opt-babylm2-subset-default-1e-3 / train_results.json
kanishka's picture
End of training
380b940 verified
raw
history blame contribute delete
238 Bytes
{
"epoch": 10.0,
"total_flos": 5.9582336320512e+17,
"train_loss": 2.1796733167279037,
"train_runtime": 31005.8847,
"train_samples": 452524,
"train_samples_per_second": 145.948,
"train_steps_per_second": 4.561
}