dfine-cppe5 / train_results.json
nonl's picture
End of training
482f1df verified
{
"epoch": 300.0,
"total_flos": 1.9198779273216e+19,
"train_loss": 11.328886562834647,
"train_runtime": 8885.8834,
"train_samples_per_second": 28.697,
"train_steps_per_second": 3.612
}