[NeurIPS 2025 Oral] Official Code for Exploring Diffusion Transformer Designs via Grafting

FastCache: Fast Caching for Diffusion Transformer Through Learnable Linear Approximation [Efficient ML Model]

edge-dit.cpp — a native C/C++ inference engine for Diffusion Transformers (DiT), designed for local and resource-constrained devices with automatic VRAM-aware precision, placement, and offloading.

Educational, modular, and high-performance Diffusion Transformers (DiT) in JAX and Flax NNX.

Official repository for the "LAViG-FLOW: Latent Autoregressive Video Generation for Fluid Flow Simulations" paper.

Curated DiT paper landscape: architecture, video, systems, RL, agents and omni generation

基于 bytetriper/RAE 的个人研究工作区，记录 Representation Autoencoders 相关实验；非官方仓库。

Official implementation of PrediT (ACM MM 2026): training-free linear multistep feature forecasting for efficient diffusion transformers.

Compute-optimal data selection, adaptation, evaluation, and systems engineering for video-generation models.

[NeurIPS 2025] Official code for "Exploring Diffusion Transformer Designs via Grafting"

This repository implements multiple generative diffusion frameworks (EDM, Consistency Models, etc.). It also implements some architectures (U-Net, Diffusion Transformers, etc.).

Evaluation of Diffusion & Transformer-Based Generative Models. Forked repository of the 18Dec2025 version of our group work from https://github.com/juhimgupta/F25-Deep-Learning-Project.

Course Project for CSE676 - Fall 2025 University at Buffalo

Benchmark and comparison dashboard for FLUX.1-dev spatial acceleration methods (Baseline vs. TaylorSeer vs. RALU) built with Streamlit.