HBERTv1_48_L12_H256_A4 / train_results.json
gokuls's picture
End of training
1dd85cf
raw
history blame contribute delete
202 Bytes
{
"epoch": 5.45,
"train_loss": 5.580446514515485,
"train_runtime": 197998.7388,
"train_samples": 5858758,
"train_samples_per_second": 2958.988,
"train_steps_per_second": 46.235
}