HBERTv1_48_L6_H512_A8 / train_results.json
gokuls's picture
End of training
3ad56d1
{
"epoch": 4.83,
"train_loss": 5.167583625387428,
"train_runtime": 197999.0273,
"train_samples": 5858758,
"train_samples_per_second": 2958.983,
"train_steps_per_second": 30.823
}