Files
Qwen3-1.7B-MATH-GDPO/train_results.json

8 lines
200 B
JSON
Raw Normal View History

{
"total_flos": 0.0,
"train_loss": -0.09854654141236097,
"train_runtime": 3102.2738,
"train_samples": 1348,
"train_samples_per_second": 0.869,
"train_steps_per_second": 0.054
}