HBERTv1_48_L10_H768_A12 / train_results.json
gokuls's picture
End of training
8deb113
{
"epoch": 2.15,
"train_loss": 4.290504575642074,
"train_runtime": 197999.5648,
"train_samples": 5858758,
"train_samples_per_second": 2958.975,
"train_steps_per_second": 52.839
}