为什么我出来的效果非常糟糕呢？ #162
Replies: 4 comments 3 replies
-
|
貌似加载的是LLaMA模型？对话问答的话用Alpaca模型体验会更好。 |
Beta Was this translation helpful? Give feedback.
-
|
7B Alpaca 模型，效果也十分糟糕，基本处于不可用的情况。13B之后测试。
|
Beta Was this translation helpful? Give feedback.
-
|
效果还可以的，你是不是没用alpaca prompt模板 |
Beta Was this translation helpful? Give feedback.
-
|
额，这个效果一言难尽。和会不会关系不大，感觉对中文的指令理解有点错乱。用的网友提供的merge后的13B模型。 |
Beta Was this translation helpful? Give feedback.
-
|
我也是这种情况，胡言乱语 |
Beta Was this translation helpful? Give feedback.
-
|
试试这个，基于gpt-x-vicuna做的，效果还不错 |
Beta Was this translation helpful? Give feedback.
-
如果只用chinese llama alpaca 由於我都拿來用代碼 所以只問代碼問題 😂 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
我先把facebook原始模型转成hf格式，然后再使用merge_llama_with_chinese_lora_to_hf.py合并模型，两个模型都是13B的。最后使用transformers做推理。好像看起来非常糟糕。
Beta Was this translation helpful? Give feedback.
All reactions