Open Fabric Interfaces

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.

TransferBench is a utility capable of benchmarking simultaneous copies between user-specified devices (CPUs/GPUs)

PyTorch Extension Library of Optimized Scatter Operations

The lightweight PyTorch wrapper for high-performance AI research. Scale your models, not the boilerplate.