Feature request of quantization of starchat-alpha

#2
by delas - opened

Hi!

Firstly, thank you for your work implementing starcoder for ggml. Working fine with the 4 bit quantized version.

I would like to make a feature request related with the implementation of the 4 bit quantize version of starchat-alpha for enabling the usage with ggml.

Best Regards,

Jordi

Hello

How to run that starcoder ggml? I tried with llama.cpp and koboldcpp ... no working

This comment has been hidden

Thank you .
Is working so bad for you like for me?

I'm asking for c++ Fibonacci code and getting always for python.... and other ask also is giving me python code....

I suppose starcoder is not optimized for chat interaction but for autocompletion, that's why the finetuned variation starchat-alpha probably is more useful for interaction purposes.

Oh I see . Thanks .

Sign up or log in to comment