Remove vLLM test dependency

vLLM can only currently be run on GPU (unless you want to go to extreme
lengths to make it work on CPU), and we thus cannot run the tests in CI.
We thus separate the test dependencies in "with GPU" and "without
GPU" (a subset of "with GPU") that the user has to pick manually.

Author: rlouf

docs/community/contribute.md

@@ -18,7 +18,7 @@ Note that the [issue tracker][issues] is only intended for actionable items. In
 
 First, [fork the repository on GitHub](https://github.com/dottxt-ai/outlines/fork) and clone the fork locally:
 
-```bash
+```shell
 git clone [email protected]/YourUserName/outlines.git
 cd outlines
 ```
@@ -27,35 +27,49 @@ Create a new virtual environment:
 
 *If you are using `uv`*:
 
-```bash
+```shell
 uv venv
 source .venv/bin/activate
 alias pip="uv pip" # ... or just remember to prepend any pip command with uv in the rest of this guide
 ```
 
 *If you are using `venv`*:
 
-```bash
+```shell
 python -m venv .venv
 source .venv/bin/activate
 ```
 
 *If you are using `conda`*:
 
-```bash
+```shell
 conda env create -f environment.yml
 ```
 
 Then install the dependencies in editable mode, and install the `pre-commit` hooks:
 
-```bash
+```shell
+python -m venv .venv
+source .venv/bin/activate
+```
+
+Then install the dependencies in editable mode, and install the pre-commit hooks:
+
+```shell
 pip install -e ".[test]"
 pre-commit install
 ```
+If you own a GPU and want to run the vLLM tests you will have to run:
+
+```shell
+pip install -e ".[test-gpu]"
+```
+
+instead.
 
 Outlines provides optional dependencies for different supported backends, which you can install with
 
-```bash
+```shell
 pip install ".[vllm]"
 ```
 
@@ -85,13 +99,13 @@ You will not have access to a GPU, but you'll be able to make basic contribution
 
 Run the tests:
 
-```bash
+```shell
 pytest
 ```
 
 And run the code style checks:
 
-```bash
+```shell
 pre-commit run --all-files
 ```
 
@@ -101,7 +115,7 @@ Outlines uses [asv](https://asv.readthedocs.io) for automated benchmark testing.
 
 You can run the benchmark test suite locally with the following command:
 
-```bash
+```shell
 asv run --config benchmarks/asv.conf.json
 ```
 
@@ -112,19 +126,19 @@ Caveats:
 
 #### Run a specific test:
 
-```bash
+```shell
 asv run --config benchmarks/asv.conf.json -b bench_json_schema.JsonSchemaBenchmark.time_json_schema_to_fsm
 ```
 
 #### Profile a specific test:
 
-```bash
+```shell
 asv run --config benchmarks/asv.conf.json --profile -b bench_json_schema.JsonSchemaBenchmark.time_json_schema_to_fsm
 ```
 
 #### Compare to `origin/main`
 
-```bash
+```shell
 get fetch origin
 asv continuous origin/main HEAD --config benchmarks/asv.conf.json
 ```
@@ -140,13 +154,13 @@ asv continuous origin/main HEAD --config benchmarks/asv.conf.json
 
 To work on the *documentation* you will need to install the related dependencies:
 
-```bash
+```shell
 pip install -r requirements-doc.txt
 ```
 
 To build the documentation and serve it locally, run the following command in the repository's root folder:
 
-```bash
+```shell
 mkdocs serve
 ```
 
@@ -157,7 +171,7 @@ It will be updated every time you make a change.
 
 Create a new branch on your fork, commit and push the changes:
 
-```bash
+```shell
 git checkout -b new-branch
 git add .
 git commit -m "Changes I made"

pyproject.toml

@@ -69,12 +69,12 @@ test = [
     "huggingface_hub",
     "openai>=1.0.0",
     "datasets",
-    "vllm; sys_platform == 'linux'",
     "transformers",
     "pillow",
     "exllamav2",
     "jax",
 ]
+test-gpu=["outlines[test]", "vllm; sys_platform == 'linux'"]
 serve = [
     "vllm>=0.3.0",
     "uvicorn",