HBERTv1_48_L2_H128_A2_ffn_3 / train_results.json
gokuls's picture
End of training
40d588e
raw
history blame contribute delete
203 Bytes
{
"epoch": 14.08,
"train_loss": 6.209815649055619,
"train_runtime": 197999.0208,
"train_samples": 5858758,
"train_samples_per_second": 2958.983,
"train_steps_per_second": 16.439
}