Mistral-7B-Instruct-v0.3-ORPO-SFT / train_results.json
chchen's picture
End of training
d4981e1 verified
raw
history blame
222 Bytes
{
"epoch": 2.986666666666667,
"total_flos": 1.5795369631678464e+16,
"train_loss": 0.11316087664592833,
"train_runtime": 355.6972,
"train_samples_per_second": 7.591,
"train_steps_per_second": 0.472
}