phi3-4k-chinese-orpo / train_results.json
postitive666
orpo chinese phi3 4K
3c7b14a
{
"epoch": 2.994601079784043,
"total_flos": 132590267662336.0,
"train_loss": 0.7937506708579186,
"train_runtime": 49781.9259,
"train_samples_per_second": 1.205,
"train_steps_per_second": 0.025
}