--- license: mit tags: - moe --- * [This is DPO improved version of cloudyu/Mixtral_11Bx2_MoE_19B](https://huggingface.co/cloudyu/Mixtral_11Bx2_MoE_19B) * [DPO Trainer](https://huggingface.co/docs/trl/main/en/dpo_trainer)