HBERTv1_48_L2_H128_A2_ffn_5 / train_results.json
gokuls's picture
End of training
a50aeef
raw
history blame contribute delete
203 Bytes
{
"epoch": 13.83,
"train_loss": 6.195823538107642,
"train_runtime": 197999.5037,
"train_samples": 5858758,
"train_samples_per_second": 2958.976,
"train_steps_per_second": 18.494
}