HBERTv1_48_L6_H256_A4 / train_results.json
gokuls's picture
End of training
18c217f
{
"epoch": 8.19,
"train_loss": 5.961271763625358,
"train_runtime": 197998.9382,
"train_samples": 5858758,
"train_samples_per_second": 2958.985,
"train_steps_per_second": 26.9
}