,057[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5647136265735843[0m
|
| 68 |
+
[[36m2026-07-22 16:49:53,696[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:49:53,703[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.89it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.44it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.70it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.10it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.40it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.20it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.10it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.84it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.21it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.21it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.28it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.43it/s]
|
| 86 |
+
[[36m2026-07-22 16:49:55,600[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:49:55,607[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:49:55,607[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -97.71719542463732[0m
|
| 89 |
+
[[36m2026-07-22 16:49:55,923[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:49:56,993[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.2998217429587749[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.1_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.10it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.26it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 11.57it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 11.71it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 11.98it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:01<00:00, 11.73it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 11.48it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 11.36it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 11.68it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 11.72it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 11.74it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:01<00:00, 11.86it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[[36m2026-07-21 07:48:03,589[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha5_epoch10[0m
|
| 2 |
+
[[36m2026-07-21 07:48:03,591[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 3 |
+
[[36m2026-07-21 07:48:03,591[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 4 |
+
[[36m2026-07-21 07:48:03,591[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 5 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 6 |
+
[[36m2026-07-21 07:48:03,663[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 7 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 8 |
+
[[36m2026-07-21 07:48:03,664[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 9 |
+
[[36m2026-07-21 07:48:03,675[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 10 |
+
|
| 11 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-21 07:48:06,899[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 14 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 15 |
+
[[36m2026-07-21 07:48:06,902[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 16 |
+
[[36m2026-07-21 07:48:06,907[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 17 |
+
|
| 18 |
+
Evaluated 400 examples
|
| 19 |
+
[[36m2026-07-21 07:48:20,041[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 20 |
+
[[36m2026-07-21 07:48:20,042[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4891395807403171[0m
|
| 21 |
+
[[36m2026-07-21 07:48:20,050[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 22 |
+
[[36m2026-07-21 07:48:20,051[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 23 |
+
[[36m2026-07-21 07:48:20,066[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 24 |
+
[[36m2026-07-21 07:48:20,067[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 8.513756583610105e-19[0m
|
| 25 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 26 |
+
[[36m2026-07-21 07:48:20,113[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 27 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 28 |
+
[[36m2026-07-21 07:48:20,115[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 29 |
+
[[36m2026-07-21 07:48:20,120[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 30 |
+
|
| 31 |
+
Evaluated 400 examples
|
| 32 |
+
[[36m2026-07-21 07:48:22,784[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.682255859375[0m
|
| 33 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 34 |
+
[[36m2026-07-21 07:48:22,831[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 35 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 36 |
+
[[36m2026-07-21 07:48:22,833[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 37 |
+
[[36m2026-07-21 07:48:22,838[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 38 |
+
|
| 39 |
+
Evaluated 400 examples
|
| 40 |
+
[[36m2026-07-21 07:48:30,756[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.34626664496042736[0m
|
| 41 |
+
[[36m2026-07-21 07:48:31,463[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 42 |
+
|
| 43 |
+
Evaluated 400 examples
|
| 44 |
+
[[36m2026-07-21 07:48:34,302[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 45 |
+
|
| 46 |
+
Evaluated 400 examples
|
| 47 |
+
[[36m2026-07-21 07:48:41,752[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 400 examples
|
| 50 |
+
[[36m2026-07-21 07:48:44,650[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 51 |
+
|
| 52 |
+
Evaluated 400 examples
|
| 53 |
+
[[36m2026-07-21 07:48:57,479[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 54 |
+
[[36m2026-07-21 07:48:57,889[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 55 |
+
|
| 56 |
+
Evaluated 100 examples
|
| 57 |
+
[[36m2026-07-21 07:48:58,656[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 58 |
+
|
| 59 |
+
Evaluated 100 examples
|
| 60 |
+
[[36m2026-07-21 07:49:00,159[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 61 |
+
[[36m2026-07-21 07:49:00,419[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 62 |
+
|
| 63 |
+
Evaluated 100 examples
|
| 64 |
+
[[36m2026-07-21 07:49:01,861[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-21 07:49:01,862[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 66 |
+
[[36m2026-07-21 07:49:01,862[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 67 |
+
[[36m2026-07-21 07:49:02,145[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 68 |
+
|
| 69 |
+
Evaluated 117 examples
|
| 70 |
+
[[36m2026-07-21 07:49:03,109[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 71 |
+
|
| 72 |
+
Evaluated 117 examples
|
| 73 |
+
[[36m2026-07-21 07:49:04,587[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 74 |
+
[[36m2026-07-21 07:49:04,852[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 75 |
+
|
| 76 |
+
Evaluated 117 examples
|
| 77 |
+
[[36m2026-07-21 07:49:06,632[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 78 |
+
[[36m2026-07-21 07:49:06,632[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 79 |
+
[[36m2026-07-21 07:49:06,632[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 80 |
+
[[36m2026-07-21 07:49:06,633[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 81 |
+
[[36m2026-07-21 07:49:06,633[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5773267601055421[0m
|
| 82 |
+
[[36m2026-07-21 07:49:07,192[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 83 |
+
[[36m2026-07-21 07:49:07,209[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 84 |
+
|
| 85 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 86 |
15%|█▌ | 2/13 [00:00<00:00, 13.10it/s]
|
| 87 |
31%|███ | 4/13 [00:00<00:00, 11.26it/s]
|
| 88 |
46%|████▌ | 6/13 [00:00<00:00, 11.57it/s]
|
| 89 |
62%|██████▏ | 8/13 [00:00<00:00, 11.71it/s]
|
| 90 |
77%|███████▋ | 10/13 [00:00<00:00, 11.98it/s]
|
| 91 |
92%|█████████▏| 12/13 [00:01<00:00, 11.73it/s]
|
| 92 |
+
|
| 93 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 94 |
15%|█▌ | 2/13 [00:00<00:00, 11.48it/s]
|
| 95 |
31%|███ | 4/13 [00:00<00:00, 11.36it/s]
|
| 96 |
46%|████▌ | 6/13 [00:00<00:00, 11.68it/s]
|
| 97 |
62%|██████▏ | 8/13 [00:00<00:00, 11.72it/s]
|
| 98 |
77%|███████▋ | 10/13 [00:00<00:00, 11.74it/s]
|
| 99 |
92%|█████████▏| 12/13 [00:01<00:00, 11.86it/s]
|
| 100 |
+
[[36m2026-07-21 07:49:09,345[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 101 |
+
[[36m2026-07-21 07:49:09,359[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 102 |
+
[[36m2026-07-21 07:49:09,360[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -98.40051829711476[0m
|
| 103 |
+
[[36m2026-07-21 07:49:09,652[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 104 |
+
|
| 105 |
+
Evaluated 400 examples
|
| 106 |
+
[[36m2026-07-21 07:49:10,867[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.3632035856954073[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.1_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.22it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.96it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.60it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.94it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.34it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 14.05it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.77it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.60it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 14.03it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 14.00it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 14.08it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 14.12it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:51:55,587[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:51:55,589[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:51:55,589[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:51:55,589[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.1_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:51:56,533[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:52:00,441[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:52:14,837[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:52:14,838[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4886691839664853[0m
|
| 15 |
+
[[36m2026-07-22 16:52:14,842[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:52:14,842[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:52:14,849[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:52:14,850[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 6.159412995752159e-18[0m
|
| 19 |
+
[[36m2026-07-22 16:52:15,266[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:52:18,124[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.7102978515625[0m
|
| 23 |
+
[[36m2026-07-22 16:52:18,508[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:52:22,653[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.35476941618082974[0m
|
| 27 |
+
[[36m2026-07-22 16:52:23,005[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:52:26,065[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:52:30,259[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:52:33,367[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:52:47,382[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:52:47,857[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:52:48,710[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:52:50,299[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:52:50,602[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:52:51,362[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:52:51,362[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:52:51,362[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:52:51,669[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:52:52,452[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:52:54,030[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:52:54,335[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:52:55,222[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:52:55,222[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:52:55,222[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:52:55,222[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:52:55,223[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5745077151823373[0m
|
| 68 |
+
[[36m2026-07-22 16:52:55,839[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:52:55,845[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.22it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.96it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.60it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.94it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.34it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 14.05it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.77it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.60it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 14.03it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 14.00it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 14.08it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 14.12it/s]
|
| 86 |
+
[[36m2026-07-22 16:52:57,639[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:52:57,646[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:52:57,646[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -98.65461317265823[0m
|
| 89 |
+
[[36m2026-07-22 16:52:58,032[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:52:59,086[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.39110193166022245[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.5_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.38it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.16it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.75it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.07it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.44it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.13it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.76it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.61it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.03it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.09it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.17it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.17it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:50:35,401[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:50:35,402[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:50:35,402[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:50:35,402[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:50:36,166[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:50:39,675[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:50:54,012[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:50:54,012[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4742987749796278[0m
|
| 15 |
+
[[36m2026-07-22 16:50:54,016[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:50:54,017[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:50:54,023[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:50:54,024[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 3.9054713571083378e-22[0m
|
| 19 |
+
[[36m2026-07-22 16:50:54,428[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:50:57,296[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.8646875[0m
|
| 23 |
+
[[36m2026-07-22 16:50:57,621[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:51:02,087[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.7672024557544972[0m
|
| 27 |
+
[[36m2026-07-22 16:51:02,423[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:51:05,501[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:51:09,605[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:51:12,739[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:51:26,725[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:51:27,284[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:51:28,174[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:51:29,770[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:51:30,081[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:51:30,879[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:51:30,880[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:51:30,880[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:51:31,185[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:51:31,977[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:51:33,518[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:51:33,823[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:51:34,723[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:51:34,723[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:51:34,723[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:51:34,723[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:51:34,723[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5902351342263401[0m
|
| 68 |
+
[[36m2026-07-22 16:51:35,491[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:51:35,499[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.38it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.16it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.75it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.07it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.44it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.13it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.76it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.61it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.03it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.09it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.17it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.17it/s]
|
| 86 |
+
[[36m2026-07-22 16:51:37,420[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:51:37,427[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:51:37,427[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.4270211011031[0m
|
| 89 |
+
[[36m2026-07-22 16:51:37,745[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:51:38,813[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.6288417167408347[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.5_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.14it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.80it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.41it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.74it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.11it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.80it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.47it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.32it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.71it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.64it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.53it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.61it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-23 08:15:21,911[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-23 08:15:21,912[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-23 08:15:21,912[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-23 08:15:21,912[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-23 08:15:22,665[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-23 08:15:26,481[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-23 08:15:40,918[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-23 08:15:40,919[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4736869636207382[0m
|
| 15 |
+
[[36m2026-07-23 08:15:40,923[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-23 08:15:40,923[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-23 08:15:40,930[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-23 08:15:40,931[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 3.4004838027527454e-21[0m
|
| 19 |
+
[[36m2026-07-23 08:15:41,340[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-23 08:15:44,215[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.863173828125[0m
|
| 23 |
+
[[36m2026-07-23 08:15:44,578[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-23 08:15:49,310[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.7605217674818161[0m
|
| 27 |
+
[[36m2026-07-23 08:15:49,673[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-23 08:15:52,952[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-23 08:15:57,293[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-23 08:16:00,425[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-23 08:16:14,471[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-23 08:16:14,945[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-23 08:16:15,770[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-23 08:16:17,366[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-23 08:16:17,658[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-23 08:16:18,537[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-23 08:16:18,537[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-23 08:16:18,537[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-23 08:16:18,833[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-23 08:16:19,620[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-23 08:16:21,199[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-23 08:16:21,492[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-23 08:16:22,424[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-23 08:16:22,424[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-23 08:16:22,424[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-23 08:16:22,424[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-23 08:16:22,424[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5859760539860527[0m
|
| 68 |
+
[[36m2026-07-23 08:16:23,287[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-23 08:16:23,294[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.14it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.80it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.41it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.74it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.11it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.80it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.47it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.32it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.71it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.64it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.53it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.61it/s]
|
| 86 |
+
[[36m2026-07-23 08:16:25,133[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-23 08:16:25,141[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-23 08:16:25,141[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.41284847856284[0m
|
| 89 |
+
[[36m2026-07-23 08:16:25,458[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-23 08:16:26,489[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.6303303136195457[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.5_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.71it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.52it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.27it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.70it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.10it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.86it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.62it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.37it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.87it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.87it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.92it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 14.01it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:53:32,201[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha2_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:53:32,202[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:53:32,202[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:53:32,202[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:53:33,139[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:53:36,729[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:53:51,089[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:53:51,090[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.47410013845091875[0m
|
| 15 |
+
[[36m2026-07-22 16:53:51,094[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:53:51,095[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:53:51,102[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:53:51,103[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 8.080285566431044e-22[0m
|
| 19 |
+
[[36m2026-07-22 16:53:51,516[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:53:54,393[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.8659033203125[0m
|
| 23 |
+
[[36m2026-07-22 16:53:54,701[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:53:59,223[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.7683979443717897[0m
|
| 27 |
+
[[36m2026-07-22 16:53:59,541[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:54:02,623[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:54:06,754[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:54:09,910[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:54:23,892[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:54:24,314[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:54:25,154[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:54:26,722[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:54:27,082[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:54:27,830[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:54:27,830[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:54:27,830[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:54:28,130[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:54:28,916[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:54:30,470[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:54:30,778[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:54:31,837[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:54:31,837[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:54:31,837[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:54:31,837[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:54:31,838[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5902324739785918[0m
|
| 68 |
+
[[36m2026-07-22 16:54:32,457[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:54:32,465[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.71it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.52it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.27it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.70it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.10it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.86it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.62it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.37it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.87it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.87it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.92it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 14.01it/s]
|
| 86 |
+
[[36m2026-07-22 16:54:34,285[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:54:34,292[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:54:34,292[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.44828003491352[0m
|
| 89 |
+
[[36m2026-07-22 16:54:34,613[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:54:35,582[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.6355132650627258[0m
|
logs/eval_IdkDPO_lr1e-05_beta0.5_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.31it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.95it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.49it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.90it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.34it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.17it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.07it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.92it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.35it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.34it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.38it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.48it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:52:09,598[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:52:09,599[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:52:09,599[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:52:09,599[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr1e-05_beta0.5_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:52:10,357[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:52:13,973[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:52:28,286[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:52:28,286[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4730085111991299[0m
|
| 15 |
+
[[36m2026-07-22 16:52:28,290[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:52:28,291[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:52:28,298[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:52:28,298[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 3.4004838027527454e-21[0m
|
| 19 |
+
[[36m2026-07-22 16:52:28,713[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:52:31,578[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.864326171875[0m
|
| 23 |
+
[[36m2026-07-22 16:52:31,886[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:52:36,491[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.7618058533136057[0m
|
| 27 |
+
[[36m2026-07-22 16:52:36,860[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:52:39,931[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:52:44,057[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:52:47,217[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:53:01,201[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:53:01,685[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:53:02,583[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:53:04,154[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:53:04,457[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:53:05,205[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:53:05,206[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:53:05,206[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:53:05,507[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:53:06,394[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:53:07,947[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:53:08,260[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:53:09,333[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:53:09,333[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:53:09,333[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:53:09,333[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:53:09,334[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5867730176524717[0m
|
| 68 |
+
[[36m2026-07-22 16:53:09,971[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:53:09,984[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.31it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 11.95it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.49it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.90it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.34it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.17it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.07it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.92it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.35it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.34it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.38it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.48it/s]
|
| 86 |
+
[[36m2026-07-22 16:53:11,888[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:53:11,895[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:53:11,895[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.41791012947009[0m
|
| 89 |
+
[[36m2026-07-22 16:53:12,216[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:53:13,234[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.6269286107126597[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.05_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
8%|▊ | 1/13 [00:00<00:01, 6.19it/s]
|
| 2 |
15%|█▌ | 2/13 [00:00<00:01, 5.74it/s]
|
| 3 |
23%|██▎ | 3/13 [00:00<00:01, 6.67it/s]
|
| 4 |
31%|███ | 4/13 [00:00<00:01, 6.91it/s]
|
| 5 |
46%|████▌ | 6/13 [00:00<00:00, 8.77it/s]
|
| 6 |
54%|█████▍ | 7/13 [00:00<00:00, 8.18it/s]
|
| 7 |
69%|██████▉ | 9/13 [00:01<00:00, 9.88it/s]
|
| 8 |
85%|████████▍ | 11/13 [00:01<00:00, 10.55it/s]
|
|
|
|
| 9 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 10 |
15%|█▌ | 2/13 [00:00<00:00, 11.48it/s]
|
| 11 |
31%|███ | 4/13 [00:00<00:00, 11.35it/s]
|
| 12 |
46%|████▌ | 6/13 [00:00<00:00, 11.65it/s]
|
| 13 |
62%|██████▏ | 8/13 [00:00<00:00, 11.67it/s]
|
| 14 |
77%|███████▋ | 10/13 [00:00<00:00, 11.67it/s]
|
| 15 |
92%|█████████▏| 12/13 [00:01<00:00, 11.78it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 14:37:30,082[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-20 14:37:30,083[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 14:37:30,084[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 14:37:30,084[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 14:37:30,749[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 14:37:36,924[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 14:37:58,293[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 14:37:58,294[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6079179597770539[0m
|
| 15 |
+
[[36m2026-07-20 14:37:58,302[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 14:37:58,303[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 14:37:58,319[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 14:37:58,320[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.9843558982396782[0m
|
| 19 |
+
[[36m2026-07-20 14:37:58,690[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 14:38:02,608[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.01086092198267579[0m
|
| 23 |
+
[[36m2026-07-20 14:38:02,884[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 14:38:09,863[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.04677014857282305[0m
|
| 27 |
+
[[36m2026-07-20 14:38:10,339[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-20 14:38:14,372[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-20 14:38:25,823[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-20 14:38:31,216[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-20 14:38:54,614[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-20 14:38:54,981[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-20 14:38:55,957[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-20 14:38:58,559[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-20 14:38:58,817[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-20 14:39:00,831[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-20 14:39:00,831[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-20 14:39:00,831[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-20 14:39:01,101[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-20 14:39:01,917[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 14:39:03,898[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-20 14:39:04,168[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-20 14:39:07,378[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-20 14:39:07,378[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-20 14:39:07,378[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-20 14:39:07,378[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-20 14:39:07,379[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5738479174848408[0m
|
| 68 |
+
[[36m2026-07-20 14:39:07,938[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-20 14:39:07,953[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
8%|▊ | 1/13 [00:00<00:01, 6.19it/s]
|
| 73 |
15%|█▌ | 2/13 [00:00<00:01, 5.74it/s]
|
| 74 |
23%|██▎ | 3/13 [00:00<00:01, 6.67it/s]
|
| 75 |
31%|███ | 4/13 [00:00<00:01, 6.91it/s]
|
| 76 |
46%|████▌ | 6/13 [00:00<00:00, 8.77it/s]
|
| 77 |
54%|█████▍ | 7/13 [00:00<00:00, 8.18it/s]
|
| 78 |
69%|██████▉ | 9/13 [00:01<00:00, 9.88it/s]
|
| 79 |
85%|████████▍ | 11/13 [00:01<00:00, 10.55it/s]
|
| 80 |
+
|
| 81 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 82 |
15%|█▌ | 2/13 [00:00<00:00, 11.48it/s]
|
| 83 |
31%|███ | 4/13 [00:00<00:00, 11.35it/s]
|
| 84 |
46%|████▌ | 6/13 [00:00<00:00, 11.65it/s]
|
| 85 |
62%|██████▏ | 8/13 [00:00<00:00, 11.67it/s]
|
| 86 |
77%|███████▋ | 10/13 [00:00<00:00, 11.67it/s]
|
| 87 |
92%|█████████▏| 12/13 [00:01<00:00, 11.78it/s]
|
| 88 |
+
[[36m2026-07-20 14:39:10,404[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 89 |
+
[[36m2026-07-20 14:39:10,418[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 90 |
+
[[36m2026-07-20 14:39:10,418[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 57.23512379871069[0m
|
| 91 |
+
[[36m2026-07-20 14:39:10,736[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 92 |
+
|
| 93 |
+
Evaluated 400 examples
|
| 94 |
+
[[36m2026-07-20 14:39:14,062[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.056570082696203915[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.05_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.30it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.01it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.61it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.92it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.32it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.07it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.14it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.91it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.29it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.29it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.36it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.51it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:53:42,102[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:53:42,103[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:53:42,103[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:53:42,103[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:53:42,949[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:53:46,492[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:54:00,846[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:54:00,846[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6220788118434154[0m
|
| 15 |
+
[[36m2026-07-22 16:54:00,851[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:54:00,851[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:54:00,858[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:54:00,859[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.4680435421303692[0m
|
| 19 |
+
[[36m2026-07-22 16:54:01,270[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:54:04,118[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.022943640872836114[0m
|
| 23 |
+
[[36m2026-07-22 16:54:04,424[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:54:06,985[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.03527222107065795[0m
|
| 27 |
+
[[36m2026-07-22 16:54:07,325[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:54:10,399[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:54:14,898[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:54:18,037[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:54:31,986[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:54:32,413[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:54:33,234[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:54:34,797[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:54:35,097[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:54:35,911[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:54:35,911[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:54:35,911[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:54:36,226[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:54:37,027[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:54:38,580[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:54:38,889[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:54:40,200[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:54:40,200[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:54:40,200[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:54:40,201[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:54:40,201[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5404892091238207[0m
|
| 68 |
+
[[36m2026-07-22 16:54:40,819[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:54:40,826[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.30it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.01it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.61it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.92it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.32it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.07it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.14it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.91it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.29it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.29it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.36it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.51it/s]
|
| 86 |
+
[[36m2026-07-22 16:54:42,732[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:54:42,738[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:54:42,739[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 51.871292332308116[0m
|
| 89 |
+
[[36m2026-07-22 16:54:43,082[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:54:44,172[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.05958994460193524[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.05_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,96 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.41it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.55it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 11.86it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.00it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 12.26it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 11.98it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 11.83it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 11.66it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 11.96it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 12.00it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 12.01it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 12.13it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[[36m2026-07-21 07:48:00,964[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha2_epoch10[0m
|
| 2 |
+
[[36m2026-07-21 07:48:00,965[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 3 |
+
[[36m2026-07-21 07:48:00,966[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 4 |
+
[[36m2026-07-21 07:48:00,966[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 5 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 6 |
+
[[36m2026-07-21 07:48:01,037[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 7 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 8 |
+
[[36m2026-07-21 07:48:01,039[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 9 |
+
[[36m2026-07-21 07:48:01,049[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 10 |
+
|
| 11 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-21 07:48:04,272[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 14 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 15 |
+
[[36m2026-07-21 07:48:04,274[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 16 |
+
[[36m2026-07-21 07:48:04,279[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 17 |
+
|
| 18 |
+
Evaluated 400 examples
|
| 19 |
+
[[36m2026-07-21 07:48:17,395[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 20 |
+
[[36m2026-07-21 07:48:17,396[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6190772377979225[0m
|
| 21 |
+
[[36m2026-07-21 07:48:17,404[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 22 |
+
[[36m2026-07-21 07:48:17,405[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 23 |
+
[[36m2026-07-21 07:48:17,420[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 24 |
+
[[36m2026-07-21 07:48:17,422[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.5234101030810641[0m
|
| 25 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 26 |
+
[[36m2026-07-21 07:48:17,467[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 27 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 28 |
+
[[36m2026-07-21 07:48:17,468[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 29 |
+
[[36m2026-07-21 07:48:17,473[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 30 |
+
|
| 31 |
+
Evaluated 400 examples
|
| 32 |
+
[[36m2026-07-21 07:48:20,130[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.011751832466106862[0m
|
| 33 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 34 |
+
[[36m2026-07-21 07:48:20,179[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 35 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 36 |
+
[[36m2026-07-21 07:48:20,181[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 37 |
+
[[36m2026-07-21 07:48:20,186[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 38 |
+
|
| 39 |
+
Evaluated 400 examples
|
| 40 |
+
[[36m2026-07-21 07:48:24,939[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.057885012603382385[0m
|
| 41 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 42 |
+
[[36m2026-07-21 07:48:24,991[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 43 |
+
Found the latest cached dataset configuration 'retain_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/retain_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:22:39 2026).
|
| 44 |
+
[[36m2026-07-21 07:48:25,011[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'retain_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/retain_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:22:39 2026).[0m
|
| 45 |
+
[[36m2026-07-21 07:48:25,016[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 46 |
+
|
| 47 |
+
Evaluated 400 examples
|
| 48 |
+
[[36m2026-07-21 07:48:28,192[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 49 |
+
|
| 50 |
+
Evaluated 400 examples
|
| 51 |
+
[[36m2026-07-21 07:48:36,365[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 52 |
+
|
| 53 |
+
Evaluated 400 examples
|
| 54 |
+
[[36m2026-07-21 07:48:39,265[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 55 |
+
|
| 56 |
+
Evaluated 400 examples
|
| 57 |
+
[[36m2026-07-21 07:48:52,068[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 58 |
+
[[36m2026-07-21 07:48:52,520[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 59 |
+
|
| 60 |
+
Evaluated 100 examples
|
| 61 |
+
[[36m2026-07-21 07:48:53,285[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 62 |
+
|
| 63 |
+
Evaluated 100 examples
|
| 64 |
+
[[36m2026-07-21 07:48:54,784[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 65 |
+
[[36m2026-07-21 07:48:55,051[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 66 |
+
|
| 67 |
+
Evaluated 100 examples
|
| 68 |
+
[[36m2026-07-21 07:48:56,822[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 69 |
+
[[36m2026-07-21 07:48:56,822[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 70 |
+
[[36m2026-07-21 07:48:56,822[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 71 |
+
[[36m2026-07-21 07:48:57,104[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 72 |
+
|
| 73 |
+
Evaluated 117 examples
|
| 74 |
+
[[36m2026-07-21 07:48:57,827[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 75 |
+
|
| 76 |
+
Evaluated 117 examples
|
| 77 |
+
[[36m2026-07-21 07:48:59,297[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 78 |
+
[[36m2026-07-21 07:48:59,553[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 79 |
+
|
| 80 |
+
Evaluated 117 examples
|
| 81 |
+
[[36m2026-07-21 07:49:01,687[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 82 |
+
[[36m2026-07-21 07:49:01,688[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 83 |
+
[[36m2026-07-21 07:49:01,688[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 84 |
+
[[36m2026-07-21 07:49:01,688[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 85 |
+
[[36m2026-07-21 07:49:01,689[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5807218035273619[0m
|
| 86 |
+
[[36m2026-07-21 07:49:02,231[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-21 07:49:02,246[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 88 |
+
|
| 89 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 90 |
15%|█▌ | 2/13 [00:00<00:00, 13.41it/s]
|
| 91 |
31%|███ | 4/13 [00:00<00:00, 11.55it/s]
|
| 92 |
46%|████▌ | 6/13 [00:00<00:00, 11.86it/s]
|
| 93 |
62%|██████▏ | 8/13 [00:00<00:00, 12.00it/s]
|
| 94 |
77%|███████▋ | 10/13 [00:00<00:00, 12.26it/s]
|
| 95 |
92%|█████████▏| 12/13 [00:00<00:00, 11.98it/s]
|
| 96 |
+
|
| 97 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 98 |
15%|█▌ | 2/13 [00:00<00:00, 11.83it/s]
|
| 99 |
31%|███ | 4/13 [00:00<00:00, 11.66it/s]
|
| 100 |
46%|████▌ | 6/13 [00:00<00:00, 11.96it/s]
|
| 101 |
62%|██████▏ | 8/13 [00:00<00:00, 12.00it/s]
|
| 102 |
77%|███████▋ | 10/13 [00:00<00:00, 12.01it/s]
|
| 103 |
92%|█████████▏| 12/13 [00:00<00:00, 12.13it/s]
|
| 104 |
+
[[36m2026-07-21 07:49:04,333[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 105 |
+
[[36m2026-07-21 07:49:04,347[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 106 |
+
[[36m2026-07-21 07:49:04,348[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 55.68929561163949[0m
|
| 107 |
+
[[36m2026-07-21 07:49:04,644[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 108 |
+
|
| 109 |
+
Evaluated 400 examples
|
| 110 |
+
[[36m2026-07-21 07:49:05,908[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.05330560985766884[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.05_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.69it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.26it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.80it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.83it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.11it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 12.89it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.78it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.60it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 12.98it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.19it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.40it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.60it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:47:12,730[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:47:12,731[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:47:12,731[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:47:12,731[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:47:13,575[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:47:17,461[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:47:31,877[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:47:31,878[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6340846037338883[0m
|
| 15 |
+
[[36m2026-07-22 16:47:31,883[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:47:31,883[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:47:31,890[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:47:31,891[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.9843558982396782[0m
|
| 19 |
+
[[36m2026-07-22 16:47:32,331[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:47:35,185[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.020661789551377298[0m
|
| 23 |
+
[[36m2026-07-22 16:47:35,487[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:47:37,766[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.04726549129895672[0m
|
| 27 |
+
[[36m2026-07-22 16:47:38,077[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:47:41,220[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:47:45,654[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:47:48,835[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:48:02,859[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:48:03,290[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:48:04,189[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:48:05,794[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:48:06,109[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:48:07,014[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:48:07,014[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:48:07,014[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:48:07,356[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:48:08,201[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:48:09,762[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:48:10,108[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:48:11,266[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:48:11,266[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:48:11,266[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:48:11,266[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:48:11,267[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5701628088154713[0m
|
| 68 |
+
[[36m2026-07-22 16:48:11,958[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:48:11,965[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.69it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.26it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.80it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.83it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.11it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 12.89it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.78it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.60it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 12.98it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.19it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.40it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.60it/s]
|
| 86 |
+
[[36m2026-07-22 16:48:13,880[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:48:13,886[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:48:13,886[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 49.485736259725755[0m
|
| 89 |
+
[[36m2026-07-22 16:48:14,302[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:48:15,418[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.054280161871334213[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.05_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.23it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.92it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.53it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.87it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.28it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.76it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.60it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.42it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.79it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.80it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.89it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.95it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:56:57,226[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha5_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:56:57,227[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:56:57,227[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:56:57,227[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:56:57,977[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:57:01,602[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:57:15,982[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:57:15,982[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6188798690648576[0m
|
| 15 |
+
[[36m2026-07-22 16:57:15,987[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:57:15,987[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:57:15,994[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:57:15,995[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.3670864569817871[0m
|
| 19 |
+
[[36m2026-07-22 16:57:16,335[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:57:19,218[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.03412718772888183[0m
|
| 23 |
+
[[36m2026-07-22 16:57:19,495[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:57:22,373[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.06810551645015558[0m
|
| 27 |
+
[[36m2026-07-22 16:57:22,683[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:57:25,694[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:57:30,519[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:57:33,603[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:57:47,595[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:57:48,040[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:57:48,824[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:57:50,435[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:57:50,670[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:57:51,462[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:57:51,463[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:57:51,463[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:57:51,877[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:57:52,675[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:57:54,227[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:57:54,461[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:57:55,398[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:57:55,398[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:57:55,398[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:57:55,398[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:57:55,398[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5693805233880749[0m
|
| 68 |
+
[[36m2026-07-22 16:57:55,939[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:57:55,946[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.23it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.92it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.53it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.87it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.28it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.76it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.60it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.42it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.79it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.80it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.89it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.95it/s]
|
| 86 |
+
[[36m2026-07-22 16:57:57,763[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:57:57,770[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:57:57,770[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 37.368650152883134[0m
|
| 89 |
+
[[36m2026-07-22 16:57:58,020[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:57:59,171[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.059179382074830596[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.05_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.92it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.49it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.98it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.26it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.62it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.11it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.96it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.82it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.20it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.19it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.21it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.31it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:55:17,837[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:55:17,838[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:55:17,838[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:55:17,838[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.05_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:55:18,467[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:55:21,960[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:55:36,307[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:55:36,308[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6246550004943207[0m
|
| 15 |
+
[[36m2026-07-22 16:55:36,312[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:55:36,313[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:55:36,322[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:55:36,323[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.28118158736530297[0m
|
| 19 |
+
[[36m2026-07-22 16:55:36,720[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:55:39,603[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.053291425704956055[0m
|
| 23 |
+
[[36m2026-07-22 16:55:39,844[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:55:42,932[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.0646789111165221[0m
|
| 27 |
+
[[36m2026-07-22 16:55:43,227[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:55:46,243[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:55:50,681[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:55:53,746[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:56:07,765[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:56:08,114[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:56:08,934[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:56:10,530[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:56:10,853[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:56:11,711[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:56:11,711[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:56:11,711[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:56:12,001[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:56:12,742[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:56:14,319[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:56:14,564[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:56:15,568[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:56:15,568[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:56:15,568[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:56:15,568[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:56:15,569[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5622993324390867[0m
|
| 68 |
+
[[36m2026-07-22 16:56:16,060[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:56:16,067[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.92it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.49it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.98it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.26it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.62it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.11it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.96it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.82it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.20it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.19it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.21it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.31it/s]
|
| 86 |
+
[[36m2026-07-22 16:56:17,969[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:56:17,976[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:56:17,976[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 23.286124998759124[0m
|
| 89 |
+
[[36m2026-07-22 16:56:18,297[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:56:19,463[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.06581347670590504[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.1_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
8%|▊ | 1/13 [00:00<00:03, 3.70it/s]
|
| 2 |
15%|█▌ | 2/13 [00:00<00:02, 3.67it/s]
|
| 3 |
23%|██▎ | 3/13 [00:00<00:02, 3.49it/s]
|
| 4 |
31%|███ | 4/13 [00:01<00:02, 3.56it/s]
|
| 5 |
38%|███▊ | 5/13 [00:01<00:02, 3.61it/s]
|
| 6 |
46%|████▌ | 6/13 [00:01<00:01, 3.60it/s]
|
| 7 |
54%|█████▍ | 7/13 [00:01<00:01, 3.54it/s]
|
| 8 |
62%|██████▏ | 8/13 [00:02<00:01, 3.57it/s]
|
| 9 |
69%|██████▉ | 9/13 [00:02<00:01, 3.60it/s]
|
| 10 |
77%|███████▋ | 10/13 [00:02<00:00, 3.62it/s]
|
| 11 |
85%|████████▍ | 11/13 [00:03<00:00, 3.54it/s]
|
| 12 |
92%|█████████▏| 12/13 [00:03<00:00, 3.56it/s]
|
|
|
|
| 13 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 14 |
15%|█▌ | 2/13 [00:00<00:00, 12.06it/s]
|
| 15 |
31%|███ | 4/13 [00:00<00:00, 11.73it/s]
|
| 16 |
46%|████▌ | 6/13 [00:00<00:01, 6.85it/s]
|
| 17 |
54%|█████▍ | 7/13 [00:01<00:01, 5.36it/s]
|
| 18 |
62%|██████▏ | 8/13 [00:01<00:01, 4.53it/s]
|
| 19 |
69%|██████▉ | 9/13 [00:01<00:00, 4.42it/s]
|
| 20 |
85%|████████▍ | 11/13 [00:02<00:00, 4.57it/s]
|
| 21 |
92%|█████████▏| 12/13 [00:02<00:00, 4.18it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 14:33:08,347[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-20 14:33:08,349[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 14:33:08,349[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 14:33:08,349[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 14:33:09,145[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 14:33:14,219[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 14:33:30,305[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 14:33:30,306[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5672324813101506[0m
|
| 15 |
+
[[36m2026-07-20 14:33:30,314[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 14:33:30,315[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 14:33:30,331[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 14:33:30,332[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 2.061070745515147e-07[0m
|
| 19 |
+
[[36m2026-07-20 14:33:30,688[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 14:33:33,481[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.23092559814453126[0m
|
| 23 |
+
[[36m2026-07-20 14:33:33,757[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 14:33:43,208[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.20133480492786052[0m
|
| 27 |
+
[[36m2026-07-20 14:33:43,582[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-20 14:33:48,376[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-20 14:34:01,843[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-20 14:34:06,777[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-20 14:34:26,392[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-20 14:34:26,746[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-20 14:34:28,083[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-20 14:34:30,567[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-20 14:34:30,833[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-20 14:34:32,720[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-20 14:34:32,720[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-20 14:34:32,720[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-20 14:34:33,081[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-20 14:34:34,040[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 14:34:36,187[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-20 14:34:36,457[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-20 14:34:38,725[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-20 14:34:38,726[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-20 14:34:38,726[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-20 14:34:38,726[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-20 14:34:38,727[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.575305796992338[0m
|
| 68 |
+
[[36m2026-07-20 14:34:39,312[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-20 14:34:39,327[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
8%|▊ | 1/13 [00:00<00:03, 3.70it/s]
|
| 73 |
15%|█▌ | 2/13 [00:00<00:02, 3.67it/s]
|
| 74 |
23%|██▎ | 3/13 [00:00<00:02, 3.49it/s]
|
| 75 |
31%|███ | 4/13 [00:01<00:02, 3.56it/s]
|
| 76 |
38%|███▊ | 5/13 [00:01<00:02, 3.61it/s]
|
| 77 |
46%|████▌ | 6/13 [00:01<00:01, 3.60it/s]
|
| 78 |
54%|█████▍ | 7/13 [00:01<00:01, 3.54it/s]
|
| 79 |
62%|██████▏ | 8/13 [00:02<00:01, 3.57it/s]
|
| 80 |
69%|██████▉ | 9/13 [00:02<00:01, 3.60it/s]
|
| 81 |
77%|███████▋ | 10/13 [00:02<00:00, 3.62it/s]
|
| 82 |
85%|████████▍ | 11/13 [00:03<00:00, 3.54it/s]
|
| 83 |
92%|█████████▏| 12/13 [00:03<00:00, 3.56it/s]
|
| 84 |
+
|
| 85 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 86 |
15%|█▌ | 2/13 [00:00<00:00, 12.06it/s]
|
| 87 |
31%|███ | 4/13 [00:00<00:00, 11.73it/s]
|
| 88 |
46%|████▌ | 6/13 [00:00<00:01, 6.85it/s]
|
| 89 |
54%|█████▍ | 7/13 [00:01<00:01, 5.36it/s]
|
| 90 |
62%|██████▏ | 8/13 [00:01<00:01, 4.53it/s]
|
| 91 |
69%|██████▉ | 9/13 [00:01<00:00, 4.42it/s]
|
| 92 |
85%|████████▍ | 11/13 [00:02<00:00, 4.57it/s]
|
| 93 |
92%|█████████▏| 12/13 [00:02<00:00, 4.18it/s]
|
| 94 |
+
[[36m2026-07-20 14:34:45,384[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 95 |
+
[[36m2026-07-20 14:34:45,399[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 96 |
+
[[36m2026-07-20 14:34:45,399[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -69.01054846931457[0m
|
| 97 |
+
[[36m2026-07-20 14:34:45,805[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 98 |
+
|
| 99 |
+
Evaluated 400 examples
|
| 100 |
+
[[36m2026-07-20 14:34:57,713[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.10659769849182027[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.1_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.12it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.75it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.35it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.65it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.01it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.69it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.02it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.99it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.29it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.33it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.42it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.53it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:48:41,033[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:48:41,034[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:48:41,034[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:48:41,034[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:48:47,041[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:48:50,574[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:49:04,998[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:49:04,999[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5638023050526287[0m
|
| 15 |
+
[[36m2026-07-22 16:49:05,003[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:49:05,003[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:49:05,010[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:49:05,011[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 2.5496506594547965e-08[0m
|
| 19 |
+
[[36m2026-07-22 16:49:05,433[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:49:08,297[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.2597616577148438[0m
|
| 23 |
+
[[36m2026-07-22 16:49:08,618[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:49:13,084[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.17659373799786976[0m
|
| 27 |
+
[[36m2026-07-22 16:49:13,402[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:49:16,466[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:49:21,466[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:49:24,605[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:49:38,645[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:49:39,082[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:49:39,916[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:49:41,515[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:49:41,821[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:49:42,762[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:49:42,762[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:49:42,762[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:49:43,082[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:49:43,887[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:49:45,450[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:49:45,748[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:49:46,712[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:49:46,712[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:49:46,712[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:49:46,712[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:49:46,712[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5678740971855692[0m
|
| 68 |
+
[[36m2026-07-22 16:49:47,423[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:49:47,429[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.12it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.75it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.35it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.65it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.01it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.69it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.02it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.99it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.29it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.33it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.42it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.53it/s]
|
| 86 |
+
[[36m2026-07-22 16:49:49,288[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:49:49,296[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:49:49,296[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -73.66777346906638[0m
|
| 89 |
+
[[36m2026-07-22 16:49:49,657[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:49:50,797[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.11471381429943271[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.1_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.93it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.67it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.31it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.68it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.02it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.76it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.51it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.96it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.37it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.43it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.51it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:58:39,230[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha2_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:58:39,231[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:58:39,231[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:58:39,231[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:58:39,953[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:58:43,477[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:58:57,840[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:58:57,841[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5410996766317079[0m
|
| 15 |
+
[[36m2026-07-22 16:58:57,845[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:58:57,846[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:58:57,853[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:58:57,854[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 6.85594912218188e-10[0m
|
| 19 |
+
[[36m2026-07-22 16:58:58,195[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:59:01,087[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.304361572265625[0m
|
| 23 |
+
[[36m2026-07-22 16:59:01,381[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:59:05,818[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.23501160256068665[0m
|
| 27 |
+
[[36m2026-07-22 16:59:06,108[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:59:09,147[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:59:13,300[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:59:16,412[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:59:30,401[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:59:30,760[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:59:31,553[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:59:33,156[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:59:33,391[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:59:34,139[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:59:34,139[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:59:34,139[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:59:34,386[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:59:35,096[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:59:36,648[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:59:36,929[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:59:37,884[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:59:37,884[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:59:37,884[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:59:37,885[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:59:37,885[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.573824869114754[0m
|
| 68 |
+
[[36m2026-07-22 16:59:38,452[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:59:38,461[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.93it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.67it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.31it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.68it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.02it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.76it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.51it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.96it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.37it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.43it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.51it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
| 86 |
+
[[36m2026-07-22 16:59:40,309[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:59:40,315[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:59:40,316[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -81.63936748287912[0m
|
| 89 |
+
[[36m2026-07-22 16:59:40,626[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:59:41,695[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.13706057902050917[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.1_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.58it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.18it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.72it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.06it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.34it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.09it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.73it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.65it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 12.96it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 12.90it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.05it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.18it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:56:58,172[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:56:58,173[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:56:58,173[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:56:58,173[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:56:58,865[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:57:02,552[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:57:16,905[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:57:16,905[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5383143845746179[0m
|
| 15 |
+
[[36m2026-07-22 16:57:16,909[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:57:16,910[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:57:16,917[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:57:16,917[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 9.906736183347326e-11[0m
|
| 19 |
+
[[36m2026-07-22 16:57:17,287[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:57:20,181[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.336982421875[0m
|
| 23 |
+
[[36m2026-07-22 16:57:20,428[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:57:24,563[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.20400387363704553[0m
|
| 27 |
+
[[36m2026-07-22 16:57:24,895[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:57:27,953[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:57:31,872[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:57:34,955[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:57:48,958[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:57:49,307[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:57:50,124[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:57:51,721[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:57:52,023[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:57:52,918[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:57:52,918[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:57:52,918[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:57:53,182[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:57:53,888[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:57:55,425[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:57:55,722[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:57:56,644[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:57:56,645[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:57:56,645[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:57:56,645[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:57:56,645[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5684966547761952[0m
|
| 68 |
+
[[36m2026-07-22 16:57:57,232[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:57:57,244[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.58it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.18it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.72it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.06it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.34it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.09it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.73it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.65it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 12.96it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 12.90it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.05it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.18it/s]
|
| 86 |
+
[[36m2026-07-22 16:57:59,170[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:57:59,177[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:57:59,177[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -85.61984975633297[0m
|
| 89 |
+
[[36m2026-07-22 16:57:59,449[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:58:00,597[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.1462223353706454[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.1_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.83it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.46it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.90it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.27it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.57it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.20it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.61it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.87it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.23it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.13it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.30it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.17it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:50:22,568[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha5_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:50:22,569[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:50:22,569[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:50:22,569[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:50:28,397[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:50:32,101[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:50:46,495[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:50:46,495[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5166679354142635[0m
|
| 15 |
+
[[36m2026-07-22 16:50:46,500[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:50:46,500[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:50:46,507[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:50:46,508[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.0229984068189572e-13[0m
|
| 19 |
+
[[36m2026-07-22 16:50:46,933[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:50:49,817[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.441942138671875[0m
|
| 23 |
+
[[36m2026-07-22 16:50:50,127[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:50:54,878[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.3003619044610984[0m
|
| 27 |
+
[[36m2026-07-22 16:50:55,200[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:50:58,291[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:51:02,394[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:51:05,537[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:51:19,555[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:51:19,994[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:51:20,826[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:51:22,546[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:51:22,888[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:51:23,672[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:51:23,673[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:51:23,673[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:51:24,034[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:51:24,815[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:51:26,356[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:51:26,674[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:51:27,689[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:51:27,689[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:51:27,689[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:51:27,689[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:51:27,690[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5768412422779892[0m
|
| 68 |
+
[[36m2026-07-22 16:51:28,395[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:51:28,403[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.83it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.46it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.90it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.27it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.57it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.20it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.61it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.87it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.23it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.13it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.30it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.17it/s]
|
| 86 |
+
[[36m2026-07-22 16:51:30,314[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:51:30,321[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:51:30,321[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -93.02099571289975[0m
|
| 89 |
+
[[36m2026-07-22 16:51:30,640[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:51:31,740[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.198816879429523[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.1_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 12.95it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.32it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.76it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.35it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.66it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.09it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.28it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.11it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.52it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.49it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.65it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.89it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:00:19,596[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:00:19,597[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:00:19,597[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:00:19,597[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.1_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:00:20,243[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:00:23,774[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:00:38,148[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:00:38,148[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5126241146891207[0m
|
| 15 |
+
[[36m2026-07-22 17:00:38,152[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:00:38,153[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:00:38,160[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:00:38,161[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 5.7046977773929285e-15[0m
|
| 19 |
+
[[36m2026-07-22 17:00:38,517[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:00:41,388[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.4754931640625[0m
|
| 23 |
+
[[36m2026-07-22 17:00:41,707[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:00:46,024[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.2769141873896573[0m
|
| 27 |
+
[[36m2026-07-22 17:00:46,312[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:00:49,360[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:00:53,275[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:00:56,384[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:01:10,364[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:01:10,715[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:01:11,680[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:01:13,265[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:01:13,498[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:01:14,219[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:01:14,219[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:01:14,219[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:01:14,460[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:01:15,195[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:01:16,749[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:01:16,997[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:01:17,921[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:01:17,921[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:01:17,921[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:01:17,922[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:01:17,922[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5706611610726314[0m
|
| 68 |
+
[[36m2026-07-22 17:01:18,712[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:01:18,720[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 12.95it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.32it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.76it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.35it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.66it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.09it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.28it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.11it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.52it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.49it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.65it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.89it/s]
|
| 86 |
+
[[36m2026-07-22 17:01:20,594[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:01:20,601[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:01:20,601[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -94.43522097638271[0m
|
| 89 |
+
[[36m2026-07-22 17:01:20,959[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:01:22,006[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.2104713825301646[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.5_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.04it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.18it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:01, 5.43it/s]
|
| 4 |
54%|█████▍ | 7/13 [00:01<00:01, 4.59it/s]
|
| 5 |
62%|██████▏ | 8/13 [00:01<00:01, 4.25it/s]
|
| 6 |
69%|██████▉ | 9/13 [00:01<00:00, 4.06it/s]
|
| 7 |
77%|███████▋ | 10/13 [00:02<00:00, 3.88it/s]
|
| 8 |
85%|████████▍ | 11/13 [00:02<00:00, 3.75it/s]
|
| 9 |
92%|█████████▏| 12/13 [00:02<00:00, 3.72it/s]
|
|
|
|
| 10 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 11 |
8%|▊ | 1/13 [00:00<00:01, 7.19it/s]
|
| 12 |
15%|█▌ | 2/13 [00:00<00:02, 4.37it/s]
|
| 13 |
23%|██▎ | 3/13 [00:00<00:02, 3.82it/s]
|
| 14 |
31%|███ | 4/13 [00:01<00:02, 3.59it/s]
|
| 15 |
38%|███▊ | 5/13 [00:01<00:02, 3.74it/s]
|
| 16 |
54%|█████▍ | 7/13 [00:01<00:01, 5.97it/s]
|
| 17 |
69%|██████▉ | 9/13 [00:01<00:00, 7.49it/s]
|
| 18 |
85%|████████▍ | 11/13 [00:01<00:00, 7.30it/s]
|
| 19 |
92%|█████████▏| 12/13 [00:02<00:00, 5.82it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 14:39:33,394[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-20 14:39:33,396[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 14:39:33,396[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 14:39:33,396[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 14:39:34,288[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 14:39:39,117[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 14:39:59,036[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 14:39:59,037[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.48179290369189476[0m
|
| 15 |
+
[[36m2026-07-20 14:39:59,045[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 14:39:59,046[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 14:39:59,061[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 14:39:59,063[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 4.353260441808186e-19[0m
|
| 19 |
+
[[36m2026-07-20 14:39:59,553[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 14:40:03,411[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.8301806640625[0m
|
| 23 |
+
[[36m2026-07-20 14:40:03,947[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 14:40:17,922[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.6960141660595527[0m
|
| 27 |
+
[[36m2026-07-20 14:40:18,587[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-20 14:40:23,882[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-20 14:40:35,423[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-20 14:40:40,160[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-20 14:40:56,205[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-20 14:40:56,598[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-20 14:40:57,533[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-20 14:40:59,044[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-20 14:40:59,322[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-20 14:41:00,970[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-20 14:41:00,970[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-20 14:41:00,970[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-20 14:41:01,226[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-20 14:41:01,976[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 14:41:03,452[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-20 14:41:03,714[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-20 14:41:05,505[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-20 14:41:05,505[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-20 14:41:05,505[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-20 14:41:05,505[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-20 14:41:05,506[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5841219734973736[0m
|
| 68 |
+
[[36m2026-07-20 14:41:06,082[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-20 14:41:06,098[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.04it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 11.18it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:01, 5.43it/s]
|
| 75 |
54%|█████▍ | 7/13 [00:01<00:01, 4.59it/s]
|
| 76 |
62%|██████▏ | 8/13 [00:01<00:01, 4.25it/s]
|
| 77 |
69%|██████▉ | 9/13 [00:01<00:00, 4.06it/s]
|
| 78 |
77%|███████▋ | 10/13 [00:02<00:00, 3.88it/s]
|
| 79 |
85%|████████▍ | 11/13 [00:02<00:00, 3.75it/s]
|
| 80 |
92%|█████████▏| 12/13 [00:02<00:00, 3.72it/s]
|
| 81 |
+
|
| 82 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 83 |
8%|▊ | 1/13 [00:00<00:01, 7.19it/s]
|
| 84 |
15%|█▌ | 2/13 [00:00<00:02, 4.37it/s]
|
| 85 |
23%|██▎ | 3/13 [00:00<00:02, 3.82it/s]
|
| 86 |
31%|███ | 4/13 [00:01<00:02, 3.59it/s]
|
| 87 |
38%|███▊ | 5/13 [00:01<00:02, 3.74it/s]
|
| 88 |
54%|█████▍ | 7/13 [00:01<00:01, 5.97it/s]
|
| 89 |
69%|██████▉ | 9/13 [00:01<00:00, 7.49it/s]
|
| 90 |
85%|████████▍ | 11/13 [00:01<00:00, 7.30it/s]
|
| 91 |
92%|█████████▏| 12/13 [00:02<00:00, 5.82it/s]
|
| 92 |
+
[[36m2026-07-20 14:41:11,266[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 93 |
+
[[36m2026-07-20 14:41:11,280[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 94 |
+
[[36m2026-07-20 14:41:11,280[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.12939602775738[0m
|
| 95 |
+
[[36m2026-07-20 14:41:11,668[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 96 |
+
|
| 97 |
+
Evaluated 400 examples
|
| 98 |
+
[[36m2026-07-20 14:41:18,947[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.5493364170346915[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.5_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.56it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.23it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.77it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.02it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.23it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 12.76it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.74it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.50it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 12.85it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 12.62it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 12.91it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 12.96it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:58:38,415[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:58:38,416[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:58:38,416[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:58:38,416[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:58:39,059[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:58:42,504[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:58:56,840[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:58:56,841[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4827526607798302[0m
|
| 15 |
+
[[36m2026-07-22 16:58:56,845[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:58:56,845[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:58:56,853[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:58:56,853[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 8.513756583610105e-19[0m
|
| 19 |
+
[[36m2026-07-22 16:58:57,247[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:59:00,136[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.82377197265625[0m
|
| 23 |
+
[[36m2026-07-22 16:59:00,401[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:59:04,847[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.6658492819742965[0m
|
| 27 |
+
[[36m2026-07-22 16:59:05,117[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:59:08,141[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:59:12,206[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:59:15,274[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:59:29,237[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:59:29,671[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:59:30,448[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:59:32,235[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:59:32,470[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:59:33,217[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:59:33,217[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:59:33,217[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:59:33,448[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:59:34,159[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:59:35,696[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:59:35,953[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:59:36,849[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:59:36,849[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:59:36,849[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:59:36,849[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:59:36,850[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5816914374335537[0m
|
| 68 |
+
[[36m2026-07-22 16:59:37,396[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:59:37,404[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.56it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.23it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.77it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.02it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.23it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 12.76it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.74it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.50it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 12.85it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 12.62it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 12.91it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 12.96it/s]
|
| 86 |
+
[[36m2026-07-22 16:59:39,354[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:59:39,361[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:59:39,361[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.11724806557999[0m
|
| 89 |
+
[[36m2026-07-22 16:59:39,635[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:59:40,696[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.5340673801514764[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.5_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 12.92it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.10it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 11.37it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 11.48it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 11.74it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:01<00:00, 11.48it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 11.29it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 11.14it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 11.42it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 11.47it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 11.48it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:01<00:00, 11.60it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[[36m2026-07-21 07:48:01,700[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha2_epoch10[0m
|
| 2 |
+
[[36m2026-07-21 07:48:01,702[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 3 |
+
[[36m2026-07-21 07:48:01,702[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 4 |
+
[[36m2026-07-21 07:48:01,702[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 5 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 6 |
+
[[36m2026-07-21 07:48:01,773[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 7 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 8 |
+
[[36m2026-07-21 07:48:01,774[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 9 |
+
[[36m2026-07-21 07:48:01,784[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 10 |
+
|
| 11 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-21 07:48:05,060[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 14 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 15 |
+
[[36m2026-07-21 07:48:05,062[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 16 |
+
[[36m2026-07-21 07:48:05,067[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 17 |
+
|
| 18 |
+
Evaluated 400 examples
|
| 19 |
+
[[36m2026-07-21 07:48:18,282[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 20 |
+
[[36m2026-07-21 07:48:18,283[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.48285022639509834[0m
|
| 21 |
+
[[36m2026-07-21 07:48:18,291[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 22 |
+
[[36m2026-07-21 07:48:18,292[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 23 |
+
[[36m2026-07-21 07:48:18,307[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 24 |
+
[[36m2026-07-21 07:48:18,309[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.11924339764853e-19[0m
|
| 25 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 26 |
+
[[36m2026-07-21 07:48:18,356[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 27 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 28 |
+
[[36m2026-07-21 07:48:18,359[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 29 |
+
[[36m2026-07-21 07:48:18,364[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 30 |
+
|
| 31 |
+
Evaluated 400 examples
|
| 32 |
+
[[36m2026-07-21 07:48:21,041[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.83384033203125[0m
|
| 33 |
+
Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub
|
| 34 |
+
[[36m2026-07-21 07:48:21,086[0m][[34mdatasets.load[0m][[33mWARNING[0m] - Using the latest cached version of the dataset since locuslab/TOFU couldn't be found on the Hugging Face Hub[0m
|
| 35 |
+
Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).
|
| 36 |
+
[[36m2026-07-21 07:48:21,087[0m][[34mdatasets.packaged_modules.cache.cache[0m][[33mWARNING[0m] - Found the latest cached dataset configuration 'forget10_perturbed' at /workspace/hf_cache_ou/locuslab___tofu/forget10_perturbed/0.0.0/324592d84ae4f482ac7249b9285c2ecdb53e3a68 (last modified on Mon Jul 20 13:17:15 2026).[0m
|
| 37 |
+
[[36m2026-07-21 07:48:21,093[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 38 |
+
|
| 39 |
+
Evaluated 400 examples
|
| 40 |
+
[[36m2026-07-21 07:48:29,820[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.6996836195457175[0m
|
| 41 |
+
[[36m2026-07-21 07:48:30,456[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 42 |
+
|
| 43 |
+
Evaluated 400 examples
|
| 44 |
+
[[36m2026-07-21 07:48:33,321[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 45 |
+
|
| 46 |
+
Evaluated 400 examples
|
| 47 |
+
[[36m2026-07-21 07:48:40,996[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 400 examples
|
| 50 |
+
[[36m2026-07-21 07:48:43,903[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 51 |
+
|
| 52 |
+
Evaluated 400 examples
|
| 53 |
+
[[36m2026-07-21 07:48:56,797[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 54 |
+
[[36m2026-07-21 07:48:57,165[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 55 |
+
|
| 56 |
+
Evaluated 100 examples
|
| 57 |
+
[[36m2026-07-21 07:48:57,927[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 58 |
+
|
| 59 |
+
Evaluated 100 examples
|
| 60 |
+
[[36m2026-07-21 07:48:59,435[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 61 |
+
[[36m2026-07-21 07:48:59,691[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 62 |
+
|
| 63 |
+
Evaluated 100 examples
|
| 64 |
+
[[36m2026-07-21 07:49:01,286[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-21 07:49:01,286[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 66 |
+
[[36m2026-07-21 07:49:01,286[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 67 |
+
[[36m2026-07-21 07:49:01,539[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 68 |
+
|
| 69 |
+
Evaluated 117 examples
|
| 70 |
+
[[36m2026-07-21 07:49:02,255[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 71 |
+
|
| 72 |
+
Evaluated 117 examples
|
| 73 |
+
[[36m2026-07-21 07:49:03,735[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 74 |
+
[[36m2026-07-21 07:49:03,982[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 75 |
+
|
| 76 |
+
Evaluated 117 examples
|
| 77 |
+
[[36m2026-07-21 07:49:05,675[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 78 |
+
[[36m2026-07-21 07:49:05,675[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 79 |
+
[[36m2026-07-21 07:49:05,675[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 80 |
+
[[36m2026-07-21 07:49:05,676[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 81 |
+
[[36m2026-07-21 07:49:05,676[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5864416017123132[0m
|
| 82 |
+
[[36m2026-07-21 07:49:06,278[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 83 |
+
[[36m2026-07-21 07:49:06,294[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 84 |
+
|
| 85 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 86 |
15%|█▌ | 2/13 [00:00<00:00, 12.92it/s]
|
| 87 |
31%|███ | 4/13 [00:00<00:00, 11.10it/s]
|
| 88 |
46%|████▌ | 6/13 [00:00<00:00, 11.37it/s]
|
| 89 |
62%|██████▏ | 8/13 [00:00<00:00, 11.48it/s]
|
| 90 |
77%|███████▋ | 10/13 [00:00<00:00, 11.74it/s]
|
| 91 |
92%|█████████▏| 12/13 [00:01<00:00, 11.48it/s]
|
| 92 |
+
|
| 93 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 94 |
15%|█▌ | 2/13 [00:00<00:00, 11.29it/s]
|
| 95 |
31%|███ | 4/13 [00:00<00:00, 11.14it/s]
|
| 96 |
46%|████▌ | 6/13 [00:00<00:00, 11.42it/s]
|
| 97 |
62%|██████▏ | 8/13 [00:00<00:00, 11.47it/s]
|
| 98 |
77%|███████▋ | 10/13 [00:00<00:00, 11.48it/s]
|
| 99 |
92%|█████████▏| 12/13 [00:01<00:00, 11.60it/s]
|
| 100 |
+
[[36m2026-07-21 07:49:08,475[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 101 |
+
[[36m2026-07-21 07:49:08,489[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 102 |
+
[[36m2026-07-21 07:49:08,489[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.19013583864424[0m
|
| 103 |
+
[[36m2026-07-21 07:49:08,793[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 104 |
+
|
| 105 |
+
Evaluated 400 examples
|
| 106 |
+
[[36m2026-07-21 07:49:09,982[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.5507798576893507[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.5_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.45it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.20it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.75it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.08it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.54it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.35it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.29it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.14it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.42it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.51it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.60it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.67it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:52:01,322[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:52:01,323[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:52:01,323[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:52:01,323[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:52:02,223[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:52:05,842[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:52:20,225[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:52:20,226[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.483027025002518[0m
|
| 15 |
+
[[36m2026-07-22 16:52:20,230[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:52:20,231[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:52:20,238[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:52:20,238[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 8.513756583610105e-19[0m
|
| 19 |
+
[[36m2026-07-22 16:52:20,696[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:52:23,565[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.8267724609375[0m
|
| 23 |
+
[[36m2026-07-22 16:52:23,871[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:52:28,452[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.6644333843947845[0m
|
| 27 |
+
[[36m2026-07-22 16:52:28,763[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:52:31,841[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:52:35,994[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:52:39,162[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:52:53,185[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:52:53,593[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:52:54,431[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:52:56,001[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:52:56,312[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:52:56,992[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:52:56,992[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:52:56,993[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:52:57,295[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:52:58,095[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:52:59,652[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:52:59,966[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:53:00,936[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:53:00,936[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:53:00,936[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:53:00,936[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:53:00,936[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5814537991536521[0m
|
| 68 |
+
[[36m2026-07-22 16:53:01,578[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:53:01,585[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.45it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.20it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.75it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.08it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.54it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.35it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.29it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.14it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.42it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.51it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.60it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.67it/s]
|
| 86 |
+
[[36m2026-07-22 16:53:03,463[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:53:03,471[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:53:03,471[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.18507418773703[0m
|
| 89 |
+
[[36m2026-07-22 16:53:03,795[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:53:04,837[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.5437310727272766[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.5_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.23it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.97it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.62it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.97it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.34it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 14.05it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.75it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.59it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 14.01it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 14.04it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 14.09it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 14.21it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:02:00,787[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha5_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 17:02:00,788[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:02:00,788[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:02:00,788[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:02:01,440[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:02:04,989[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:02:19,351[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:02:19,351[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.48170726420692106[0m
|
| 15 |
+
[[36m2026-07-22 17:02:19,356[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:02:19,356[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:02:19,364[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:02:19,365[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.11924339764853e-19[0m
|
| 19 |
+
[[36m2026-07-22 17:02:19,720[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:02:22,568[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.83840087890625[0m
|
| 23 |
+
[[36m2026-07-22 17:02:22,816[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:02:27,569[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.7016515632062874[0m
|
| 27 |
+
[[36m2026-07-22 17:02:27,822[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:02:30,823[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:02:35,093[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:02:38,185[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:02:52,184[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:02:52,537[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:02:53,295[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:02:54,859[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:02:55,105[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:02:55,874[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:02:55,874[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:02:55,874[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:02:56,120[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:02:56,872[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:02:58,429[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:02:58,717[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:02:59,697[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:02:59,697[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:02:59,697[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:02:59,698[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:02:59,698[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5860929391683252[0m
|
| 68 |
+
[[36m2026-07-22 17:03:00,333[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:03:00,342[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.23it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.97it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.62it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.97it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.34it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 14.05it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.75it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.59it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 14.01it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 14.04it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 14.09it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 14.21it/s]
|
| 86 |
+
[[36m2026-07-22 17:03:02,131[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:03:02,138[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:03:02,138[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.28630720588184[0m
|
| 89 |
+
[[36m2026-07-22 17:03:02,423[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:03:03,412[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.5558251547191049[0m
|
logs/eval_IdkDPO_lr2e-05_beta0.5_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.62it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.16it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.57it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.75it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.05it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 12.75it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.57it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.53it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 12.98it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.05it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.21it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.36it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:00:19,470[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:00:19,471[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:00:19,471[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:00:19,471[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr2e-05_beta0.5_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:00:20,146[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:00:23,678[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:00:37,990[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:00:37,991[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4819710448264702[0m
|
| 15 |
+
[[36m2026-07-22 17:00:37,995[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:00:37,995[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:00:38,003[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:00:38,003[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 4.353260441808186e-19[0m
|
| 19 |
+
[[36m2026-07-22 17:00:38,350[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:00:41,223[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.832109375[0m
|
| 23 |
+
[[36m2026-07-22 17:00:41,526[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:00:46,229[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.6830541373583775[0m
|
| 27 |
+
[[36m2026-07-22 17:00:46,495[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:00:49,524[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:00:53,858[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:00:56,991[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:01:11,000[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:01:11,401[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:01:12,265[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:01:13,833[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:01:14,081[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:01:14,728[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:01:14,728[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:01:14,728[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:01:14,956[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:01:15,672[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:01:17,224[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:01:17,465[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:01:18,484[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:01:18,484[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:01:18,484[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:01:18,484[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:01:18,484[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5826105713484052[0m
|
| 68 |
+
[[36m2026-07-22 17:01:19,074[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:01:19,082[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.62it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.16it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.57it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.75it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.05it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 12.75it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.57it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.53it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 12.98it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.05it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.21it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.36it/s]
|
| 86 |
+
[[36m2026-07-22 17:01:21,018[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:01:21,025[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:01:21,025[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.28833186624469[0m
|
| 89 |
+
[[36m2026-07-22 17:01:21,296[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:01:22,328[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.5468642874980272[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.05_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.98it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.69it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.27it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.59it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.98it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.68it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.36it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.24it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.64it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.61it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.66it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.75it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:53:32,815[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:53:32,816[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:53:32,816[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:53:32,816[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:53:34,266[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:53:37,851[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:53:52,260[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:53:52,260[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6690860558012102[0m
|
| 15 |
+
[[36m2026-07-22 16:53:52,264[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:53:52,265[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:53:52,272[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:53:52,273[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.0973970510505411e-09[0m
|
| 19 |
+
[[36m2026-07-22 16:53:52,679[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:53:55,531[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.0037765012102317997[0m
|
| 23 |
+
[[36m2026-07-22 16:53:55,852[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:54:00,773[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.12740131381455744[0m
|
| 27 |
+
[[36m2026-07-22 16:54:01,080[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:54:04,157[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:54:08,441[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:54:11,581[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:54:25,615[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:54:26,015[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:54:26,919[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:54:28,506[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:54:28,807[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:54:29,622[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:54:29,622[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:54:29,622[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:54:29,948[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:54:30,733[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:54:32,302[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:54:32,623[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:54:33,632[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:54:33,632[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:54:33,632[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:54:33,632[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:54:33,633[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5230881885397125[0m
|
| 68 |
+
[[36m2026-07-22 16:54:34,282[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:54:34,289[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.98it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.69it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.27it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.59it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.98it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.68it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.36it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.24it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.64it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.61it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.66it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.75it/s]
|
| 86 |
+
[[36m2026-07-22 16:54:36,131[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:54:36,138[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:54:36,138[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 61.641797078554106[0m
|
| 89 |
+
[[36m2026-07-22 16:54:36,477[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:54:37,614[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.04009040604588732[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.05_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.91it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.84it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.39it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.79it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.16it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.87it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.58it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.09it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.59it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.61it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.58it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.72it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:03:41,826[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:03:41,827[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:03:41,827[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:03:41,827[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:03:42,558[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:03:46,130[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:04:00,507[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:04:00,508[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6773695702553476[0m
|
| 15 |
+
[[36m2026-07-22 17:04:00,512[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:04:00,513[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:04:00,519[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:04:00,520[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.7473118094396964e-09[0m
|
| 19 |
+
[[36m2026-07-22 17:04:00,871[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:04:03,730[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.0051828776131151244[0m
|
| 23 |
+
[[36m2026-07-22 17:04:03,990[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:04:09,337[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.12324596348940699[0m
|
| 27 |
+
[[36m2026-07-22 17:04:09,579[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:04:12,592[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:04:17,147[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:04:20,212[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:04:34,207[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:04:34,631[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:04:35,392[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:04:36,994[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:04:37,243[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:04:38,271[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:04:38,271[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:04:38,271[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:04:38,587[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:04:39,335[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:04:40,911[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:04:41,172[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:04:42,166[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:04:42,166[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:04:42,166[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:04:42,166[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:04:42,167[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.4885568179815585[0m
|
| 68 |
+
[[36m2026-07-22 17:04:42,706[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:04:42,713[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.91it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.84it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.39it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.79it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.16it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.87it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.58it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.09it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.59it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.61it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.58it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.72it/s]
|
| 86 |
+
[[36m2026-07-22 17:04:44,547[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:04:44,554[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:04:44,554[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 61.57802027712288[0m
|
| 89 |
+
[[36m2026-07-22 17:04:44,850[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:04:45,956[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.039749124221918605[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.05_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
8%|▊ | 1/13 [00:00<00:02, 5.81it/s]
|
| 2 |
15%|█▌ | 2/13 [00:00<00:02, 3.81it/s]
|
| 3 |
23%|██▎ | 3/13 [00:00<00:02, 3.41it/s]
|
| 4 |
31%|███ | 4/13 [00:01<00:02, 3.40it/s]
|
| 5 |
38%|███▊ | 5/13 [00:01<00:02, 3.46it/s]
|
| 6 |
46%|████▌ | 6/13 [00:01<00:02, 3.46it/s]
|
| 7 |
54%|█████▍ | 7/13 [00:01<00:01, 3.51it/s]
|
| 8 |
62%|██████▏ | 8/13 [00:02<00:01, 3.59it/s]
|
| 9 |
69%|██████▉ | 9/13 [00:02<00:01, 3.61it/s]
|
| 10 |
77%|███████▋ | 10/13 [00:02<00:00, 3.57it/s]
|
| 11 |
85%|████████▍ | 11/13 [00:03<00:00, 3.49it/s]
|
| 12 |
92%|█████████▏| 12/13 [00:03<00:00, 3.68it/s]
|
|
|
|
| 13 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 14 |
15%|█▌ | 2/13 [00:00<00:00, 12.67it/s]
|
| 15 |
31%|███ | 4/13 [00:00<00:00, 11.89it/s]
|
| 16 |
46%|████▌ | 6/13 [00:00<00:00, 9.04it/s]
|
| 17 |
54%|█████▍ | 7/13 [00:00<00:00, 8.51it/s]
|
| 18 |
62%|██████▏ | 8/13 [00:00<00:00, 7.89it/s]
|
| 19 |
69%|██████▉ | 9/13 [00:01<00:00, 7.43it/s]
|
| 20 |
77%|███████▋ | 10/13 [00:01<00:00, 7.39it/s]
|
| 21 |
85%|████████▍ | 11/13 [00:01<00:00, 7.21it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 14:41:40,455[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha2_epoch10[0m
|
| 4 |
+
[[36m2026-07-20 14:41:40,457[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 14:41:40,457[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 14:41:40,457[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 14:41:41,206[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 14:41:46,095[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 14:42:07,508[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 14:42:07,509[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6693142101444849[0m
|
| 15 |
+
[[36m2026-07-20 14:42:07,517[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 14:42:07,519[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 14:42:07,534[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 14:42:07,536[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 4.2606809322061477e-10[0m
|
| 19 |
+
[[36m2026-07-20 14:42:07,948[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 14:42:12,959[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.006944010932929814[0m
|
| 23 |
+
[[36m2026-07-20 14:42:13,367[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 14:42:28,792[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.13328360608257625[0m
|
| 27 |
+
[[36m2026-07-20 14:42:29,199[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-20 14:42:33,651[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-20 14:42:43,836[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-20 14:42:46,765[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-20 14:43:05,675[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-20 14:43:06,040[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-20 14:43:07,405[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-20 14:43:10,499[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-20 14:43:10,791[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-20 14:43:13,002[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-20 14:43:13,002[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-20 14:43:13,002[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-20 14:43:13,306[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-20 14:43:14,492[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 14:43:17,375[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-20 14:43:17,726[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-20 14:43:20,648[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-20 14:43:20,648[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-20 14:43:20,649[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-20 14:43:20,649[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-20 14:43:20,650[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5558230070579687[0m
|
| 68 |
+
[[36m2026-07-20 14:43:21,288[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-20 14:43:21,303[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
8%|▊ | 1/13 [00:00<00:02, 5.81it/s]
|
| 73 |
15%|█▌ | 2/13 [00:00<00:02, 3.81it/s]
|
| 74 |
23%|██▎ | 3/13 [00:00<00:02, 3.41it/s]
|
| 75 |
31%|███ | 4/13 [00:01<00:02, 3.40it/s]
|
| 76 |
38%|███▊ | 5/13 [00:01<00:02, 3.46it/s]
|
| 77 |
46%|████▌ | 6/13 [00:01<00:02, 3.46it/s]
|
| 78 |
54%|█████▍ | 7/13 [00:01<00:01, 3.51it/s]
|
| 79 |
62%|██████▏ | 8/13 [00:02<00:01, 3.59it/s]
|
| 80 |
69%|██████▉ | 9/13 [00:02<00:01, 3.61it/s]
|
| 81 |
77%|███████▋ | 10/13 [00:02<00:00, 3.57it/s]
|
| 82 |
85%|████████▍ | 11/13 [00:03<00:00, 3.49it/s]
|
| 83 |
92%|█████████▏| 12/13 [00:03<00:00, 3.68it/s]
|
| 84 |
+
|
| 85 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 86 |
15%|█▌ | 2/13 [00:00<00:00, 12.67it/s]
|
| 87 |
31%|███ | 4/13 [00:00<00:00, 11.89it/s]
|
| 88 |
46%|████▌ | 6/13 [00:00<00:00, 9.04it/s]
|
| 89 |
54%|█████▍ | 7/13 [00:00<00:00, 8.51it/s]
|
| 90 |
62%|██████▏ | 8/13 [00:00<00:00, 7.89it/s]
|
| 91 |
69%|██████▉ | 9/13 [00:01<00:00, 7.43it/s]
|
| 92 |
77%|███████▋ | 10/13 [00:01<00:00, 7.39it/s]
|
| 93 |
85%|████████▍ | 11/13 [00:01<00:00, 7.21it/s]
|
| 94 |
+
[[36m2026-07-20 14:43:26,222[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 95 |
+
[[36m2026-07-20 14:43:26,236[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 96 |
+
[[36m2026-07-20 14:43:26,236[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 61.11133606347532[0m
|
| 97 |
+
[[36m2026-07-20 14:43:26,540[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 98 |
+
|
| 99 |
+
Evaluated 400 examples
|
| 100 |
+
[[36m2026-07-20 14:43:30,618[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.04379600170478851[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.05_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.72it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.12it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.53it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 12.76it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.21it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 12.97it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.96it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.81it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.15it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.05it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.11it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.22it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:02:02,895[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:02:02,896[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:02:02,896[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:02:02,896[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:02:03,512[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:02:07,000[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:02:21,352[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:02:21,352[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6766758383301827[0m
|
| 15 |
+
[[36m2026-07-22 17:02:21,357[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:02:21,357[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:02:21,364[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:02:21,365[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.652551353350331e-08[0m
|
| 19 |
+
[[36m2026-07-22 17:02:21,743[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:02:24,602[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.009342645797878503[0m
|
| 23 |
+
[[36m2026-07-22 17:02:24,839[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:02:30,120[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.1423345148832089[0m
|
| 27 |
+
[[36m2026-07-22 17:02:30,384[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:02:33,404[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:02:38,054[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:02:41,121[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:02:55,119[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:02:55,453[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:02:56,233[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:02:57,839[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:02:58,088[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:02:58,918[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:02:58,918[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:02:58,918[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:02:59,228[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:03:00,048[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:03:01,592[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:03:01,851[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:03:02,920[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:03:02,920[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:03:02,920[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:03:02,920[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:03:02,921[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5342224046244235[0m
|
| 68 |
+
[[36m2026-07-22 17:03:03,441[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:03:03,448[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.72it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.12it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.53it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 12.76it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.21it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 12.97it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.96it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.81it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.15it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.05it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.11it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.22it/s]
|
| 86 |
+
[[36m2026-07-22 17:03:05,377[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:03:05,383[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:03:05,384[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 60.83193293339565[0m
|
| 89 |
+
[[36m2026-07-22 17:03:05,641[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:03:06,822[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.04338479703493455[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.05_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 14.00it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.69it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.28it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.62it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.92it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.61it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.29it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.11it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.56it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.59it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.57it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.65it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:55:14,945[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha5_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:55:14,946[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:55:14,946[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:55:14,946[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:55:15,731[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:55:19,208[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:55:33,628[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:55:33,628[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6560693064868462[0m
|
| 15 |
+
[[36m2026-07-22 16:55:33,633[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:55:33,633[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:55:33,640[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:55:33,641[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.001024214776236788[0m
|
| 19 |
+
[[36m2026-07-22 16:55:34,036[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:55:36,915[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.016583357490599155[0m
|
| 23 |
+
[[36m2026-07-22 16:55:37,191[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:55:42,026[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.14422862303880127[0m
|
| 27 |
+
[[36m2026-07-22 16:55:42,271[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:55:45,289[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:55:49,868[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:55:52,945[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:56:06,995[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:56:07,337[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:56:08,108[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:56:09,697[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:56:09,957[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:56:10,686[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:56:10,686[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:56:10,686[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:56:10,932[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:56:11,661[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:56:13,245[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:56:13,488[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:56:14,343[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:56:14,343[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:56:14,343[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:56:14,344[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:56:14,344[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5654775896008905[0m
|
| 68 |
+
[[36m2026-07-22 16:56:14,948[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:56:14,955[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 14.00it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.69it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.28it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.62it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.92it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.61it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.29it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.11it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.56it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.59it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.57it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.65it/s]
|
| 86 |
+
[[36m2026-07-22 16:56:16,807[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:56:16,813[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:56:16,813[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 57.29788826996046[0m
|
| 89 |
+
[[36m2026-07-22 16:56:17,076[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:56:18,215[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.047392134456931884[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.05_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.78it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.31it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.04it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.41it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.73it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.52it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.33it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.26it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.69it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.72it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.79it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.87it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:05:23,866[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:05:23,866[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:05:23,866[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:05:23,867[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.05_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:05:24,491[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:05:27,916[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:05:42,284[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:05:42,284[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6564655390404865[0m
|
| 15 |
+
[[36m2026-07-22 17:05:42,288[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:05:42,289[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:05:42,296[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:05:42,296[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.0007748993373290397[0m
|
| 19 |
+
[[36m2026-07-22 17:05:42,632[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:05:45,521[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.021908438429236413[0m
|
| 23 |
+
[[36m2026-07-22 17:05:45,766[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:05:50,347[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.14343149867927152[0m
|
| 27 |
+
[[36m2026-07-22 17:05:50,604[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:05:53,618[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:05:58,127[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:06:01,197[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:06:15,183[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:06:15,532[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:06:16,317[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:06:17,925[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:06:18,224[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:06:18,977[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:06:18,978[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:06:18,978[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:06:19,227[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:06:19,968[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:06:21,509[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:06:21,812[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:06:22,859[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:06:22,859[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:06:22,859[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:06:22,859[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:06:22,860[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5460970059002894[0m
|
| 68 |
+
[[36m2026-07-22 17:06:23,377[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:06:23,386[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.78it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.31it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.04it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.41it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.73it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.52it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.33it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.26it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.69it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.72it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.79it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.87it/s]
|
| 86 |
+
[[36m2026-07-22 17:06:25,235[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:06:25,242[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:06:25,242[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: 55.76977586106461[0m
|
| 89 |
+
[[36m2026-07-22 17:06:25,574[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:06:26,761[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.048888956826320544[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.1_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
8%|▊ | 1/13 [00:00<00:03, 3.18it/s]
|
| 2 |
15%|█▌ | 2/13 [00:00<00:03, 3.09it/s]
|
| 3 |
23%|██▎ | 3/13 [00:00<00:02, 4.20it/s]
|
| 4 |
38%|███▊ | 5/13 [00:00<00:01, 6.61it/s]
|
| 5 |
54%|█████▍ | 7/13 [00:01<00:00, 8.21it/s]
|
| 6 |
62%|██████▏ | 8/13 [00:01<00:00, 5.99it/s]
|
| 7 |
69%|██████▉ | 9/13 [00:01<00:00, 5.34it/s]
|
| 8 |
77%|███████▋ | 10/13 [00:01<00:00, 5.20it/s]
|
| 9 |
92%|█████████▏| 12/13 [00:02<00:00, 6.98it/s]
|
|
|
|
| 10 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 11 |
15%|█▌ | 2/13 [00:00<00:01, 6.07it/s]
|
| 12 |
23%|██▎ | 3/13 [00:00<00:02, 4.46it/s]
|
| 13 |
31%|███ | 4/13 [00:00<00:02, 3.83it/s]
|
| 14 |
46%|████▌ | 6/13 [00:01<00:01, 4.51it/s]
|
| 15 |
54%|█████▍ | 7/13 [00:01<00:01, 3.52it/s]
|
| 16 |
62%|██████▏ | 8/13 [00:02<00:01, 3.54it/s]
|
| 17 |
69%|██████▉ | 9/13 [00:02<00:00, 4.11it/s]
|
| 18 |
77%|███████▋ | 10/13 [00:02<00:00, 4.22it/s]
|
| 19 |
92%|█████████▏| 12/13 [00:02<00:00, 6.15it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 14:35:17,403[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-20 14:35:17,404[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 14:35:17,404[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 14:35:17,404[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 14:35:18,044[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 14:35:21,625[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 14:35:46,102[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 14:35:46,103[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6251863838312173[0m
|
| 15 |
+
[[36m2026-07-20 14:35:46,112[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 14:35:46,113[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 14:35:46,130[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 14:35:46,132[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.02424717195471133[0m
|
| 19 |
+
[[36m2026-07-20 14:35:46,469[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 14:35:51,641[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.13517303466796876[0m
|
| 23 |
+
[[36m2026-07-20 14:35:51,942[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 14:36:04,568[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.24437904088907494[0m
|
| 27 |
+
[[36m2026-07-20 14:36:04,933[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-20 14:36:08,999[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-20 14:36:18,992[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-20 14:36:23,554[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-20 14:36:44,206[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-20 14:36:44,613[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-20 14:36:46,164[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-20 14:36:49,330[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-20 14:36:49,625[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-20 14:36:52,028[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-20 14:36:52,029[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-20 14:36:52,029[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-20 14:36:52,302[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-20 14:36:53,505[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 14:36:56,423[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-20 14:36:56,694[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-20 14:36:59,339[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-20 14:36:59,339[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-20 14:36:59,339[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-20 14:36:59,340[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-20 14:36:59,341[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5589265062187224[0m
|
| 68 |
+
[[36m2026-07-20 14:36:59,937[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-20 14:36:59,952[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
8%|▊ | 1/13 [00:00<00:03, 3.18it/s]
|
| 73 |
15%|█▌ | 2/13 [00:00<00:03, 3.09it/s]
|
| 74 |
23%|██▎ | 3/13 [00:00<00:02, 4.20it/s]
|
| 75 |
38%|███▊ | 5/13 [00:00<00:01, 6.61it/s]
|
| 76 |
54%|█████▍ | 7/13 [00:01<00:00, 8.21it/s]
|
| 77 |
62%|██████▏ | 8/13 [00:01<00:00, 5.99it/s]
|
| 78 |
69%|██████▉ | 9/13 [00:01<00:00, 5.34it/s]
|
| 79 |
77%|███████▋ | 10/13 [00:01<00:00, 5.20it/s]
|
| 80 |
92%|█████████▏| 12/13 [00:02<00:00, 6.98it/s]
|
| 81 |
+
|
| 82 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 83 |
15%|█▌ | 2/13 [00:00<00:01, 6.07it/s]
|
| 84 |
23%|██▎ | 3/13 [00:00<00:02, 4.46it/s]
|
| 85 |
31%|███ | 4/13 [00:00<00:02, 3.83it/s]
|
| 86 |
46%|████▌ | 6/13 [00:01<00:01, 4.51it/s]
|
| 87 |
54%|█████▍ | 7/13 [00:01<00:01, 3.52it/s]
|
| 88 |
62%|██████▏ | 8/13 [00:02<00:01, 3.54it/s]
|
| 89 |
69%|██████▉ | 9/13 [00:02<00:00, 4.11it/s]
|
| 90 |
77%|███████▋ | 10/13 [00:02<00:00, 4.22it/s]
|
| 91 |
92%|█████████▏| 12/13 [00:02<00:00, 6.15it/s]
|
| 92 |
+
[[36m2026-07-20 14:37:04,655[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 93 |
+
[[36m2026-07-20 14:37:04,670[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 94 |
+
[[36m2026-07-20 14:37:04,670[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -14.290559006411584[0m
|
| 95 |
+
[[36m2026-07-20 14:37:04,961[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 96 |
+
|
| 97 |
+
Evaluated 400 examples
|
| 98 |
+
[[36m2026-07-20 14:37:08,014[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.07439229037631959[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.1_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.98it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.53it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.08it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.39it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.69it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.35it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.08it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.89it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.22it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.23it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.24it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.11it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:03:42,445[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:03:42,446[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:03:42,446[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:03:42,446[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:03:43,122[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:03:46,535[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:04:00,897[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:04:00,897[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6298913263610946[0m
|
| 15 |
+
[[36m2026-07-22 17:04:00,901[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:04:00,902[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:04:00,908[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:04:00,909[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.02985302150862578[0m
|
| 19 |
+
[[36m2026-07-22 17:04:01,248[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:04:04,109[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.14977981567382812[0m
|
| 23 |
+
[[36m2026-07-22 17:04:04,427[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:04:09,700[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.2525018851151177[0m
|
| 27 |
+
[[36m2026-07-22 17:04:09,961[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:04:12,981[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:04:17,305[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:04:20,438[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:04:34,425[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:04:34,816[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:04:35,577[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:04:37,172[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:04:37,433[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:04:38,155[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:04:38,156[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:04:38,156[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:04:38,397[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:04:39,126[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:04:40,702[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:04:40,939[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:04:41,808[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:04:41,808[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:04:41,808[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:04:41,808[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:04:41,808[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5397392532622398[0m
|
| 68 |
+
[[36m2026-07-22 17:04:42,449[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:04:42,456[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.98it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.53it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.08it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.39it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.69it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.35it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.08it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.89it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.22it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.23it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.24it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.11it/s]
|
| 86 |
+
[[36m2026-07-22 17:04:44,356[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:04:44,363[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:04:44,363[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -19.132534264277687[0m
|
| 89 |
+
[[36m2026-07-22 17:04:44,702[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:04:45,829[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.0740434781230234[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.1_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,76 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.39it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.52it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 11.77it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 11.90it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 12.14it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:01<00:00, 11.88it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 11.62it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 11.48it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 11.79it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 11.83it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 11.84it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:01<00:00, 11.96it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[[36m2026-07-21 07:49:22,342[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha2_epoch10[0m
|
| 2 |
+
[[36m2026-07-21 07:49:22,344[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 3 |
+
[[36m2026-07-21 07:49:22,344[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 4 |
+
[[36m2026-07-21 07:49:22,344[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 5 |
+
[[36m2026-07-21 07:49:22,992[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 6 |
+
|
| 7 |
+
Evaluated 400 examples
|
| 8 |
+
[[36m2026-07-21 07:49:26,482[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 9 |
+
|
| 10 |
+
Evaluated 400 examples
|
| 11 |
+
[[36m2026-07-21 07:49:39,585[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 12 |
+
[[36m2026-07-21 07:49:39,586[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6119044179566158[0m
|
| 13 |
+
[[36m2026-07-21 07:49:39,594[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 14 |
+
[[36m2026-07-21 07:49:39,595[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 15 |
+
[[36m2026-07-21 07:49:39,610[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 16 |
+
[[36m2026-07-21 07:49:39,612[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.001024214776236788[0m
|
| 17 |
+
[[36m2026-07-21 07:49:39,989[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 18 |
+
|
| 19 |
+
Evaluated 400 examples
|
| 20 |
+
[[36m2026-07-21 07:49:42,651[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.17952091217041016[0m
|
| 21 |
+
[[36m2026-07-21 07:49:42,927[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 22 |
+
|
| 23 |
+
Evaluated 400 examples
|
| 24 |
+
[[36m2026-07-21 07:49:52,339[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.270815563435448[0m
|
| 25 |
+
[[36m2026-07-21 07:49:52,702[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 26 |
+
|
| 27 |
+
Evaluated 400 examples
|
| 28 |
+
[[36m2026-07-21 07:49:55,580[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 29 |
+
|
| 30 |
+
Evaluated 400 examples
|
| 31 |
+
[[36m2026-07-21 07:50:03,746[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 32 |
+
|
| 33 |
+
Evaluated 400 examples
|
| 34 |
+
[[36m2026-07-21 07:50:06,646[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 35 |
+
|
| 36 |
+
Evaluated 400 examples
|
| 37 |
+
[[36m2026-07-21 07:50:19,422[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 38 |
+
[[36m2026-07-21 07:50:19,826[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 39 |
+
|
| 40 |
+
Evaluated 100 examples
|
| 41 |
+
[[36m2026-07-21 07:50:20,597[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 42 |
+
|
| 43 |
+
Evaluated 100 examples
|
| 44 |
+
[[36m2026-07-21 07:50:22,093[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 45 |
+
[[36m2026-07-21 07:50:22,347[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 46 |
+
|
| 47 |
+
Evaluated 100 examples
|
| 48 |
+
[[36m2026-07-21 07:50:23,718[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 49 |
+
[[36m2026-07-21 07:50:23,718[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 50 |
+
[[36m2026-07-21 07:50:23,718[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 51 |
+
[[36m2026-07-21 07:50:24,050[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 52 |
+
|
| 53 |
+
Evaluated 117 examples
|
| 54 |
+
[[36m2026-07-21 07:50:24,790[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 55 |
+
|
| 56 |
+
Evaluated 117 examples
|
| 57 |
+
[[36m2026-07-21 07:50:26,261[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 58 |
+
[[36m2026-07-21 07:50:26,587[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 59 |
+
|
| 60 |
+
Evaluated 117 examples
|
| 61 |
+
[[36m2026-07-21 07:50:28,229[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 62 |
+
[[36m2026-07-21 07:50:28,229[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 63 |
+
[[36m2026-07-21 07:50:28,229[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 64 |
+
[[36m2026-07-21 07:50:28,230[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 65 |
+
[[36m2026-07-21 07:50:28,231[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5532968844484002[0m
|
| 66 |
+
[[36m2026-07-21 07:50:28,784[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 67 |
+
[[36m2026-07-21 07:50:28,799[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 68 |
+
|
| 69 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 70 |
15%|█▌ | 2/13 [00:00<00:00, 13.39it/s]
|
| 71 |
31%|███ | 4/13 [00:00<00:00, 11.52it/s]
|
| 72 |
46%|████▌ | 6/13 [00:00<00:00, 11.77it/s]
|
| 73 |
62%|██████▏ | 8/13 [00:00<00:00, 11.90it/s]
|
| 74 |
77%|███████▋ | 10/13 [00:00<00:00, 12.14it/s]
|
| 75 |
92%|█████████▏| 12/13 [00:01<00:00, 11.88it/s]
|
| 76 |
+
|
| 77 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 78 |
15%|█▌ | 2/13 [00:00<00:00, 11.62it/s]
|
| 79 |
31%|███ | 4/13 [00:00<00:00, 11.48it/s]
|
| 80 |
46%|████▌ | 6/13 [00:00<00:00, 11.79it/s]
|
| 81 |
62%|██████▏ | 8/13 [00:00<00:00, 11.83it/s]
|
| 82 |
77%|███████▋ | 10/13 [00:00<00:00, 11.84it/s]
|
| 83 |
92%|█████████▏| 12/13 [00:01<00:00, 11.96it/s]
|
| 84 |
+
[[36m2026-07-21 07:50:30,911[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 85 |
+
[[36m2026-07-21 07:50:30,930[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 86 |
+
[[36m2026-07-21 07:50:30,931[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -36.79111578436701[0m
|
| 87 |
+
[[36m2026-07-21 07:50:31,243[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 88 |
+
|
| 89 |
+
Evaluated 400 examples
|
| 90 |
+
[[36m2026-07-21 07:50:32,523[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.08240799751904848[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.1_alpha2_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.89it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.57it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.22it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.60it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.97it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.42it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.34it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.66it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.58it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.69it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:06:56,163[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha2_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:06:56,163[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:06:56,163[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha2_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:06:56,163[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha2_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:06:56,922[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:07:01,165[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:07:15,554[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:07:15,555[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.6182819370651267[0m
|
| 15 |
+
[[36m2026-07-22 17:07:15,559[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:07:15,560[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:07:15,567[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:07:15,568[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 0.0017621899205092087[0m
|
| 19 |
+
[[36m2026-07-22 17:07:15,910[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:07:18,814[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.19928009033203126[0m
|
| 23 |
+
[[36m2026-07-22 17:07:19,049[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:07:23,493[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.2769266733660994[0m
|
| 27 |
+
[[36m2026-07-22 17:07:23,768[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:07:26,795[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:07:31,013[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:07:34,086[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:07:48,111[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:07:48,449[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:07:49,226[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:07:50,907[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:07:51,142[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:07:51,790[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:07:51,790[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:07:51,790[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:07:52,077[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:07:52,794[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:07:54,343[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:07:54,581[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:07:55,474[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:07:55,475[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:07:55,475[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:07:55,475[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:07:55,475[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5369199366782811[0m
|
| 68 |
+
[[36m2026-07-22 17:07:55,988[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:07:55,995[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.89it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.57it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.22it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.60it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.97it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.42it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.34it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.66it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.58it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.69it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.66it/s]
|
| 86 |
+
[[36m2026-07-22 17:07:57,840[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:07:57,847[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:07:57,847[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -41.999048402826595[0m
|
| 89 |
+
[[36m2026-07-22 17:07:58,103[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:07:59,280[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.08561120637238975[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.1_alpha5_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.92it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.43it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.97it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.27it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.63it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.33it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.06it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.91it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.28it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.28it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.36it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.49it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:05:29,681[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha5_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 17:05:29,682[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:05:29,682[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha5_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:05:29,682[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha5_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:05:30,420[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:05:33,861[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:05:48,213[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:05:48,213[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.593876697461511[0m
|
| 15 |
+
[[36m2026-07-22 17:05:48,218[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:05:48,218[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:05:48,225[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:05:48,225[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 4.455792684961036e-06[0m
|
| 19 |
+
[[36m2026-07-22 17:05:48,572[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:05:51,427[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.2642172241210938[0m
|
| 23 |
+
[[36m2026-07-22 17:05:51,712[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:05:56,228[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.3021807405137598[0m
|
| 27 |
+
[[36m2026-07-22 17:05:56,483[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:05:59,484[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:06:04,006[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:06:07,115[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:06:21,134[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:06:21,515[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:06:22,279[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:06:23,868[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:06:24,118[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:06:24,787[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:06:24,797[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:06:24,797[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:06:25,035[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:06:25,758[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:06:27,315[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:06:27,565[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:06:28,515[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:06:28,515[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:06:28,515[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:06:28,516[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:06:28,516[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.558210072885203[0m
|
| 68 |
+
[[36m2026-07-22 17:06:29,050[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:06:29,057[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.92it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.43it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.97it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.27it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.63it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.33it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.06it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.91it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.28it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.28it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.36it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.49it/s]
|
| 86 |
+
[[36m2026-07-22 17:06:30,941[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:06:30,948[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:06:30,948[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -64.31941040848423[0m
|
| 89 |
+
[[36m2026-07-22 17:06:31,249[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:06:32,341[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.09716642151573382[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.1_alpha5_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.37it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.31it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.93it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.18it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.48it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.05it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.42it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.12it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.46it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 12.98it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 12.54it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 12.50it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:56:58,666[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha5_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 16:56:58,666[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:56:58,666[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha5_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:56:58,666[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.1_alpha5_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:56:59,398[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:57:02,853[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:57:17,259[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:57:17,259[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5992297050026533[0m
|
| 15 |
+
[[36m2026-07-22 16:57:17,263[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:57:17,264[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:57:17,271[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:57:17,272[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 9.9886659393885e-07[0m
|
| 19 |
+
[[36m2026-07-22 16:57:17,627[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:57:20,538[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.2813580322265625[0m
|
| 23 |
+
[[36m2026-07-22 16:57:20,808[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:57:25,305[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.31014771568074245[0m
|
| 27 |
+
[[36m2026-07-22 16:57:25,621[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:57:28,661[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:57:33,104[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:57:36,211[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:57:50,233[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:57:50,590[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:57:51,402[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:57:53,058[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:57:53,313[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:57:53,951[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:57:53,951[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:57:53,951[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:57:54,431[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:57:55,190[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:57:56,743[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:57:57,056[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:57:58,015[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:57:58,015[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:57:58,015[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:57:58,015[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:57:58,015[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.543855549899428[0m
|
| 68 |
+
[[36m2026-07-22 16:57:58,590[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:57:58,597[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.37it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.31it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.93it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.18it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.48it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.05it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.42it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.12it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.46it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 12.98it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 12.54it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 12.50it/s]
|
| 86 |
+
[[36m2026-07-22 16:58:00,548[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:58:00,555[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:58:00,555[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -67.25010628377639[0m
|
| 89 |
+
[[36m2026-07-22 16:58:00,816[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:58:01,979[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.10046612315340754[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.5_alpha1_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.71it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.61it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 13.26it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.63it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 14.00it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.71it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 13.16it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 13.20it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.45it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.39it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.53it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.75it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:08:28,173[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha1_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 17:08:28,174[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:08:28,174[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha1_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:08:28,174[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha1_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:08:28,819[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:08:32,486[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:08:46,896[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:08:46,896[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5247954468604213[0m
|
| 15 |
+
[[36m2026-07-22 17:08:46,900[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:08:46,900[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:08:46,907[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:08:46,908[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 5.137999089404319e-16[0m
|
| 19 |
+
[[36m2026-07-22 17:08:47,369[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:08:50,274[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.6555126953125[0m
|
| 23 |
+
[[36m2026-07-22 17:08:50,512[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:08:55,139[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.5076679442064704[0m
|
| 27 |
+
[[36m2026-07-22 17:08:55,445[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:08:58,493[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:09:02,710[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:09:05,786[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:09:19,796[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:09:20,153[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:09:20,939[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:09:22,586[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:09:22,829[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:09:23,597[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:09:23,597[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:09:23,597[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:09:23,875[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:09:24,585[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:09:26,130[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:09:26,444[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:09:27,401[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:09:27,401[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:09:27,401[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:09:27,401[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:09:27,402[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5535670888647204[0m
|
| 68 |
+
[[36m2026-07-22 17:09:28,071[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:09:28,082[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.71it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.61it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 13.26it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.63it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 14.00it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.71it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 13.16it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 13.20it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.45it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.39it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.53it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.75it/s]
|
| 86 |
+
[[36m2026-07-22 17:09:29,931[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:09:29,938[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:09:29,938[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -97.83968737659254[0m
|
| 89 |
+
[[36m2026-07-22 17:09:30,425[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:09:31,534[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.2770861231130759[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.5_alpha1_epoch5.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.84it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.43it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.96it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.22it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.57it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.25it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.96it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.48it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 12.89it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 12.98it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.01it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.13it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 17:07:10,039[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha1_epoch5[0m
|
| 4 |
+
[[36m2026-07-22 17:07:10,040[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 17:07:10,040[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha1_epoch5/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 17:07:10,040[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha1_epoch5/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 17:07:10,762[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 17:07:14,225[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 17:07:28,572[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 17:07:28,572[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5421558270185806[0m
|
| 15 |
+
[[36m2026-07-22 17:07:28,576[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 17:07:28,577[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 17:07:28,584[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 17:07:28,584[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 3.277026375806354e-14[0m
|
| 19 |
+
[[36m2026-07-22 17:07:28,929[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 17:07:31,789[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.618076171875[0m
|
| 23 |
+
[[36m2026-07-22 17:07:32,031[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 17:07:36,826[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.4743630049698625[0m
|
| 27 |
+
[[36m2026-07-22 17:07:37,060[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 17:07:40,064[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 17:07:44,685[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 17:07:47,757[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 17:08:01,768[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 17:08:02,136[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 17:08:02,935[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 17:08:04,522[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 17:08:04,786[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 17:08:05,601[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 17:08:05,601[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 17:08:05,601[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 17:08:05,901[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 17:08:06,727[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 17:08:08,307[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 17:08:08,665[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 17:08:09,721[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 17:08:09,721[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 17:08:09,721[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 17:08:09,721[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 17:08:09,722[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5456160393493406[0m
|
| 68 |
+
[[36m2026-07-22 17:08:10,967[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 17:08:10,974[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.84it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.43it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.96it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.22it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.57it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.25it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.96it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.48it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 12.89it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 12.98it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.01it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.13it/s]
|
| 86 |
+
[[36m2026-07-22 17:08:12,888[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 17:08:12,896[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 17:08:12,896[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -97.50359375635178[0m
|
| 89 |
+
[[36m2026-07-22 17:08:13,239[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 17:08:14,368[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.2586622973210397[0m
|
logs/eval_IdkDPO_lr5e-05_beta0.5_alpha2_epoch10.log
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.62it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 12.18it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 12.83it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 13.17it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 13.23it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 13.16it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 12.33it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 12.33it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 13.03it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 13.19it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 13.24it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:00<00:00, 13.18it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-22 16:58:37,970[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha2_epoch10[0m
|
| 4 |
+
[[36m2026-07-22 16:58:37,971[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-22 16:58:37,971[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha2_epoch10/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-22 16:58:37,971[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/IdkDPO_lr5e-05_beta0.5_alpha2_epoch10/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-22 16:58:38,622[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-22 16:58:42,146[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-22 16:58:56,546[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-22 16:58:56,547[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.5249729961214809[0m
|
| 15 |
+
[[36m2026-07-22 16:58:56,551[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-22 16:58:56,552[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-22 16:58:56,559[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-22 16:58:56,559[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.02724452984523e-14[0m
|
| 19 |
+
[[36m2026-07-22 16:58:56,936[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-22 16:58:59,831[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.662333984375[0m
|
| 23 |
+
[[36m2026-07-22 16:59:00,120[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-22 16:59:04,961[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.523768418130635[0m
|
| 27 |
+
[[36m2026-07-22 16:59:05,217[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 28 |
+
|
| 29 |
+
Evaluated 400 examples
|
| 30 |
+
[[36m2026-07-22 16:59:08,242[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 31 |
+
|
| 32 |
+
Evaluated 400 examples
|
| 33 |
+
[[36m2026-07-22 16:59:12,452[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 34 |
+
|
| 35 |
+
Evaluated 400 examples
|
| 36 |
+
[[36m2026-07-22 16:59:15,553[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 37 |
+
|
| 38 |
+
Evaluated 400 examples
|
| 39 |
+
[[36m2026-07-22 16:59:29,564[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 40 |
+
[[36m2026-07-22 16:59:29,915[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 41 |
+
|
| 42 |
+
Evaluated 100 examples
|
| 43 |
+
[[36m2026-07-22 16:59:30,697[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 44 |
+
|
| 45 |
+
Evaluated 100 examples
|
| 46 |
+
[[36m2026-07-22 16:59:32,468[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 47 |
+
[[36m2026-07-22 16:59:32,738[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 48 |
+
|
| 49 |
+
Evaluated 100 examples
|
| 50 |
+
[[36m2026-07-22 16:59:33,420[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 51 |
+
[[36m2026-07-22 16:59:33,421[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 52 |
+
[[36m2026-07-22 16:59:33,421[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 53 |
+
[[36m2026-07-22 16:59:33,655[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 54 |
+
|
| 55 |
+
Evaluated 117 examples
|
| 56 |
+
[[36m2026-07-22 16:59:34,453[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-22 16:59:35,996[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 60 |
+
[[36m2026-07-22 16:59:36,254[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 61 |
+
|
| 62 |
+
Evaluated 117 examples
|
| 63 |
+
[[36m2026-07-22 16:59:37,257[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 64 |
+
[[36m2026-07-22 16:59:37,257[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 65 |
+
[[36m2026-07-22 16:59:37,257[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 66 |
+
[[36m2026-07-22 16:59:37,257[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 67 |
+
[[36m2026-07-22 16:59:37,257[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5553763944252108[0m
|
| 68 |
+
[[36m2026-07-22 16:59:37,791[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 69 |
+
[[36m2026-07-22 16:59:37,798[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 70 |
+
|
| 71 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 72 |
15%|█▌ | 2/13 [00:00<00:00, 13.62it/s]
|
| 73 |
31%|███ | 4/13 [00:00<00:00, 12.18it/s]
|
| 74 |
46%|████▌ | 6/13 [00:00<00:00, 12.83it/s]
|
| 75 |
62%|██████▏ | 8/13 [00:00<00:00, 13.17it/s]
|
| 76 |
77%|███████▋ | 10/13 [00:00<00:00, 13.23it/s]
|
| 77 |
92%|█████████▏| 12/13 [00:00<00:00, 13.16it/s]
|
| 78 |
+
|
| 79 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 80 |
15%|█▌ | 2/13 [00:00<00:00, 12.33it/s]
|
| 81 |
31%|███ | 4/13 [00:00<00:00, 12.33it/s]
|
| 82 |
46%|████▌ | 6/13 [00:00<00:00, 13.03it/s]
|
| 83 |
62%|██████▏ | 8/13 [00:00<00:00, 13.19it/s]
|
| 84 |
77%|███████▋ | 10/13 [00:00<00:00, 13.24it/s]
|
| 85 |
92%|█████████▏| 12/13 [00:00<00:00, 13.18it/s]
|
| 86 |
+
[[36m2026-07-22 16:59:39,719[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 87 |
+
[[36m2026-07-22 16:59:39,728[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 88 |
+
[[36m2026-07-22 16:59:39,728[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -98.07049865796273[0m
|
| 89 |
+
[[36m2026-07-22 16:59:40,007[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 90 |
+
|
| 91 |
+
Evaluated 400 examples
|
| 92 |
+
[[36m2026-07-22 16:59:41,121[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.28232532436914226[0m
|
logs/eval_sanity_full.log
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 1 |
15%|█▌ | 2/13 [00:00<00:00, 13.41it/s]
|
| 2 |
31%|███ | 4/13 [00:00<00:00, 11.55it/s]
|
| 3 |
46%|████▌ | 6/13 [00:00<00:00, 11.84it/s]
|
| 4 |
62%|██████▏ | 8/13 [00:00<00:00, 11.97it/s]
|
| 5 |
77%|███████▋ | 10/13 [00:00<00:00, 12.27it/s]
|
| 6 |
92%|█████████▏| 12/13 [00:00<00:00, 11.97it/s]
|
|
|
|
| 7 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 8 |
15%|█▌ | 2/13 [00:00<00:00, 11.67it/s]
|
| 9 |
31%|███ | 4/13 [00:00<00:00, 11.44it/s]
|
| 10 |
46%|████▌ | 6/13 [00:00<00:00, 11.81it/s]
|
| 11 |
62%|██████▏ | 8/13 [00:00<00:00, 11.80it/s]
|
| 12 |
77%|███████▋ | 10/13 [00:00<00:00, 11.92it/s]
|
| 13 |
92%|█████████▏| 12/13 [00:01<00:00, 12.03it/s]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
/opt/conda/envs/copygen/lib/python3.12/site-packages/requests/__init__.py:113: RequestsDependencyWarning: urllib3 (2.6.3) or chardet (6.0.0.post1)/charset_normalizer (3.4.4) doesn't match a supported version!
|
| 2 |
+
warnings.warn(
|
| 3 |
+
[[36m2026-07-20 13:22:08,612[0m][[34mevaluator[0m][[32mINFO[0m] - Evaluations stored in the experiment directory: ./saves/eval/sanity_full[0m
|
| 4 |
+
[[36m2026-07-20 13:22:08,614[0m][[34mevaluator[0m][[32mINFO[0m] - ***** Running TOFU evaluation suite *****[0m
|
| 5 |
+
[[36m2026-07-20 13:22:08,614[0m][[34mevaluator[0m][[32mINFO[0m] - Fine-grained evaluations will be saved to: ./saves/eval/sanity_full/TOFU_EVAL.json[0m
|
| 6 |
+
[[36m2026-07-20 13:22:08,614[0m][[34mevaluator[0m][[32mINFO[0m] - Aggregated evaluations will be summarised in: ./saves/eval/sanity_full/TOFU_SUMMARY.json[0m
|
| 7 |
+
[[36m2026-07-20 13:22:09,780[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PARA_Prob[0m
|
| 8 |
+
|
| 9 |
+
Evaluated 400 examples
|
| 10 |
+
[[36m2026-07-20 13:22:13,329[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_PERT_Prob[0m
|
| 11 |
+
|
| 12 |
+
Evaluated 400 examples
|
| 13 |
+
[[36m2026-07-20 13:22:26,596[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_truth_ratio[0m
|
| 14 |
+
[[36m2026-07-20 13:22:26,597[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_truth_ratio: 0.4752863258160055[0m
|
| 15 |
+
[[36m2026-07-20 13:22:26,605[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping forget_quality's precompute forget_truth_ratio, already evaluated.[0m
|
| 16 |
+
[[36m2026-07-20 13:22:26,606[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 17 |
+
[[36m2026-07-20 13:22:26,622[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_quality[0m
|
| 18 |
+
[[36m2026-07-20 13:22:26,624[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_quality: 1.876936326891253e-22[0m
|
| 19 |
+
[[36m2026-07-20 13:22:27,065[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_Prob[0m
|
| 20 |
+
|
| 21 |
+
Evaluated 400 examples
|
| 22 |
+
[[36m2026-07-20 13:22:29,732[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_Prob: 0.880693359375[0m
|
| 23 |
+
[[36m2026-07-20 13:22:30,049[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating forget_Q_A_ROUGE[0m
|
| 24 |
+
|
| 25 |
+
Evaluated 400 examples
|
| 26 |
+
[[36m2026-07-20 13:22:39,127[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric forget_Q_A_ROUGE: 0.8206965949827894[0m
|
| 27 |
+
|
| 28 |
+
[[36m2026-07-20 13:22:39,630[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_Prob[0m
|
| 29 |
+
|
| 30 |
+
Evaluated 400 examples
|
| 31 |
+
[[36m2026-07-20 13:22:42,611[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_ROUGE[0m
|
| 32 |
+
|
| 33 |
+
Evaluated 400 examples
|
| 34 |
+
[[36m2026-07-20 13:22:51,099[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PARA_Prob[0m
|
| 35 |
+
|
| 36 |
+
Evaluated 400 examples
|
| 37 |
+
[[36m2026-07-20 13:22:54,044[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Q_A_PERT_Prob[0m
|
| 38 |
+
|
| 39 |
+
Evaluated 400 examples
|
| 40 |
+
[[36m2026-07-20 13:23:06,912[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating retain_Truth_Ratio[0m
|
| 41 |
+
|
| 42 |
+
[[36m2026-07-20 13:23:07,415[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob[0m
|
| 43 |
+
|
| 44 |
+
Evaluated 100 examples
|
| 45 |
+
[[36m2026-07-20 13:23:08,271[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_PERT_Prob[0m
|
| 46 |
+
|
| 47 |
+
Evaluated 100 examples
|
| 48 |
+
[[36m2026-07-20 13:23:09,772[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_Prob_normalised[0m
|
| 49 |
+
[[36m2026-07-20 13:23:10,314[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Q_A_ROUGE[0m
|
| 50 |
+
|
| 51 |
+
Evaluated 100 examples
|
| 52 |
+
[[36m2026-07-20 13:23:11,705[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_Prob, already evaluated.[0m
|
| 53 |
+
[[36m2026-07-20 13:23:11,705[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping ra_Truth_Ratio's precompute ra_Q_A_PERT_Prob, already evaluated.[0m
|
| 54 |
+
[[36m2026-07-20 13:23:11,705[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating ra_Truth_Ratio[0m
|
| 55 |
+
|
| 56 |
+
[[36m2026-07-20 13:23:12,271[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob[0m
|
| 57 |
+
|
| 58 |
+
Evaluated 117 examples
|
| 59 |
+
[[36m2026-07-20 13:23:13,117[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_PERT_Prob[0m
|
| 60 |
+
|
| 61 |
+
Evaluated 117 examples
|
| 62 |
+
[[36m2026-07-20 13:23:14,597[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_Prob_normalised[0m
|
| 63 |
+
[[36m2026-07-20 13:23:14,855[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Q_A_ROUGE[0m
|
| 64 |
+
|
| 65 |
+
Evaluated 117 examples
|
| 66 |
+
[[36m2026-07-20 13:23:16,684[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_Prob, already evaluated.[0m
|
| 67 |
+
[[36m2026-07-20 13:23:16,684[0m][[34mmetrics[0m][[32mINFO[0m] - Skipping wf_Truth_Ratio's precompute wf_Q_A_PERT_Prob, already evaluated.[0m
|
| 68 |
+
[[36m2026-07-20 13:23:16,684[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating wf_Truth_Ratio[0m
|
| 69 |
+
[[36m2026-07-20 13:23:16,684[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating model_utility[0m
|
| 70 |
+
[[36m2026-07-20 13:23:16,685[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric model_utility: 0.5994289783093724[0m
|
| 71 |
+
|
| 72 |
+
[[36m2026-07-20 13:23:17,399[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 73 |
+
[[36m2026-07-20 13:23:17,416[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating mia_min_k[0m
|
| 74 |
+
|
| 75 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 76 |
15%|█▌ | 2/13 [00:00<00:00, 13.41it/s]
|
| 77 |
31%|███ | 4/13 [00:00<00:00, 11.55it/s]
|
| 78 |
46%|████▌ | 6/13 [00:00<00:00, 11.84it/s]
|
| 79 |
62%|██████▏ | 8/13 [00:00<00:00, 11.97it/s]
|
| 80 |
77%|███████▋ | 10/13 [00:00<00:00, 12.27it/s]
|
| 81 |
92%|█████████▏| 12/13 [00:00<00:00, 11.97it/s]
|
| 82 |
+
|
| 83 |
0%| | 0/13 [00:00<?, ?it/s]
|
| 84 |
15%|█▌ | 2/13 [00:00<00:00, 11.67it/s]
|
| 85 |
31%|███ | 4/13 [00:00<00:00, 11.44it/s]
|
| 86 |
46%|████▌ | 6/13 [00:00<00:00, 11.81it/s]
|
| 87 |
62%|██████▏ | 8/13 [00:00<00:00, 11.80it/s]
|
| 88 |
77%|███████▋ | 10/13 [00:00<00:00, 11.92it/s]
|
| 89 |
92%|█████████▏| 12/13 [00:01<00:00, 12.03it/s]
|
| 90 |
+
[[36m2026-07-20 13:23:19,516[0m][[34mmetrics[0m][[32mINFO[0m] - Loading evaluations from /workspace/jspace-unlearning/refs/tofu_Llama-3.2-1B-Instruct_retain90__TOFU_EVAL.json[0m
|
| 91 |
+
[[36m2026-07-20 13:23:19,530[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating privleak[0m
|
| 92 |
+
[[36m2026-07-20 13:23:19,531[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric privleak: -99.4563786763651[0m
|
| 93 |
+
[[36m2026-07-20 13:23:19,811[0m][[34mmetrics[0m][[32mINFO[0m] - Evaluating extraction_strength[0m
|
| 94 |
+
|
| 95 |
+
Evaluated 400 examples
|
| 96 |
+
[[36m2026-07-20 13:23:20,930[0m][[34mevaluator[0m][[32mINFO[0m] - Result for metric extraction_strength: 0.7064597846379354[0m
|
logs/h2_train_h3rl_WD_npo_l0.log
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
| 0 |
0%| | 0/6 [00:00<?, ?it/s]
|
| 1 |
17%|█▋ | 1/6 [00:00<00:04, 1.07it/s]
|
| 2 |
33%|███▎ | 2/6 [00:03<00:07, 1.89s/it]
|
| 3 |
50%|█████ | 3/6 [00:05<00:06, 2.16s/it]
|
| 4 |
67%|██████▋ | 4/6 [00:08<00:04, 2.22s/it]
|
| 5 |
83%|████████▎ | 5/6 [00:10<00:02, 2.28s/it]
|
| 6 |
|
| 7 |
83%|████████▎ | 5/6 [00:10<00:02, 2.28s/it]
|
| 8 |
|
|
|
|
|
|
|
|
|
| 1 |
+
[[36m2026-07-21 12:37:56,565[0m][[34mtrainer[0m][[32mINFO[0m] - FinetuneTrainer Trainer loaded, output_dir: /workspace/models/h2_tmp/h3rl_WD_npo_l0[0m
|
| 2 |
+
|
| 3 |
0%| | 0/6 [00:00<?, ?it/s]
|
| 4 |
17%|█▋ | 1/6 [00:00<00:04, 1.07it/s]
|
| 5 |
33%|███▎ | 2/6 [00:03<00:07, 1.89s/it]
|
| 6 |
50%|█████ | 3/6 [00:05<00:06, 2.16s/it]
|
| 7 |
67%|██████▋ | 4/6 [00:08<00:04, 2.22s/it]
|
| 8 |
83%|████████▎ | 5/6 [00:10<00:02, 2.28s/it]
|
| 9 |
|
| 10 |
83%|████████▎ | 5/6 [00:10<00:02, 2.28s/it]
|
| 11 |
|
| 12 |
+
{'loss': 1.3583, 'grad_norm': 17.125, 'learning_rate': 1.3333333333333333e-05, 'epoch': 0.8}
|
| 13 |
+
{'train_runtime': 15.1667, 'train_samples_per_second': 13.187, 'train_steps_per_second': 0.396, 'train_loss': 1.2528751492500305, 'epoch': 0.96}
|