trainer_output / train_results.json
sunmengjie's picture
Add checkpoint -1 post-trained on curated_deepscaler
16db530 verified
{
"total_flos": 0.0,
"train_loss": 0.0,
"train_runtime": 0.7184,
"train_samples_per_second": 41.762,
"train_steps_per_second": 8.352
}