HBERTv1_L12_H768_A12_ffn_1 / train_results.json
gokuls's picture
End of training
8e63567
raw
history blame contribute delete
201 Bytes
{
"epoch": 2.4,
"train_loss": 4.907397945397096,
"train_runtime": 197998.9762,
"train_samples": 5858758,
"train_samples_per_second": 2958.984,
"train_steps_per_second": 61.646
}