Llama3.1-Mamba2-8B-dpo / all_results.json
Junxiong Wang
add models
7ae6913
raw
history blame contribute delete
764 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": -1.3477604389190674,
"eval_logits/rejected": -1.3952159881591797,
"eval_logps/chosen": -486.49871826171875,
"eval_logps/rejected": -697.171875,
"eval_loss": 0.41439929604530334,
"eval_rewards/accuracies": 0.8339285850524902,
"eval_rewards/chosen": -2.313478469848633,
"eval_rewards/margins": 2.180910587310791,
"eval_rewards/rejected": -4.494389057159424,
"eval_runtime": 176.8003,
"eval_samples": 4461,
"eval_samples_per_second": 25.232,
"eval_steps_per_second": 0.792,
"total_flos": 0.0,
"train_loss": 0.4822349088434523,
"train_runtime": 13347.0673,
"train_samples": 133368,
"train_samples_per_second": 9.992,
"train_steps_per_second": 0.312
}