HBERTv1_48_L12_H512_A8 / train_results.json
gokuls's picture
End of training
d34d050
{
"epoch": 3.02,
"train_loss": 4.694392849819643,
"train_runtime": 197999.2569,
"train_samples": 5858758,
"train_samples_per_second": 2958.98,
"train_steps_per_second": 52.839
}