RubiaCx/Megatron-LM
Ongoing research training transformer language models at scale, including: BERT & GPT-2
Hi~ o(*￣▽￣*)ブ
Repositories
RubiaCx/MagiAttention
A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training
RubiaCx/NotionNext
一个使用 NextJS + Notion API 实现的，部署在 Vercel 上的静态博客系统。
RubiaCx/sglang
SGLang is a fast serving framework for large language models and vision language models.
RubiaCx/claude-code
An independent Python feature port of Claude Code, entirely rewritten from scratch. Educational Purpose only.
RubiaCx/TurboDiffusion
TurboDiffusion: 100–200× Acceleration for Video Diffusion Models
RubiaCx/FastVideo
A unified inference and post-training framework for accelerated video generation.
RubiaCx/LightX2V
Light Image Video Generation Inference Framework
RubiaCx/TransformerEngine
A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.
RubiaCx/CUTE_NOTE
RubiaCx/Awesome-ML-SYS-Tutorial
My learning notes/codes for ML SYS.
RubiaCx/CNTaxCalculator
2025 个税计算器
RubiaCx/infra-skills
A collection of specialized agent skills for AI infrastructure development, enabling Claude Code to write, optimize, and debug high-performance systems.
RubiaCx/Sparse-Attention-Video-Bench
RubiaCx/RubiaCx
RubiaCx/x-attention
[ICML 2025] XAttention: Block Sparse Attention with Antidiagonal Scoring
RubiaCx/tilelang
Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels
RubiaCx/DiffSynthStudio
Enjoy the magic of Diffusion models!
RubiaCx/slime
slime is an LLM post-training framework for RL Scaling.
RubiaCx/quack
A Quirky Assortment of CuTe Kernels
RubiaCx/finetrainers
Scalable and memory-optimized training of diffusion models
RubiaCx/CogVideo
text and image to video generation: CogVideoX (2024) and CogVideo (ICLR 2023)
RubiaCx/FlagGems
FlagGems is an operator library for large language models implemented in Triton Language.
RubiaCx/tritonbench
Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.
RubiaCx/Benchmark
RubiaCx/triton_meta
trition-lang/triton -> facebookexperimental/triton
RubiaCx/SageAttention
Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.
RubiaCx/CUDA-Learn-Notes
📚Tensor/CUDA Cores, 📖150+ CUDA Kernels, ⚡️⚡️toy-hgemm library with WMMA, MMA and CuTe (98%~100% TFLOPS of cuBLAS 🎉🎉).