Making large AI models cheaper, faster and more accessible

A PyTorch implementation of EfficientNet

Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.

Flickr-Faces-HQ Dataset (FFHQ)

This repo is implemented based on detectron2 and centernet

(ECCVW 2025)GPT4RoI: Instruction Tuning Large Language Model on Region-of-Interest

DVIS: Decoupled Video Instance Segmentation Framework

Differentiable Multi-Granularity Human Representation Learning for Instance-Aware Human Semantic Parsing