How did you convert it?

#1
by vbuhoijymzoi - opened

Thanks for the upload.
Can you explain how did you convert it? llama.cpp refuses to convert it for me. See: https://huggingface.co/ise-uiuc/Magicoder-S-DS-6.7B/discussions/2

Happy to help! I used the command provided in the other thread. I always make sure I pull latest master and re-build with llama.

(llama.cpp)$ ./quantize --allow-requantize Magicoder-S-DS-6.7B_q8_0.gguf <output_requantized_model>.gguf q4_k_m

The issue I'm having is that llama.cpp refused to convert Magicoder's safetensors format into 16bit gguf for further quantization.
Where did you get q8 quant from?

Sign up or log in to comment