GPTQ / Python - Test #1493

	name: GPTQ / Python - Test

	on:
	workflow_dispatch:
	push:
	branches: [main]
	paths:
	- tests/gptq/**
	- optimum/gptq/**
	- .github/workflows/test_gptq.yml
	pull_request:
	branches: [main]
	paths:
	- tests/gptq/**
	- optimum/gptq/**
	- .github/workflows/test_gptq.yml
	schedule:
	# every day at midnight
	- cron: "0 0 * * *"

	jobs:
	test_gptq:
	runs-on:
	group: aws-g6-4xlarge-plus

	steps:
	- name: Checkout code
	uses: actions/checkout@v4

	- name: Run tests
	uses: addnab/docker-run-action@v3
	with:
	image: pytorch/pytorch:2.2.2-cuda12.1-cudnn8-runtime
	# latest auto-gptq was built with pytorch 2.2 and cuda 12.1
	options: \|
	--rm
	--gpus all
	--shm-size 16G
	--env RUN_SLOW=1
	--env HF_HOME=/mnt/cache/
	--volume /mnt/cache/:/mnt/cache/
	--volume ${{ github.workspace }}:/workspace
	--workdir /workspace
	run: \|
	pip install auto-gptq
	pip install -e .[tests]
	pytest tests/gptq -s -vvvv --durations=0

Provide feedback