Bert Maher

simplegemm. Cuda

llama2.so. Inference Llama 2 with a model compiled to native code by TorchInductor

tf32_gemm. Example of binding a TF32 CUTLASS GEMM kernel to PyTorch

bitserial. Hacking around with ultra-low precision GEMM using TVM

pytorch. Tensors and Dynamic neural networks in Python with strong GPU acceleration

membench. C++

midwit-matmul. A simplistic approach to high-performance GPU matmul

learn_cuda. Simple programs for learning CUDA

pyhpc-benchmarks. A suite of benchmarks for CPU and GPU performance of the most popular high-performance libraries for Python :rocket:

deepstream-video-pipeline. Python