Files
k8s-rca-orpo/training_summary.json

10 lines
240 B
JSON
Raw Permalink Normal View History

{
"method": "ORPO",
"base_model": "Qwen/Qwen2.5-1.5B-Instruct",
"dataset": "/workspace/dataset/dpo_dataset_v2.jsonl",
"lambda_orpo": 0.1,
"lr": 8e-06,
"epochs": 3,
"train_loss": 0.5324908440029071,
"runtime_min": 50.051775
}