运行复现代码，为什么报错？

Shenzhi-Wang / Llama3-Chinese-Chat

This is the first Chinese chat model specifically fine-tuned for Chinese through ORPO based on the Meta-Llama-3-8B-Instruct model.

300 stars 16 forks source link

运行复现代码，为什么报错？ #13

Open dignfei opened 3 months ago

dignfei commented 3 months ago

数据格式明显是错的

"dpo_mix_en": { "hf_hub_url": "hiyouga/DPO-En-Zh-20k", "subset": "en", "ranking": true, "formatting": "sharegpt", "columns": { "messages": "conversations", "chosen": "chosen", "rejected": "rejected" } }, "dpo_mix_zh": { "hf_hub_url": "hiyouga/DPO-En-Zh-20k", "subset": "zh", "ranking": true, "formatting": "sharegpt", "columns": { "messages": "conversations", "chosen": "chosen", "rejected": "rejected" } },

dignfei commented 3 months ago

数据的格式，和 dataset_info.json 中设置的格式不一样

hiyouga commented 3 months ago

你的代码没有更新到最新

dignfei commented 3 months ago

你的代码没有更新到最新

我更新到了复现代码中指定的版本 git clone https://github.com/hiyouga/LLaMA-Factory.git git reset --hard 25aeaae51b6d08a747e222bbcb27e75c4d56a856
没有用最新版，因为最新版不支持orpo参数

dignfei commented 3 months ago

[rank4]: ValueError: Some specified arguments are not used by the HfArgumentParser: ['--pref_loss:', 'orpo', '--orpo_beta', '0.05']