Fast Transformer

Simplifying transformer model integration into FastAPI applications for seamless NLP development.

Simplified microservice approach to deploy models with fastAPI wrapper

Uses transformer and pytorch (support for configuring tensorflow coming soon..)

Prerequisite

• Python
• Poetry

Running in local

uvicorn app.main:app 

If everything is fine, you should be able to see output as below

INFO:     Started server process [38901]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)

example request

curl --location 'http://localhost:8000/analyze-sentiments/' \
--header 'Content-Type: application/json' \
--data '["bad programmer"]'

example response

[
    {
        "label": "LABEL_1",
        "score": 0.5389513969421387
    }
]

Includes Dockerfile - deploy as a microservice in cloud / kubernetes

Configuration

1. use MODEL_NAME env variable to use different model, no need of re creating image

export MODEL_NAME="distilbert-base-uncased"