HBERTv1_48_L2_H768_A12_ffn_5 / train_results.json
gokuls's picture
End of training
5df1f9a
raw
history blame contribute delete
200 Bytes
{
"epoch": 5.16,
"train_loss": 4.093998171579732,
"train_runtime": 197997.7712,
"train_samples": 5858758,
"train_samples_per_second": 2959.002,
"train_steps_per_second": 26.9
}