Find answers from the community

Updated 6 months ago

I'm creating a FastAPI to use

At a glance

The community member is creating a FastAPI application that uses VectorStoreIndex with HuggingFace LLM and embed model. When running the application on their local Mac M1, they are encountering a "RuntimeError: MPS backend out of memory" error. One of the comments suggests that this issue has been discussed on the PyTorch forum, but there is no explicitly marked answer provided.

Useful resources

RRitually

I'm creating a FastAPI to use VectorStoreIndex with HuggingFace LLM and embed model. When I am running the FastAPI app on my local Mac M1 and sending a request to it, I am getting this error:

Plain Text

RuntimeError: MPS backend out of memory

. Has anyone dealt with this before?

2 comments

bbmax

this is crazy.

bbmax

https://discuss.pytorch.org/t/mps-backend-out-of-memory/183879

Add a reply