HBERTv1_48_L2_H128_A2_ffn_1 / train_results.json
gokuls's picture
End of training
cd889f9
{
"epoch": 14.21,
"train_loss": 6.210341006701539,
"train_runtime": 197999.3651,
"train_samples": 5858758,
"train_samples_per_second": 2958.978,
"train_steps_per_second": 16.439
}