zephyr-7b-dpo-lora-r16-20k / all_results.json
LaoRay's picture
End of training
670079e verified
raw
history blame
767 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": -2.553934335708618,
"eval_logits/rejected": -2.510233163833618,
"eval_logps/chosen": -362.36248779296875,
"eval_logps/rejected": -395.8988952636719,
"eval_loss": 0.5367478132247925,
"eval_rewards/accuracies": 0.7103174328804016,
"eval_rewards/chosen": -0.7912437319755554,
"eval_rewards/margins": 0.6874489188194275,
"eval_rewards/rejected": -1.4786925315856934,
"eval_runtime": 165.9175,
"eval_samples": 500,
"eval_samples_per_second": 3.014,
"eval_steps_per_second": 0.38,
"total_flos": 0.0,
"train_loss": 0.5874967678070069,
"train_runtime": 15864.6959,
"train_samples": 20000,
"train_samples_per_second": 1.261,
"train_steps_per_second": 0.079
}