初始化项目,由ModelHub XC社区提供模型

Model: promotion/qwen3-8b-dpo-avg-beta0p01-s42
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-07-29 05:32:20 +08:00
commit 49bc018110
22 changed files with 152912 additions and 0 deletions

15
run_status.json Normal file
View File

@@ -0,0 +1,15 @@
{
"method": "dpo",
"loss_type": "dpo",
"seed": 42,
"beta": 0.01,
"backbone": "Qwen/Qwen3-8B",
"status": "completed",
"returncode": 0,
"completed_at": "2026-07-08T05:38:57Z",
"dataset": "/home/sjkim/mnpo_shared/revision_qwen3_8b/full_iter1/precomputed/avg_oracle",
"output_dir": "/ext_hdd/sjkim/mnpo/revision_qwen3_8b/full_iter1/train/dpo_avg_beta0p01_s42_odin2",
"config": "/ext_hdd/sjkim/mnpo/revision_qwen3_8b/full_iter1/train/dpo_avg_beta0p01_s42_odin2/config.yaml",
"train_metrics_exists": true,
"trainer_state_exists": true
}