Hi,
Microsoft just released the new set of Phi 3 models (inc quantized models) that are supposed to be a game changer in performance and quality in memory/cpu limited environments.
Im struggling a bit to make it work with Candle and I have a feeling its related to something in the model config that also changed.
I think supporting it with example can be tremendously valuable to the HF community.
Hi, Microsoft just released the new set of Phi 3 models (inc quantized models) that are supposed to be a game changer in performance and quality in memory/cpu limited environments. Im struggling a bit to make it work with Candle and I have a feeling its related to something in the model config that also changed. I think supporting it with example can be tremendously valuable to the HF community.
https://huggingface.co/microsoft/Phi-3-mini-4k-instruct-gguf
Thanks