https://github.com/visheshc14/stablediffusion-fastapi

NimbleBox Apprenticeship ML Engineer Task - 2. Stable diffusion doesn't actually apply noise to images, it uses a compressed low dimensional latent representation of the image and applies noise to that. The model is running in this abstract latent space, and then the autoencoder recreates the image afterwards.
https://github.com/visheshc14/stablediffusion-fastapi

Last synced: 5 days ago
JSON representation

NimbleBox Apprenticeship ML Engineer Task - 2. Stable diffusion doesn't actually apply noise to images, it uses a compressed low dimensional latent representation of the image and applies noise to that. The model is running in this abstract latent space, and then the autoencoder recreates the image afterwards.

• Host: GitHub
• URL: https://github.com/visheshc14/stablediffusion-fastapi
• Owner: visheshc14
• Created: 2023-06-01T09:45:36.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-08-30T03:37:09.000Z (over 1 year ago)
• Last Synced: 2024-11-18T13:28:57.220Z (2 months ago)
• Language: JavaScript
• Homepage:
• Size: 1.12 MB
• Stars: 4
• Watchers: 2
• Forks: 0
• Open Issues: 6
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Stable Diffusers FastAPI Application

This project is a FastAPI application that serves a Stable Diffusers model over HTTP/REST APIs. It allows users to generate images using the Stable Diffusion technique.

## Features

- Generate images using the Stable Diffusion technique.

- RESTful API endpoints to interact with the model.

- Integration with Pydantic for input data validation.

- Streaming response to efficiently serve generated imag

## Setup

- Move to the backend folder.

- Run the following:

    ```bash

    cd backend 

    touch .env 

    ```

- Open the .env file, and paste your HF token:

    ```bash

    HF_TOKEN=YOUR_TOKEN {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}

    ```

- Install requirements:

    ```bash

    pip install -r requirements.txt

    ```

- Start your server:

    ```bash

    uvicorn main:app --port 8000

    ```

- Run the following CURL command:

    ```bash

    curl -X GET "localhost:8000/api/generate" -H "Content-Type: image/png" -d '{"image": "https://huggingface.co/front/thumbnails/stable_diffusion.png"}'

    ```

 - Change to the frontend folder:

    ```bash

    cd ../frontend

    ```

- Install front end dependencies and start the server:

    ```bash

    npm install

    npm start

    ```

    

## Crux: ML Engineer

### Fine tune a Stable Diffusion model and serve it

- create a github repo

- create a `trainer.py` file that takes in arguments like file path `--fp` and some other training arguments and fintunes a stable diffusion model (protip: check it out on `huggingface`)

- the result of the training should be a model weight file

- create a file called `server.py` that serves the Model over a HTTP/REST over some APIs (protip: use `pydantic` for models)

- A `curl` command to call the model and get response

- an ipython notebook that contains the steps to run this