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

11 lines
341 B
JSON
Raw Permalink Normal View History

{
"ewc_loss": 0.041015625,
"ewc_loss_diag": 0.00014019012451171875,
"ewc_loss_parallel": 2.682209014892578e-05,
"total_flos": 5.62815163329864e+19,
"train_loss": 0.42514991610376646,
"train_runtime": 47075.3517,
"train_samples": 125770,
"train_samples_per_second": 8.015,
"train_steps_per_second": 0.501
}