llama-3-8b-instruct-sppo-iter2 / train_results.json
jcmei's picture
End of training
841a5a9 verified
{
"epoch": 1.0,
"total_flos": 0.0,
"train_loss": 124066.95723157052,
"train_runtime": 2762.769,
"train_samples": 19958,
"train_samples_per_second": 7.224,
"train_steps_per_second": 0.113
}