llama-stack
llama-stack copied to clipboard
fbgemm-gpu isn't officially supported on mac - optional dependency?
Can we make fbgemm-gpu
an optional dependency? https://pypi.org/project/fbgemm-gpu/#files It doesn't look like it's supported on a mac https://github.com/pytorch/FBGEMM/issues/1985. This means means the build is broken out of the box. Without it, it looks like we break FP8 quantization.
https://github.com/meta-llama/llama-stack/blob/eb2d8a31a5927589197c794855d7323f8f4700bc/llama_stack/providers/impls/meta_reference/inference/generation.py#L87
They had to make it optional in torchrec too: https://github.com/pytorch/torchrec/issues/1483
ERROR: Could not find a version that satisfies the requirement fbgemm-gpu==0.8.0 (from versions: none)
ERROR: No matching distribution found for fbgemm-gpu==0.8.0
Failed to build target my-local-stack with return code 1