{
 "run": "full_7b",
 "started_at": "2026-07-23T18:26:39+0000",
 "argv": [
  "scripts/run.py",
  "configs/full_7b.yaml"
 ],
 "stages": [
  "vectors",
  "data",
  "eval_m0",
  "train",
  "eval_m1"
 ],
 "config_path": "configs/full_7b.yaml",
 "config": {
  "model_id": "Qwen/Qwen2.5-7B-Instruct",
  "device": "auto",
  "dtype": "auto",
  "seed": 0,
  "layer": 14,
  "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": 40,
  "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": 2,
  "max_seq_len": 256,
  "max_new_tokens": 64,
  "eval_batch_size": 4,
  "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": "JacoDuToit/steer-full_7b",
  "hub_private": true,
  "vectors_path": "results/full_7b/vectors.pt",
  "train_examples_path": "results/full_7b/train_examples.json",
  "eval_questions_path": "results/full_7b/eval_questions.json",
  "adapter_dir": "results/full_7b/m1_resist_adapter",
  "results_dir": "results/full_7b"
 },
 "git": {
  "commit": null,
  "branch": null,
  "dirty": null,
  "remote": null
 },
 "env": {
  "python": "3.11.10",
  "platform": "Linux-6.8.0-49-generic-x86_64-with-glibc2.35",
  "accelerator": "NVIDIA GeForce RTX 3090",
  "packages": {
   "torch": "2.13.0+cu126",
   "transformers": "5.14.1",
   "peft": "0.19.1",
   "accelerate": "1.14.0",
   "datasets": "5.0.0",
   "numpy": "2.4.6",
   "wandb": null,
   "huggingface_hub": "1.24.0"
  }
 },
 "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%->20%",
 "hub_url": "https://huggingface.co/JacoDuToit/steer-full_7b",
 "finished_at": "2026-07-23T20:16:36+0000",
 "artifacts": {
  "eval_m0.jsonl": {
   "sha256": "850d1c677802f170a8e4552ab3323b6438acca7274047b8a6590c21451e69d49",
   "bytes": 4749617
  },
  "eval_m1.jsonl": {
   "sha256": "0a1ccba3eea0950ffc11429f3d8d89e58b5533db3b3605a17e77e07f8c05416f",
   "bytes": 4129347
  },
  "eval_questions.json": {
   "sha256": "0685a496f1a91ac4c4e06f4f56486ed8c615ba482cde3691002cd5ffa6d47a2b",
   "bytes": 30264
  },
  "m1_resist_adapter/README.md": {
   "sha256": "3191b465069be139a907db620ecaaee6188deea1f5e14f8ec7281f71bba535d2",
   "bytes": 5202
  },
  "m1_resist_adapter/adapter_config.json": {
   "sha256": "9a907152004d1043d9167bedfa3ba620293502340a75b5a0353fb8a4631a3531",
   "bytes": 1103
  },
  "m1_resist_adapter/adapter_model.safetensors": {
   "sha256": "63031303ded8f07d12eb164746743fd2ee2aae4d97b6e25df947ef6f5776f5f6",
   "bytes": 161533192
  },
  "summary.csv": {
   "sha256": "40dbf0f76cf633f9312039bdd6792efc2d54b0b23ee483044a0b82052e9d2626",
   "bytes": 2561
  },
  "summary.md": {
   "sha256": "7bc571ac210c16f6a6decea771082a8f3af230aae8bdc1de405012d86f09bc1f",
   "bytes": 2606
  },
  "train_examples.json": {
   "sha256": "1ba9a7154e366cab28361a413378675c232d26b906fac8e05e82d67282d9e8dd",
   "bytes": 301292
  },
  "vectors.pt": {
   "sha256": "a4efa358fc4a9d785793739be44c5355e7d29cf5f8aa17927c1c80fcac19bc4d",
   "bytes": 10884283
  }
 }
}