Simple Fast API example to load a PyTorch model and score an input tensor (vector of 8 values).

• Install just lib
• launch

just build

to build the image locally

• launch

just run

to run the docker image locally.

Then hit

http://0.0.0.0:8000/docs

and POST a data vector of 8 elements as follow

{
  "data": [
    0.1,0.2,0.3,0.4,0.1,0.2,0.3,0.4
  ]
}

You should see a prediction showing up

{
  "predictions": [
    0.09620699286460876
  ]
}

• You can also retrain the model using different parameters in the trainer.py file.

Enjoy!