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

11 lines
342 B
JSON
Raw Permalink Normal View History

{
"ewc_loss": 0.0400390625,
"ewc_loss_diag": 0.00014019012451171875,
"ewc_loss_parallel": 2.5987625122070312e-05,
"total_flos": 5.62815163329864e+19,
"train_loss": 0.4244929082353308,
"train_runtime": 47803.2787,
"train_samples": 125770,
"train_samples_per_second": 7.893,
"train_steps_per_second": 0.493
}