HBERTv1_48_L12_H768_A12_ffn_1_KD / train_results.json
gokuls's picture
End of training
a071a1c
{
"epoch": 0.25,
"train_loss": 159.1573161326461,
"train_runtime": 35999.4318,
"train_samples": 5858758,
"train_samples_per_second": 16274.585,
"train_steps_per_second": 508.583
}