Execute architecture-tuned GPU matrix multiplication kernels via NVIDIA CUTLASS.
npx skills add https://github.com/jstzwj/ai-infra-plugins --skill cutlass
Please help me install this Agent Skill. Skill: cutlass Source: https://github.com/jstzwj/ai-infra-plugins/tree/main/plugins/cutlass/skills/cutlass Command: npx skills add https://github.com/jstzwj/ai-infra-plugins --skill cutlass
This Skill provides access to NVIDIA's CUTLASS library, enabling the deployment of optimized GPU kernels for matrix multiplication and tensor operations critical in AI and scientific computing.
Provide the target matrices' pointers and dimensions to run a GEMM operation with the CUTLASS library's Python API or C++ interface, specifying data types and architecture identifiers as needed.
High-intent search queries and answers about installing and using this skill.
To optimize matrix multiplication across GPU architectures like Volta, Turing, Ampere, and Hopper, this Skill deploys architecture-tuned GPU kernels that maximize Tensor Core utilization and resource efficiency for linear algebra workloads.
Yes, you can run mixed-precision GEMM operations. It supports a versatile range of data precisions including FP64, FP32, FP16, BF16, TF32, FP8, INT8, INT4, and complex types to cater to diverse workload requirements.
The best way to accelerate large language model training with CUDA is by leveraging architecture-aware CUTLASS kernels for high-throughput, mixed-precision matrix multiplications. You provide target matrix pointers and dimensions to execute the operation.
Yes, you need to install the CUDA toolkit and the CUTLASS library. These dependencies are required to facilitate architecture-tuned GPU kernel execution for complex linear algebra tasks.
High-performance GPU computing works with both AI and scientific workloads. It enables the deployment of optimized GPU kernels for tensor operations critical in scientific computing and HPC, supporting diverse data precisions.
dotruruConfigure Zig builds for SIMD, cross-compilation, and CUDA interop.
xchang1121Optimize grid and block configurations for Triton-Ascend kernels.
mindspore-aiRun TileLang CUDA Torch examples for GEMM, GEMV, and LayerNorm kernels.
NeverSightDevelop LLVM-based compiler pipelines from frontends to optimized code generation.
