Llama-3.1-8B-Instruct-SFT-400 / train_results.json
chchen's picture
End of training
3e63deb verified
raw
history blame
221 Bytes
{
"epoch": 9.777777777777779,
"total_flos": 1.959093130965811e+16,
"train_loss": 0.4326907604932785,
"train_runtime": 310.7933,
"train_samples_per_second": 11.583,
"train_steps_per_second": 0.708
}