LTH14/mar
PyTorch implementation of MAR (Multiscale Autoregressive), an autoregressive image generation model using diffusion loss without vector quantization.
Not currently ranked — collecting fresh signals.
star history
This repository provides the official PyTorch implementation of the MAR model with DiffLoss for autoregressive image generation without vector quantization, as presented at NeurIPS 2024. The approach generates images token-by-token using a continuous diffusion loss instead of discrete quantization, combining benefits of both autoregressive and diffusion approaches. It includes pre-trained class-conditional models for ImageNet 256x256 and supports distributed training via PyTorch DDP.
Frequently asked
- What is LTH14/mar?
- PyTorch implementation of MAR (Multiscale Autoregressive), an autoregressive image generation model using diffusion loss without vector quantization.
- Is mar open source?
- Yes — LTH14/mar is open source, released under the MIT license.
- What language is mar written in?
- LTH14/mar is primarily written in Python.
- How popular is mar?
- LTH14/mar has 1.9k stars on GitHub.
- Where can I find mar?
- LTH14/mar is on GitHub at https://github.com/LTH14/mar.