GPU-Accelerated ML & CUDA Kernels
37 repos
Libraries and frameworks for machine learning acceleration on NVIDIA GPUs, centered on CUDA kernel optimization, collective communication primitives, and low-level GPU compute abstractions. Includes foundational tools like NCCL (distributed GPU communication), CUTLASS (GPU tensor operations), and cuML (GPU-accelerated ML algorithms), alongside operator implementations and GPU infrastructure components. Most repos are production Python and C++ code targeting high-performance ML workloads and system-level GPU programming.
Python · 12
C++ · 12
Rust · 4
Go · 2
C · 1
Haskell · 1
gpu ·98,029
cuda ·97,411
nvidia ·44,431
python ·38,299
machine-learning ·23,301
deep-learning ·19,954
cpp ·19,564
docker ·17,595
nvidia-docker ·17,594
pytorch ·13,378
