HBERTv1_48_L8_H256_A4 / train_results.json
gokuls's picture
End of training
c809032
{
"epoch": 7.01,
"train_loss": 5.837019857484345,
"train_runtime": 197999.0531,
"train_samples": 5858758,
"train_samples_per_second": 2958.983,
"train_steps_per_second": 30.823
}