Llama3-8B-ORPO / all_results.json
randal
init: Fine-tune Llama 3 with ORPO
00df52e
raw
history blame contribute delete
204 Bytes
{
"epoch": 3.0,
"total_flos": 126401337753600.0,
"train_loss": 0.8602390615145366,
"train_runtime": 16444.8476,
"train_samples_per_second": 3.649,
"train_steps_per_second": 0.228
}