Explore projects

Development repository for the Triton language and compiler

《神经网络与深度学习》 Neural Network and Deep Learning

KernelBench: Can LLMs Write GPU Kernels? - Benchmark with Torch -> CUDA problems

PyTorch extensions for high performance and large scale training.

llmc is an efficient LLM compression tool with various advanced compression methods, supporting multiple inference backends.

[MLSys'25] QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving; [MLSys'25] LServe: Efficient Long-sequence LLM Serving with Unified Sparse Attention

A high-throughput and memory-efficient inference and serving engine for LLMs