hBERTv1_new_pretrain_48_KD_sst2 / train_results.json
gokuls's picture
End of training
fe10aa6
{
"epoch": 6.0,
"train_loss": 0.21436870558061305,
"train_runtime": 2799.5615,
"train_samples": 67349,
"train_samples_per_second": 1202.849,
"train_steps_per_second": 9.412
}