Files
Qwen3-4B-SFT-science-1e-5/train_results.json

8 lines
208 B
JSON
Raw Normal View History

{
"epoch": 3.0,
"total_flos": 8.852766725217714e+18,
"train_loss": 0.6584745990833162,
"train_runtime": 19013.667,
"train_samples_per_second": 15.562,
"train_steps_per_second": 0.122
}