simpo-oh_v3.1_wo_camel_ai_math / all_results.json
sedrickkeh's picture
End of training
12e3fb1 verified
raw
history blame contribute delete
741 Bytes
{
"epoch": 0.9997172745264349,
"eval_logits/chosen": -1.1595183610916138,
"eval_logits/rejected": -1.1711369752883911,
"eval_logps/chosen": -4.307417392730713,
"eval_logps/rejected": -5.051736831665039,
"eval_loss": 2.3919546604156494,
"eval_rewards/accuracies": 0.8038102388381958,
"eval_rewards/chosen": -43.07417297363281,
"eval_rewards/margins": 7.443192958831787,
"eval_rewards/rejected": -50.517372131347656,
"eval_runtime": 100.3048,
"eval_samples_per_second": 29.699,
"eval_steps_per_second": 1.864,
"total_flos": 134366991482880.0,
"train_loss": 3.293350306571339,
"train_runtime": 7617.916,
"train_samples_per_second": 7.429,
"train_steps_per_second": 0.058
}