Support For ggml format for gpt4all

[chigkim]

When I convert Llama model with convert-pth-to-ggml.py, quantize to 4bit, and load it with gpt4all, I get this:
llama_model_load: invalid model file 'ggml-model-q4_0.bin' (bad magic)
Could you implement to support ggml format that gpt4all uses?
Thanks!

[linouxis9]

Try running the migrate script as detailed in this PR: #690

[chigkim]

Isn't that migrate from old format to new format? I would like to convert pth to old format. Thanks!

[prusnak]

You should open an issue in the gpt4all repository, since what you are asking is essentially them to support a certain format.