HBERTv1_48_L10_H64_A2 / train_results.json
gokuls's picture
End of training
fb91465
{
"epoch": 10.69,
"train_loss": 6.316021415674524,
"train_runtime": 197999.0367,
"train_samples": 5858758,
"train_samples_per_second": 2958.983,
"train_steps_per_second": 26.9
}