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

11 lines
314 B
JSON
Raw Normal View History

{
"ewc_loss": -2128.0,
"ewc_loss_diag": 0.0050048828125,
"ewc_loss_parallel": -2.125,
"total_flos": 5.62815163329864e+19,
"train_loss": -653.4477224089757,
"train_runtime": 47383.3285,
"train_samples": 125770,
"train_samples_per_second": 7.963,
"train_steps_per_second": 0.498
}