Tune and validate Python JIT GPU kernels using CuTe DSL.
npx skills add https://github.com/LancerLab/croqtile-tuner --skill croq-dsl-cute-dsl
Please help me install this Agent Skill. Skill: croq-dsl-cute-dsl Source: https://github.com/LancerLab/croqtile-tuner/tree/main/.claude/skills/croq-dsl-cute-dsl Command: npx skills add https://github.com/LancerLab/croqtile-tuner --skill croq-dsl-cute-dsl
This Skill provides a structured interface and guidance for tuning CuTe DSL (Python JIT) kernels, optimizing GPU performance efficiently.
Load the CuTe DSL tuning skill, validate environment, compile kernels, profile with NCU, run benchmarks, and verify correctness.
High-intent search queries and answers about installing and using this skill.
Tuning CuTe DSL kernels involves a structured workflow that validates the environment, compiles, profiles with NCU, and benchmarks Python JIT GPU kernels. This process integrates build, run, and verification to optimize kernel performance efficiently.
Profiling CuTe DSL kernels with NCU requires compiling the Python JIT kernel, running automated profiling templates, and verifying correctness. The workflow automates these steps to systematically validate and optimize GPU performance.
Yes, compiling and tuning CuTe DSL kernels requires the cutlass Python package. Additionally, Nvidia profiling tools are necessary to execute the integrated profiling and benchmarking steps within the tuning workflow.
Yes, you can automate environment validation, build, profiling, and benchmarking for Python JIT GPU kernels. The workflow establishes tuning contracts to ensure every step from compilation to verification is structured and automated.
The best way to verify CuTe DSL kernel correctness is using integrated verification templates within the tuning workflow. This ensures every profiling and benchmarking run is validated against expected outputs automatically.
CuTe DSL kernel tuning requires Nvidia profiling tools to execute the profiling phase of the workflow. NCU integration allows developers to capture performance metrics and systematically optimize GPU kernel execution.
ShekswessGuide writing, debugging, and optimizing GPU kernels with NVIDIA CuTe DSL.
vipshopImplement and optimize CuTe DSL GPU kernels across NVIDIA architectures.
m0atDevelop and benchmark optimized CUDA kernels for HuggingFace diffusers and transformers.
spacecat2002Add and validate lightweight CUDA JIT kernels in SGLang's jit_kernel module.
