🌱 Unify launcher unit testing

.github/workflows/python-code-quality.yml

@@ -34,7 +34,7 @@ jobs:
           # install black isort flake
           pip install black==24.4.2 isort==5.13.2 flake8==7.0.0
           # install pytest and other dependencies
-          pip install pytest==9.0.2 fastapi==0.116.1 pydantic==2.11.7 uvicorn==0.35.0 uvloop==0.21.0 httpx==0.28.1 nvidia-ml-py==13.590.48 kubernetes==31.0.0
+          pip install pytest==9.0.2 -r inference_server/launcher/requirements-sans-vllm.txt
 
       - name: Check for trailing whitespace
         run: |

inference_server/launcher/howto.md

@@ -1,13 +1,14 @@
 1- HOW TO RUN THE UNIT TEST:
 
-Install all the necessary packages:
+Install all the necessary packages (feel free to use a later version of pytest if you prefer):
 ```bash
-pip install -r requirements.txt
+pip install pytest==9.0.2 -r requirements-sans-vllm.txt
 ```
 
 Run the unit test doing:
 ```bash
-python -m pytest test_launcher.py -v
+python -m pytest tests/test_launcher.py -v
+python -m pytest tests/test_gputranslator.py -v
 ```
 
 2- RUN E2E TEST:

inference_server/launcher/requirements-sans-vllm.txt

@@ -0,0 +1,7 @@
+fastapi==0.135.1
+pydantic==2.12.5
+uvicorn==0.42.0
+uvloop==0.22.1
+httpx==0.28.1
+nvidia-ml-py==13.590.48
+kubernetes==31.0.0

inference_server/launcher/requirements.txt

@@ -1,8 +1,3 @@
-fastapi
-pydantic
-uvicorn
-uvloop
-nvidia-ml-py
-kubernetes==31.0.0
+-r requirements-sans-vllm.txt
 # WARNING: vllm must be built from source on a macOS Silicon
 vllm; sys_platform != "darwin" or platform_machine != "arm64"