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

11 lines
330 B
JSON
Raw Permalink Normal View History

{
"ewc_loss": 0.008638550527393818,
"ewc_loss_diag": 0.0,
"ewc_loss_parallel": 8.638550207251683e-05,
"total_flos": 5.62815163329864e+19,
"train_loss": 0.3824227017627763,
"train_runtime": 45517.1597,
"train_samples": 125770,
"train_samples_per_second": 8.289,
"train_steps_per_second": 0.518
}