search

Ngz91 / Eunomia

Analyze your code locally using a GPT LLM.
Apache License 2.0
32 stars 9 forks source link

Not all listed models work #1

Closed Beenhakker closed 1 year ago

Beenhakker commented 1 year ago

Hi,

First off: thanks for this. Exactly what I wanted to build myself.

Second: not all listed models seem to be working on my Mac (Intel).

Working:

  1. ggml-gpt4all-j-v1.3-groovy.bin

Not working: (gptj_model_load: invalid model file (bad magic))

  1. ggml-gpt4all-l13b-snoozy.bin
  2. ggml-nous-gpt4-vicuna-13b.bin
Ngz91 commented 1 year ago

Hello, after testing locally on a Windows machine (Don't have access to a Mac) I was able to replicate the error with ggml-nous-gpt4-vicuna-13b.bin.

badMagicError

I suspect that it could be the backend that you are using when loading the model. In my case I tried to load the model using the gptj backend, but it only works with the llama backend.

Try changing the backend and let me know if that solves your problem.