qlora-llama3b-iterative / train_results.json
pakphum's picture
End of training
9823daf verified
raw
history blame
220 Bytes
{
"epoch": 4.444444444444445,
"total_flos": 1.038706354200576e+16,
"train_loss": 0.1301481350355316,
"train_runtime": 1611.4177,
"train_samples_per_second": 2.482,
"train_steps_per_second": 0.31
}