Fix compatibility with pipeline when loading model with device_map on single gpu #12443

	name: Run Torch dependency tests

	on:
	pull_request:
	branches:
	- main
	paths:
	- "src/diffusers/**.py"
	push:
	branches:
	- main

	concurrency:
	group: ${{ github.workflow }}-${{ github.head_ref \|\| github.run_id }}
	cancel-in-progress: true

	jobs:
	check_torch_dependencies:
	runs-on: ubuntu-22.04
	steps:
	- uses: actions/checkout@v3
	- name: Set up Python
	uses: actions/setup-python@v4
	with:
	python-version: "3.8"
	- name: Install dependencies
	run: \|
	python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
	python -m pip install --upgrade pip uv
	python -m uv pip install -e .
	python -m uv pip install torch torchvision torchaudio
	python -m uv pip install pytest
	- name: Check for soft dependencies
	run: \|
	python -m venv /opt/venv && export PATH="/opt/venv/bin:$PATH"
	pytest tests/others/test_dependencies.py

Provide feedback