reasoning_hp_ablations_bsz256 / train_results.json
sedrickkeh's picture
End of training
3dc16e9 verified
{
"epoch": 2.9983155530600785,
"total_flos": 4617447504347136.0,
"train_loss": 0.46386746891428915,
"train_runtime": 117690.8189,
"train_samples_per_second": 2.905,
"train_steps_per_second": 0.011
}