Files
Llama-3.1-8B-Instruct_SFT_s…/train_results.json

9 lines
260 B
JSON
Raw Permalink Normal View History

{
"ewc_loss": 6.580352783203125e-05,
"total_flos": 3.1829322264618205e+19,
"train_loss": 0.9609540035613102,
"train_runtime": 37132.1425,
"train_samples": 145693,
"train_samples_per_second": 11.771,
"train_steps_per_second": 0.736
}