Files
qwen3-4b-think/train_results.json

8 lines
204 B
JSON
Raw Permalink Normal View History

{
"epoch": 3.0,
"total_flos": 1274755977576448.0,
"train_loss": 0.5495063810720958,
"train_runtime": 34913.8374,
"train_samples_per_second": 3.33,
"train_steps_per_second": 0.026
}