HBERTv1_48_L2_H128_A2_ffn_2 / train_results.json
gokuls's picture
End of training
cd29a5b
raw
history blame contribute delete
202 Bytes
{
"epoch": 14.3,
"train_loss": 6.206436465621404,
"train_runtime": 197999.3597,
"train_samples": 5858758,
"train_samples_per_second": 2958.978,
"train_steps_per_second": 16.439
}