72 lines
2.4 KiB
JSON
72 lines
2.4 KiB
JSON
|
{
|
||
|
|
"created_at": "2026-07-15 00:59:15",
|
||
|
|
"name": "v34_v29_paper_mix_lr5e7_ep10",
|
||
|
|
"base": "/home/ll/llm4rec/experiments/outputs/v29_v19_user_world_guard_lr8e7_ep018",
|
||
|
|
"dataset": "v34_paper_mix",
|
||
|
|
"dataset_manifest": {
|
||
|
|
"name": "v34_paper_mix",
|
||
|
|
"path": "/home/ll/llm4rec/experiments/data/v34_paper_mix.jsonl",
|
||
|
|
"records": 80000,
|
||
|
|
"bytes": 287964112,
|
||
|
|
"sha256": "114013600aa0113ef0ef441bc1ecbd16a8c01f1a77cf76b9d0de16040656da8f",
|
||
|
|
"requested_components": {
|
||
|
|
"general": 30400,
|
||
|
|
"r0": 19000,
|
||
|
|
"r3_think": 6000,
|
||
|
|
"r3_direct": 12000,
|
||
|
|
"old_rec": 5000,
|
||
|
|
"old_user": 3800,
|
||
|
|
"old_item": 3800
|
||
|
|
},
|
||
|
|
"groups": {
|
||
|
|
"general": 30400,
|
||
|
|
"old_rec": 5000,
|
||
|
|
"r3": 18000,
|
||
|
|
"r0": 19000,
|
||
|
|
"old_item": 3800,
|
||
|
|
"old_user": 3800
|
||
|
|
},
|
||
|
|
"variants": {
|
||
|
|
"official_general_cot": 15121,
|
||
|
|
"rec_no_think_direct_final": 2520,
|
||
|
|
"profile_goods_direct": 2647,
|
||
|
|
"caption_to_sid_direct": 9502,
|
||
|
|
"profile_live_cot": 1414,
|
||
|
|
"official_general_direct": 15279,
|
||
|
|
"profile_video/video_cot": 1876,
|
||
|
|
"sid_to_caption_cot": 4746,
|
||
|
|
"item_short_think": 699,
|
||
|
|
"profile_video/video_direct": 3676,
|
||
|
|
"original_competition": 6105,
|
||
|
|
"profile_goods_cot": 1350,
|
||
|
|
"profile_live_direct": 2834,
|
||
|
|
"profile_video/ad_direct": 2843,
|
||
|
|
"sid_to_caption_direct": 4752,
|
||
|
|
"user_strict_array": 666,
|
||
|
|
"item_no_think_direct_final": 670,
|
||
|
|
"profile_video/ad_cot": 1360,
|
||
|
|
"user_strict_logic": 542,
|
||
|
|
"user_extra_no_think_logic": 228,
|
||
|
|
"user_logic_json_strict": 529,
|
||
|
|
"user_array_json_strict": 641
|
||
|
|
},
|
||
|
|
"routes": {
|
||
|
|
"think": 31005,
|
||
|
|
"no_think": 48995
|
||
|
|
},
|
||
|
|
"target_leakage": 0,
|
||
|
|
"seed": 202607151
|
||
|
|
},
|
||
|
|
"learning_rate": "5.0e-7",
|
||
|
|
"epochs": 1.0,
|
||
|
|
"stages": null,
|
||
|
|
"note": "Anchor run using a scaled OneReason-like R0/R3/general mixture plus old-task replay.",
|
||
|
|
"cot_policy": "Preserve meaningful CoT for /think and pure final outputs for /no_think.",
|
||
|
|
"checkpoint_policy": "save_strategy=no; only the final model is retained.",
|
||
|
|
"script": "/home/ll/llm4rec/experiments/run_v34_v43.py",
|
||
|
|
"script_sha256": "fbf01773c61579cf6159f64442fc0c8902714c84ed23b488bc9a6236a154f402",
|
||
|
|
"configs": [
|
||
|
|
"/home/ll/llm4rec/experiments/configs/v34_v29_paper_mix_lr5e7_ep10.yaml"
|
||
|
|
],
|
||
|
|
"reproduce": "/home/ll/llm4rec/demo/LLaMA-Factory/.venv/bin/python /home/ll/llm4rec/experiments/run_v34_v43.py --single v34_v29_paper_mix_lr5e7_ep10 --gpu <gpu>"
|
||
|
|
}
|