HBERTv1_48_L8_H128_A2 / train_results.json
gokuls's picture
End of training
b38b45c
{
"epoch": 9.92,
"train_loss": 6.175604047275691,
"train_runtime": 197999.2443,
"train_samples": 5858758,
"train_samples_per_second": 2958.98,
"train_steps_per_second": 26.9
}