k8s-rca-orpo/training_summary.json

{
  "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
}