fostroll / modelsrv0

The FastAPI wrapper for ML model

modelsrv0

The simple FastAPI wrapper for ML model.

Requirements

pip install python-multipart
pip install fastapi
pip install uvicorn
pip install passlib
pip install python-jose

Usage

1. Edit config.yaml. At least, you wanna change jwt_secret_key value with

openssl rand -hex 512

2. Edit user list in config_users.yaml. The hash for password key may be generated with

python make_hash.py <password>

3. Replace script server/model/spec.py with your own, that is implemented methods model_load(path, device='cpu') and model_predict(text, with_intents=True, probs=True, threshold=.5, only_true=False). Scripts server/model/model*.py may be removed or replaced with your own model definition.

4. Put the model in the directory model0 (or change the model_name in config.yaml).

5. Edit run.sh or run.bat and start the server.

License

modelsrv0 is released under the Creative Commons License. See the LICENSE file for more details.