https://github.com/answerdotai/minai
A miniture AI training framework for PyTorch
https://github.com/answerdotai/minai
Last synced: about 1 month ago
JSON representation
A miniture AI training framework for PyTorch
- Host: GitHub
- URL: https://github.com/answerdotai/minai
- Owner: AnswerDotAI
- License: apache-2.0
- Created: 2023-09-16T19:44:32.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2025-02-01T03:30:38.000Z (about 1 year ago)
- Last Synced: 2025-12-01T23:38:41.867Z (4 months ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 1.69 MB
- Stars: 42
- Watchers: 4
- Forks: 9
- Open Issues: 5
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
Awesome Lists containing this project
README
# minai
The mini version of fastai's miniai PyTorch framework created during the fastai course 2022-2023.
## Installation
```bash
pip install minai
```
or to install from source clone this repo and run:
```bash
pip install -e .
```
## Usage
This is still a work in progress - I'll add example usage soon. But in general, for examples from the course where you have `from miniai.something import X` you should be able to do `from minai import X`. You can do `import minai as mi` or even `from minai import *` for quick access to all the functions and things, if you're so inclined.
Tutorial 1 has a minimal example of fitting a model using minai - open it in Google colab [here](https://colab.research.google.com/github/AnswerDotAI/minai/blob/main/tutorial_01.ipynb).
Tutorial 2 shows callbacks in action on a slightly more complex task - open it in Google colab [here](https://colab.research.google.com/github/AnswerDotAI/minai/blob/main/tutorial_02.ipynb).
An example of the library in action: [this notebook](https://colab.research.google.com/drive/1b3CeZB2FfRGr5NPYDVvk34hyZFBtgub5?usp=sharing) shows how to train a diffusion model on spectrograms to generate birdcalls, using minai. It is covered in the final lesson of Part 2 of the FastAI course.
And a lovely demo of use in the wild is [this report by Thomas Capelle](https://wandb.ai/capecape/miniai_ddpm/reports/Next-Frame-Prediction-Using-Diffusion-The-fastai-Approach--VmlldzozMzcyMTYy) where he uses diffusion models to predict the next frame of an image sequence.