Qwen1.5-MOE-sft-math7k-sft-2epochs-frozen-router/train_results.json

{
    "total_flos": 4.0376628208572826e+17,
    "train_loss": 0.22927807642276896,
    "train_runtime": 2180.6252,
    "train_samples": 6851,
    "train_samples_per_second": 6.284,
    "train_steps_per_second": 0.197
}