d1_science_gpt / train_results.json
neginr's picture
End of training
9d6c538 verified
{
"epoch": 5.0,
"total_flos": 1.836079948138283e+18,
"train_loss": 0.3704581044342836,
"train_runtime": 21246.0211,
"train_samples_per_second": 7.437,
"train_steps_per_second": 0.058
}