steer-full_3b / run /run_meta.json
JacoDuToit's picture
run directory (provenance + evals)
aa802d1 verified
Raw
History Blame Contribute Delete
5.22 kB
{
"run": "full_3b",
"started_at": "2026-08-05T17:16:57+0000",
"argv": [
"scripts/run.py",
"configs/full_3b.yaml",
"--stages",
"eval_m1"
],
"stages": [
"eval_m1"
],
"config_path": "configs/full_3b.yaml",
"config": {
"model_id": "Qwen/Qwen2.5-3B-Instruct",
"device": "auto",
"dtype": "auto",
"seed": 0,
"layer": 18,
"train_alphas": [
0.4,
0.8,
1.2
],
"eval_alphas": [
0.4,
0.8,
1.0,
1.2,
1.6
],
"efficacy_alpha": 0.8,
"efficacy_min_rate": 0.3,
"n_contrast_pairs": 10,
"qa_source": "json",
"concepts_path": "data/concepts_paper.json",
"qa_path": "data/qa_open.json",
"relevant_frac": 0.6,
"steered_frac": 0.7,
"eval_question_frac": 0.2,
"repeats_per_question": 6,
"alpaca_replay_frac": 0.5,
"eval_concepts_per_split": {
"train": 40,
"heldout": null
},
"lora_r": 16,
"lora_alpha": 32,
"lora_dropout": 0.05,
"lora_targets": [
"q_proj",
"k_proj",
"v_proj",
"o_proj",
"gate_proj",
"up_proj",
"down_proj"
],
"lr": 0.0001,
"epochs": 1,
"effective_batch_size": 16,
"train_batch_size": 4,
"max_seq_len": 256,
"max_new_tokens": 64,
"eval_batch_size": 8,
"bootstrap_resamples": 10000,
"capability_tasks": {
"mmlu": 5,
"gsm8k_cot": 8
},
"capability_apply_chat_template": true,
"capability_limit": "mmlu=15,gsm8k_cot=200",
"wandb_project": null,
"wandb_entity": null,
"hub_repo_id": null,
"hub_private": true,
"vectors_path": "results/full_3b/vectors.pt",
"train_examples_path": "results/full_3b/train_examples.json",
"eval_questions_path": "results/full_3b/eval_questions.json",
"adapter_dir": "results/full_3b/m1_resist_adapter",
"results_dir": "results/full_3b"
},
"git": {
"commit": "b16f6e72a9bb0c38dd20d0bab46579325178c793",
"branch": "main",
"dirty": true,
"remote": "https://github.com/JacoDuToit11/steering-resistance.git"
},
"env": {
"python": "3.11.10",
"platform": "Linux-6.5.0-35-generic-x86_64-with-glibc2.35",
"accelerator": "NVIDIA A100 80GB PCIe",
"packages": {
"torch": "2.6.0+cu124",
"transformers": "5.14.1",
"peft": "0.20.0",
"accelerate": "1.14.0",
"datasets": "5.0.1",
"numpy": "2.4.6",
"wandb": null,
"huggingface_hub": "1.26.0"
}
},
"data": {
"concepts_path": {
"path": "data/concepts_paper.json",
"sha256": "adc6aee9b1cc537af09ca2229ccbb470adfdaa3e73d05be69df031e042c4347b"
},
"qa_path": {
"path": "data/qa_open.json",
"sha256": "fa2356571420fd8c2a444aee6e8c879b865e0cd21f5ade2938a0308effbed8f2"
}
},
"status": "success",
"wandb_url": null,
"headline": "clean 100%->100% · steer_heldout@1.6 correct 0%->4%",
"finished_at": "2026-08-05T19:23:32+0000",
"artifacts": {
"capability/capability_summary.csv": {
"sha256": "d5bbb31095b80e60dd46d2e5cdaa7ef2a781ccc89de5766381b21af1425233b9",
"bytes": 128
},
"capability/capability_summary.md": {
"sha256": "ebf7fa98749913e9f77670b2f926ead49fd2dfdf3cacdafbc27b873e088a7218",
"bytes": 363
},
"capability/lmeval_run.log": {
"sha256": "f3d93e78d2473755612d80a0afb6eaec8ed8027f13b4545ee8ec4d59370d4731",
"bytes": 298468
},
"capability/m0/gsm8k_cot/Qwen__Qwen2.5-3B-Instruct/results_2026-07-23T18-36-29.229735.json": {
"sha256": "fa3feefea52e589b23d972c3393a9222b7affa81d04b782a111cbfb5f6e561bc",
"bytes": 14432
},
"capability/m0/mmlu/Qwen__Qwen2.5-3B-Instruct/results_2026-07-23T18-15-57.670538.json": {
"sha256": "dfd686f19edce4cb31bffd091167c599b0596b15987d25303613295099cb4562",
"bytes": 139207
},
"capability/m1/gsm8k_cot/__root__steering-resistance__results__full_3b__m1_resist_adapter/results_2026-07-23T19-07-18.393003.json": {
"sha256": "35a97896c6938ea0031d99b63a6787f635f4d88b68a52c7aec53a1849b28a25c",
"bytes": 14681
},
"capability/m1/mmlu/__root__steering-resistance__results__full_3b__m1_resist_adapter/results_2026-07-23T18-41-47.488437.json": {
"sha256": "e6e525c0832477c3729aca9663ded0bc63ab583d6c521a550365fb8656f1fc0c",
"bytes": 143701
},
"eval_m0.jsonl": {
"sha256": "1ae0cbda4668a35852e1ef234043f320d320ae9cb7cf4017b4d2f79f1427605a",
"bytes": 11318115
},
"eval_m1.jsonl": {
"sha256": "568a78889056010226b44096a76f85affbddd9670a2f814a554ad4865ec59d42",
"bytes": 9539423
},
"eval_questions.json": {
"sha256": "8771796f4c91a61a901278e799032623900657aaa8f9e44ce789dd1fadd8bf8f",
"bytes": 27617
},
"m1_resist_adapter/adapter_config.json": {
"sha256": "390e661694a2f685b6256b5d38a80aa41c5236a115df9cfb5edd4a96da0e568e",
"bytes": 1103
},
"m1_resist_adapter/adapter_model.safetensors": {
"sha256": "bc5cf2921ad4f341fc270e0b53641804b4c725c9031d5d232dd6cb1f8877a23d",
"bytes": 119801528
},
"summary.csv": {
"sha256": "5c5eba02b37159e9c6623e298df60b71ab79f55c9f394342311907bdfb3ec526",
"bytes": 2616
},
"summary.md": {
"sha256": "5335fee8d923b9d9e3d7a1f08d248e45ba0a3f067d81b3e0b6e18dd6395a587f",
"bytes": 2628
},
"train_examples.json": {
"sha256": "3e2ac73debbb5da8cadb4e114a35f6346f015ca0f22d52b9486a7fd2b2b7e824",
"bytes": 284577
},
"vectors.pt": {
"sha256": "5c9783914f6419b376d133a3872349ac5a07b42c8456894a3214ffa6618158c4",
"bytes": 7757862
}
}
}