OpenELM-1_1B-DPO-full-2 / all_results.json
CharlesLi's picture
End of training
8c9b91c verified
raw
history blame contribute delete
696 Bytes
{
"epoch": 2.998430141287284,
"eval_logits/chosen": 3.59375,
"eval_logits/rejected": 5.5,
"eval_logps/chosen": -1144.0,
"eval_logps/rejected": -1336.0,
"eval_loss": 0.7945234179496765,
"eval_rewards/accuracies": 0.732421875,
"eval_rewards/chosen": -8.3125,
"eval_rewards/margins": 2.140625,
"eval_rewards/rejected": -10.4375,
"eval_runtime": 46.9097,
"eval_samples": 2000,
"eval_samples_per_second": 42.635,
"eval_steps_per_second": 0.682,
"total_flos": 0.0,
"train_loss": 0.26257153015070145,
"train_runtime": 12376.5138,
"train_samples": 61134,
"train_samples_per_second": 14.819,
"train_steps_per_second": 0.231
}