lm-sys/FastChat

fastchat-t5-3b-v1.0 on macOS?

Open

#646 opened on Apr 29, 2023

View on GitHub
 (15 comments) (7 reactions) (1 assignee)Python (4,736 forks)batch import
buggood first issue

Repository metrics

Stars
 (38,959 stars)
PR merge metrics
 (No merged PRs in 30d)

Description

Hi, I'm trying to use fastchat-t5-3b-v1.0 on macOS following the instructions in the README.

Simply run the line below to start chatting. It will automatically download the weights from a Hugging Face repo.
python3 -m fastchat.serve.cli --model-path lmsys/fastchat-t5-3b-v1.0

When I do this, I get the error "Torch not compiled with CUDA enabled".

I added --device mps --load-8bit to run it as

python3 -m fastchat.serve.cli --model-path lmsys/fastchat-t5-3b-v1.0 --device mps --load-8bit

And that gives a different error: "Unrecognized configuration class <class 'transformers.models.t5.configuration_t5.T5Config'> for this kind of AutoModel: AutoModelForCausalLM".

Is this usable without CUDA?

Contributor guide

Research direction
Investigate why AutoModelForCausalLM is used with a T5 model and how to correctly load the model on non CUDA devices like MPS. Check if there is a need to use a different model class or if the load 8bit flag is incompatible with MPS.
Tech stack
pythonpytorch
Domain
backend
Issue type
Bug
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
PythonPyTorch
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60