parallel-gpt2-medium-wikitext / train_results.json
shivanandmn's picture
Training in progress, step 500
f1cd615 verified
{
"epoch": 4.998736842105263,
"total_flos": 1.0584067483285586e+18,
"train_loss": 3.584141966119902,
"train_runtime": 28526.973,
"train_samples_per_second": 19.98,
"train_steps_per_second": 0.624
}