Files
qwen3-4B-instruct-refiner-sft/train_results.json

8 lines
209 B
JSON
Raw Permalink Normal View History

{
"epoch": 5.0,
"total_flos": 3.0080813400754176e+18,
"train_loss": 0.12334943315905438,
"train_runtime": 40705.595,
"train_samples_per_second": 2.097,
"train_steps_per_second": 0.066
}