llama-3-86-lora-pretrain / train_results.json
ytcheng's picture
End of training
f7878d0 verified
raw
history blame contribute delete
206 Bytes
{
"epoch": 3.0,
"total_flos": 8.691117166323302e+17,
"train_loss": 2.267601152893844,
"train_runtime": 8636.3508,
"train_samples_per_second": 2.181,
"train_steps_per_second": 0.273
}