b1_math_top_4 / all_results.json
neginr's picture
End of training
dceb7d7 verified
{
"epoch": 5.0,
"total_flos": 2.190159325803053e+18,
"train_loss": 0.32256396708459506,
"train_runtime": 24829.5428,
"train_samples_per_second": 6.363,
"train_steps_per_second": 0.05
}