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

11 lines
348 B
JSON
Raw Permalink Normal View History

{
"ewc_loss": 0.07620732486248016,
"ewc_loss_diag": 3.6716461181640625e-05,
"ewc_loss_parallel": 0.0003958623274229467,
"total_flos": 5.62815163329864e+19,
"train_loss": 0.46694955879437966,
"train_runtime": 50208.5497,
"train_samples": 125770,
"train_samples_per_second": 7.515,
"train_steps_per_second": 0.47
}