madaan/minimal-text-diffusion
A minimal implementation of diffusion models for text generation, training on text corpora to generate samples.
Not currently ranked — collecting fresh signals.
star history
The repository provides a stripped-down implementation of denoising diffusion probabilistic models (DDPMs) adapted for text generation. It retains essential training and sampling code from the Diffusion-LM project, removing image-related components. Users can train diffusion models on custom text corpora and generate text samples through iterative denoising. The project includes simple test datasets for quick experimentation and supports controllable generation.
Frequently asked
- What is madaan/minimal-text-diffusion?
- A minimal implementation of diffusion models for text generation, training on text corpora to generate samples.
- Is minimal-text-diffusion open source?
- Yes — madaan/minimal-text-diffusion is open source, released under the MIT license.
- What language is minimal-text-diffusion written in?
- madaan/minimal-text-diffusion is primarily written in Python.
- How popular is minimal-text-diffusion?
- madaan/minimal-text-diffusion has 416 stars on GitHub.
- Where can I find minimal-text-diffusion?
- madaan/minimal-text-diffusion is on GitHub at https://github.com/madaan/minimal-text-diffusion.