Text Generation
Transformers
Safetensors
mistral
alignment-handbook
Generated from Trainer
conversational
text-generation-inference
mistral-7b-anthropic / eval_results.json
vwxyzjn's picture
Add HuggingFaceH4/mistral-7b-dpo-v21.0cai.3.2 checkpoint
e62786b verified
raw
history blame
574 Bytes
{
"epoch": 3.0,
"eval_logits/chosen": -2.35003924369812,
"eval_logits/rejected": -2.296473503112793,
"eval_logps/chosen": -278.26385498046875,
"eval_logps/rejected": -376.85308837890625,
"eval_loss": 0.03253357484936714,
"eval_rewards/accuracies": 0.9613811373710632,
"eval_rewards/chosen": -2.0301272869110107,
"eval_rewards/margins": 12.106389045715332,
"eval_rewards/rejected": -14.136518478393555,
"eval_runtime": 2927.9433,
"eval_samples": 3156,
"eval_samples_per_second": 22.703,
"eval_steps_per_second": 0.355
}