Files
llama3-8b-full-pretrain-c4-…/train_results.json

8 lines
200 B
JSON
Raw Permalink Normal View History

{
"epoch": 3.0,
"total_flos": 4888573747200.0,
"train_loss": 1.5217517310478648,
"train_runtime": 2117.074,
"train_samples_per_second": 3.452,
"train_steps_per_second": 0.432
}