b1_math_top_2 / train_results.json
neginr's picture
End of training
cd15c1e verified
{
"epoch": 5.0,
"total_flos": 1.9354956165367726e+18,
"train_loss": 0.2693426305707167,
"train_runtime": 23659.8967,
"train_samples_per_second": 6.678,
"train_steps_per_second": 0.052
}