Files
Qwen2.5-0.5B-Math-SFT-1024/train_results.json

8 lines
207 B
JSON
Raw Permalink Normal View History

{
"epoch": 3.0,
"total_flos": 5.359421722984448e+16,
"train_loss": 0.5856129340097016,
"train_runtime": 310.1981,
"train_samples_per_second": 41.751,
"train_steps_per_second": 0.658
}