TransferBench is a utility capable of benchmarking simultaneous copies between user-specified devices (CPUs/GPUs)
PyTorch Extension Library of Optimized Scatter Operations
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.
The lightweight PyTorch wrapper for high-performance AI research. Scale your models, not the boilerplate.
Examples demonstrating available options to program multiple GPUs in a single node or a cluster
最近更新: 6天前