PyTorch compiler that accelerates training and inference. Get built-in optimizations for performance, memory, parallelism, and easily write your own.

An Open Source Machine Learning Framework for Everyone

sparse block-attention optimization

AITemplate is a Python framework which renders neural network into high performance CUDA/HIP C++ code. Specialized for FP16 TensorCore (NVIDIA GPU) and MatrixCore (AMD GPU) inference.

K Means Clustering with Python

Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more

Multithreaded Python without the GIL

Slicing a PyTorch Tensor Into Parallel Shards

A memory balanced and communication efficient FullyConnected layer with CrossEntropyLoss model parallel implementation in PyTorch