qwen2.5-0.5b-expo-DPO-ES-0.1 / all_results.json
hZzy's picture
End of training
1390069 verified
raw
history blame
796 Bytes
{
"epoch": 1.5588096362777515,
"eval_dpo_loss": 0.6893911957740784,
"eval_logits": -2.1208713054656982,
"eval_logps": -90.9490966796875,
"eval_loss": 0.6922824382781982,
"eval_objective": 0.6893911957740784,
"eval_ranking_idealized": 0.6030020713806152,
"eval_ranking_idealized_expo": 0.5222567319869995,
"eval_ranking_simple": 0.556418240070343,
"eval_regularize": 0.6893911957740784,
"eval_runtime": 314.8886,
"eval_samples": 5790,
"eval_samples_per_second": 18.387,
"eval_steps_per_second": 1.534,
"eval_wo_beta": 7.4231791496276855,
"total_flos": 0.0,
"train_loss": 0.4483427975394509,
"train_runtime": 15087.4424,
"train_samples": 50802,
"train_samples_per_second": 16.836,
"train_steps_per_second": 0.117
}