Hey there! ๐ Welcome to #TrendingTuesday
This week we’ll look into the fastest growing repos written in CUDA, a programming model for parallel computing with NVIDIA GPUs.
CUDA programmers form a niche but very active community. They’re experiencing a lot of growth thanks to the recent boom in artificial intelligence applications. We’re excited to show you the coolest and fastest-growing repos in this category ๐
6๏ธโฃ In place number 6 we have a project form the company that brought you TikTok ๐ฑ.
bytedance/ByteTransformer is a high-performance inference library for BERT-like transformers that claims to deliver superior performance over other transformer implementations. This code has been actually deployed by ByteDance’s inference pipelines! You can read the technical details here in their ArXiv paper.
5๏ธโฃ The next two repos are brought to you by @nvidia.
NVlabs/long-video-gan is a video generation model that accurately reproduces object motion and changes in camera viewpoint over time. It improves upon other video generation models in that it is able to produce consistency in the dynamics and object persistence of the content being generated. ๐ฅ
4๏ธโฃ NVIDIA/FasterTransformer implements a highly optimised transformer layer for the encoder and decoder components in NLP models. This transformer layer is built on top of CUDA, cuBLAS, cuBLASLt and C++.
The project has ~93 issues and ~30 contributors.
3๏ธโฃ ALIEN (Artificial LIfe ENvironment) is an artificial life simulation tool that aims to better understand the conditions for (pre-)biotic evolution and the growing complexity of biological systems. The authors ambition is to make the simulator user-friendly through a visually appealing and modern UI. They’re certainly getting there ๐ฝ
Kudos to @chrxh for this wonderful project!
The project has ~5 Issues and ~4 contributors.
2๏ธโฃ In second place we have bitsandbytes, a lightweight wrapper for 8-bit optimisers and other CUDA custom functions ๐พ. Check it out if you’re looking to optimise your transformers or PyTorch models.
The project has ~95 issues and ~14 contributors.
Brought to you by @tim_dettmers.
1๏ธโฃ The fastest-growing project of the week is ashawkey/stable-dreamfusion ๐.
This is a PyTorch implementation of the text-to-3D model Dreamfusion, powered by the Stable Diffusion text-to-2D model. See here for the original paper.
The repo has ~90 issues and ~8 contributors.
Weโll write about trending repos on different topics and languages EVERY WEEK. Follow us here and on Twitter @quine_sh to stay in the loop! ๐ฆ
Do you want to build an impressive open source portfolio, but don’t know where to start? We’re here to help! Check out Quine and fast-forward your journey as an open source creator ๐ค
