add .envdefault and docker-compose.yml

.envdefault

@@ -0,0 +1,4 @@
+APP_LANG=fr en
+ASSETS_PATH_ON_HOST=./assets
+ASSETS_PATH_IN_CONTAINER=/app/assets
+WORKER_NUMBER=1

.gitignore

@@ -1 +1,2 @@
-assets
+assets
+.env

Dockerfile

@@ -3,16 +3,14 @@ LABEL maintainer="[email protected]"
 
 WORKDIR /app
 
-VOLUME /app/assets
-ENV ASSETS_PATH=/app/assets
-
 COPY ./requirements.txt /app/
 RUN pip install --no-cache-dir -r requirements.txt
 
+COPY .envdefault /app/
 COPY ./scripts /app/scripts
 COPY ./components /app/components
 
 HEALTHCHECK --interval=15s CMD curl -fs http://0.0.0.0/health || exit 1
 
-ENTRYPOINT ["/home/user/miniconda/bin/uvicorn", "scripts.main:app", "--host", "0.0.0.0", "--port", "80"]
+ENTRYPOINT ["/home/user/miniconda/bin/gunicorn", "scripts.main:app", "--worker-class", "uvicorn.workers.UvicornWorker", "--bind", "0.0.0.0:80", "--access-logfile", "-", "--error-logfile", "-"]
 CMD ["--workers", "1"]

README.md

@@ -21,51 +21,51 @@ With our proposed stack [https://github.com/linto-ai/linto-platform-stack](https
 # Develop
 
 ## Build and run
-1 Create a named volume for storaging models.
-```bash
-sudo docker volume create linto-platform-nlp-assets
-```
-
-2 Download models into `assets/` on the host machine, make sure that `git-lfs`: [Git Large File Storage](https://git-lfs.github.com/) is installed and availble at `/usr/local/bin/git-lfs`.
+1 Download models into `./assets` on the host machine (can be stored in other places), make sure that `git-lfs`: [Git Large File Storage](https://git-lfs.github.com/) is installed and availble at `/usr/local/bin/git-lfs`.
 ```bash
 cd linto-platform-nlp-keyphrase-extraction/
 bash scripts/download_models.sh
 ```
 
-3 Copy downloaded models into created volume `linto-platform-nlp-assets`
+2 configure running environment variables
 ```bash
-sudo docker container create --name cp_helper -v linto-platform-nlp-assets:/root hello-world
-sudo docker cp assets/* cp_helper:/root
-sudo docker rm cp_helper
+mv .envdefault .env
+# cat .envdefault
+# APP_LANG=fr en | Running language of application, "fr en", "fr", etc.
+# ASSETS_PATH_ON_HOST=./assets | Storage path of models on host. (only applicable when docker-compose is used)
+# ASSETS_PATH_IN_CONTAINER=/app/assets | Volume mount point of models in container. (only applicable when docker-compose is used)
+# WORKER_NUMBER=3 | Number of processing workers. (only applicable when docker-compose is used)
 ```
 
 4 Build image
 ```bash
 sudo docker build --tag lintoai/linto-platform-keyphrase-extraction:latest .
 ```
+or
+```bash
+sudo docker-compose build
+```
 
-5 Run container (with GPU), make sure that [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#installing-on-ubuntu-and-debian) and GPU driver are installed.
+5 Run container with GPU support, make sure that [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#installing-on-ubuntu-and-debian) and GPU driver are installed.
 ```bash
 sudo docker run --gpus all \
---rm -d -p 80:80 \
--v linto-platform-nlp-assets:/app/assets:ro \
---env APP_LANG="fr en" \
-lintoai/linto-platform-keyphrase-extraction:latest
+--rm -p 80:80 \
+-v $PWD/assets:/app/assets:ro \
+--env-file .env \
+lintoai/linto-platform-keyphrase-extraction:latest \
+--workers 1
+```
+or
+```bash
+sudo docker-compose up
 ```
 <details>
   <summary>Check running with CPU only setting</summary>
 
-  ```bash
-sudo docker run \
---rm -d -p 80:80 \
--v linto-platform-nlp-assets:/app/assets:ro \
---env APP_LANG="fr en" \
-lintoai/linto-platform-keyphrase-extraction:latest
-  ```
+  - remove `--gpus all` from the first command.
+  - remove `runtime: nvidia` from the `docker-compose.yml` file.
 </details>
-To specify running language of the container, modify APP_LANG="fr en", APP_LANG="fr", etc.
 
-To lanche with multiple workers, add `--workers INTEGER` in the end of the above command.
 
 6 Navigate to `http://localhost/docs` or `http://localhost/redoc` in your browser, to explore the REST API interactively. See the examples for how to query the API.
 

docker-compose.yml

@@ -0,0 +1,13 @@
+version: "3.9"
+services:
+  kpe:
+    build: .
+    image: lintoai/linto-platform-keyphrase-extraction:latest
+    ports:
+      - "80:80"
+    env_file:
+      - .env
+    volumes:
+      - $ASSETS_PATH_ON_HOST:$ASSETS_PATH_IN_CONTAINER:ro
+    command: ["--workers", $WORKER_NUMBER]
+    runtime: nvidia

scripts/main.py

@@ -7,18 +7,23 @@
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi_health import health
 from sentence_transformers import SentenceTransformer
+from dotenv import load_dotenv
 
 # To force the GPU usage: spacy.require_gpu()
 spacy.prefer_gpu()
 
+# Parse environment variables.
+# variable and its value in the .envdefault file will be set, only if the variable is missing or empty in the current enviroment.
+load_dotenv(".envdefault", override=False)
+
 # Supported languages and corresponding model names
 LM_MAP = {
     "fr": "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
     "en": "sentence-transformers/all-MiniLM-L6-v2"
     }
 
 # Load models
-MODELS = {LM_MAP[lang]: SentenceTransformer(os.environ.get("ASSETS_PATH") + '/' + LM_MAP[lang]) for lang in os.environ.get("APP_LANG").split(" ")}
+MODELS = {LM_MAP[lang]: SentenceTransformer(os.environ.get("ASSETS_PATH_IN_CONTAINER") + '/' + LM_MAP[lang]) for lang in os.environ.get("APP_LANG").split(" ")}
 print(f"Loaded {len(MODELS)} models: {MODELS.keys()}")
 
 @spacy.registry.misc("get_model")