→ Arquivo API + Criar uma imagem docker e colocar o container no Cloud Run
criar o arquivo fastapi
habilitar o Uvicorn
estabelecer as pastas de arquivos necessárias para a imagem docker
→ colocar o UI em produção ( criar um repositório diferente?) (heroku ou streamlit cloud? )
For now, all the code that we have been using for data exploration (notebooks), model training and building our prediction API is contained in a single project.
We are going to create a separate project for our website in order to keep things organised.
🤔 Why would we want to split our code into 2 projects ?
👉 Separating the code for the model training + prediction API from the code of the website will allow the package for the website to be very light👉 Using separate projects will also ease the deployment of the interface
Actually, the package for our website will not require to contain any Data Science related code, since the website will be using our own API in order to make predictions.
Splitting the code between training/prediction and website has several benefits:
We will be able to deploy our small package on light hosting solutions such as GitHub Pages and Heroku, which can operate for free 💵
Splitting the complexity will allow other team of developers (for example web developers) to work with us without requiring any Data Science related knowledge
It follows the web development pattern of separating the Front-End code (the website) from the Back-End code (the service), both communicating through an API
→ Montar estrutura das pastas
→ Criar projeto no GC
→ ML flow:
→ Arquivo API + Criar uma imagem docker e colocar o container no Cloud Run
→ colocar o UI em produção ( criar um repositório diferente?) (heroku ou streamlit cloud? )
For now, all the code that we have been using for data exploration (notebooks), model training and building our prediction API is contained in a single project.
We are going to create a separate project for our website in order to keep things organised.
🤔 Why would we want to split our code into 2 projects ?
👉 Separating the code for the model training + prediction API from the code of the website will allow the package for the website to be very light👉 Using separate projects will also ease the deployment of the interface
Actually, the package for our website will not require to contain any Data Science related code, since the website will be using our own API in order to make predictions.
Splitting the code between training/prediction and website has several benefits:
→ conectar a api com o UI