Project Overview

In this project, we'll build a translation API using deep learning. Using FastAPI, we'll create a web server that exposes a /translate route and a /results route. Clients will post their translation request to the /translate route, and get the translation results from /results. The server will use a sqlite database to store the translations. On the backend, we'll use async and a pretrained deep learning language model to run the translation job.

By the end, we'll have a web server that can run translation jobs quickly. This server can easily be extended to translate more languages, or add more options.

Project Steps

• Build API routes
• Add in models to store data to database
• Create tasks to run the translation

Code

You can find the code for this project here

File overview:

• requirements.txt - packages you'll need to install
• languages.txt - list of languages that are supported for translation
• main.py - defines the web server routes
• models.py - defines database models
• tasks.py - runs our backend tasks, including the translation

Prerequisites

To complete this project, you'll need to have a good understanding of:

• Python syntax, including functions, if statements, and data structures
• Data cleaning
• Pandas syntax
• Using Jupyter notebook
• APIs
• The basics of machine learning.

Please make sure you've completed these Dataquest courses (or know the material) before trying this project:

• Python Introduction
• For Loops and If Statements
• Dictionaries In Python
• Functions and Jupyter Notebook
• Python Intermediate
• Pandas and NumPy Fundamentals
• Data Cleaning
• Machine Learning Fundamentals
• APIs and Web Scraping

Local Setup

Installation

To follow this project, please install the following locally:

• Python 3.8+
• The packages defined in requirements.txt
• It's recommended to install PyCharm or VSCode.

Usage

To run the server locally, run uvicorn main:app --reload.

Data

During this project, we'll download a pretrained language model.

Deploying

Deploy with Docker

One strategy is to deploy to the cloud using Docker. To do this, we first need to build a Docker image. Then we can deploy the image to the cloud.

Docker image

To build and test a Docker image, run:

• docker build -t dlapi . to build the container.
  • If you're not running on 64-bit linux, instead run docker buildx build --platform linux/amd64 -t dlapi .. This will build the image using the correct architecture for Azure.
• docker run -d --name dlapi -p 80:80 dlapi to run the container.
• docker ps to view the container information.
• Run docker logs to see logs from the container. You should see Uvicorn running on http://0.0.0.0:80. If you don't see this, wait a bit and try running docker logs again.
• Visit 127.0.0.1 or localhost to see the API server. Visit localhost/docs to see API docs.
• Run docker stop dlapi to stop the container.

Azure setup

• Create an Azure account. You can create one here.
• Create an Azure subscription if you don't have one. You can sign up for the free tier.
• Install the Azure CLI.
• Run NAME="<name-here>" to set the name for your resource group. Replace <name-here> with the name you want.
• Create a resource group with az group create --location eastus --name $NAME

Push image to Azure

• Create a container registry with az acr create --resource-group $NAME --name $NAME --sku Basic
• Note the loginServer in the response.
• Login to the registry with az acr login --resource-group $NAME --name $NAME
• Run docker tag dlapi $NAME.azurecr.io/dlapi. Replace $NAME.azurecr.io with your container registry login url.
• Push the image to the registry with docker push $NAME.azurecr.io/dlapi.

Run container on Azure

• Run docker login azure to log Docker into Azure.
• Create an aci context with docker context create aci azure. Select the resource group to use with your context.
• Run docker context ls to view your contexts.
• Switch context with docker context use azure
• Run the container with docker run --name dlapi -p 80:80 -m 1.5G $NAME.azurecr.io/dlapi
• Run docker ps to get the URL of the container. You should be able to visit the URL shown to use the API.

Remove container

• Run docker stop dlapi
• Run docker rm dlapi
• Delete the resource group with az group delete --name $NAME --no-wait

Deploy with Azure App Service

Another strategy is to deploy with Azure App Service.

Deploy App

• Create an Azure account.
• Create an Azure subscription if you don't have one. You can sign up for the free tier.
• Install the Azure CLI.
• Login to Azure with az login.
• NAME="dlapi"
• Create a resource group with az group create --location eastus --name $NAME
• Create a plan with az appservice plan create --name $NAME --resource-group $NAME --sku B2
• Create app with az webapp create --name $NAME --resource-group $NAME --plan $NAME --runtime PYTHON:3.9
• az webapp up --runtime PYTHON:3.9 --sku B2 --logs
• Get the resource group name and the app name from the log output.
• az webapp config appsettings set --resource-group $NAME --name $NAME --settings SCM_DO_BUILD_DURING_DEPLOYMENT=true POST_BUILD_COMMAND="build.sh" WEBSITE_MEMORY_LIMIT_MB=3500
• az webapp config set --resource-group $NAME --name $NAME --startup-file "run.sh"
• zip -r app.zip . -x translations.db app.zip __pycache__/*
• az webapp deploy --resource-group $NAME --name $NAME --src-path app.zip
• az webapp restart --resource-group $NAME --name $NAME

Manage App

• View the app using the URL from the logs.
• Delete the app with az group delete --name $NAME --no-wait