Skip to main content
affpapa.org
RU EN
skill CX Codex

tilegym-converting-cutile-to-julia — Claude Code skill

Translates cuTile Python GPU kernels into cuTile.jl for Julia with 1-based indexing and column-major memory layout. Useful when porting workloads to Blackwell GPUs.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add NVIDIA/skills --skill tilegym-converting-cutile-to-julia -a claude-code
Open on GitHub

What tilegym-converting-cutile-to-julia does

The skill converts @ct.kernel Python routines into standalone Julia functions using cuTile.jl. It changes zero-based indices to one-based offsets, handles row-major to column-major array conversions, and resolves type differences. In Julia, functions like ct.full() do not exist, so the skill inserts fill, zeros, or ones.

Kernels live in a standalone julia/ directory with an isolated Project.toml, avoiding any Python runtime bridge. Testing relies on standard Test and NNlib.jl. The workflow also runs validate_cutile_jl.py to spot common bugs like unbroadcasted max.(a, b) calls before compiling.

Hardware limitations are strict. You need a Blackwell GPU with compute capability 10+ and a CUDA 13.1+ driver. Without this exact setup, compiling or executing the generated Julia kernels will fail.

What it helps with

  • translate Python @ct.kernel code to cuTile.jl functions
  • convert tensor indices from 0-based to 1-based
  • remap row-major memory layouts to Julia column-major arrays
  • validate cuTile.jl syntax with validate_cutile_jl.py before running tests

Example prompts for the agent

  • Port this cuTile Python kernel into julia/kernels/matmul.jl
  • Run validate_cutile_jl.py on julia/kernels/softmax.jl and fix any IRError risks
  • Convert this @ct.kernel function to cuTile.jl using standard Julia layout

How to install tilegym-converting-cutile-to-julia

1Skills CLI (Vercel)

npx skills add NVIDIA/skills --skill tilegym-converting-cutile-to-julia -a claude-code

2Manually: put the files in place

~/.claude/skills/tilegym-converting-cutile-to-julia/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/tilegym-converting-cutile-to-julia/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about tilegym-converting-cutile-to-julia

How do I install tilegym-converting-cutile-to-julia in Claude Code?

Run npx skills add NVIDIA/skills --skill tilegym-converting-cutile-to-julia -a claude-code.

What does tilegym-converting-cutile-to-julia do?

It automates translating cuTile Python kernels to Julia, resolving memory layouts, broadcast semantics, and positional argument order in ct.launch.

What hardware is required to run cuTile.jl kernels?

You need a Blackwell architecture GPU with compute capability 10+ and a CUDA 13.1+ driver. A compatible Julia version specified in julia/Project.toml is also mandatory.

Similar skills

Everything for AI and ML →
julia Mindrally Turns the agent into a Julia-aware pair programmer: multiple dispatch, type stability, docstrings and tests. Useful for numerical and scientific code. CX Codex +1 687installs julia-development Community Contribution A plugin for modern Julia work: package management, scientific computing and numerical code that has to run fast. Useful when a Julia project needs production habits. CX Codex ★ 41,729repo stars optimize-for-gpu K-Dense-AI Accelerates data-heavy Python workloads on NVIDIA GPUs and benchmarks the end-to-end speedup against the CPU baseline. CX Codex +1 2,000installs serverless-modal wanshuiyin Runs training, fine-tuning, and inference on serverless Modal GPUs. Helps pick the right GPU tier, estimates run costs upfront, and stops billing when work ends. CX Codex +1 463installs mojo-gpu-fundamentals modular Guidance for coding agents writing Mojo code for NVIDIA, AMD, and Apple silicon accelerators. It prevents models from injecting CUDA syntax into Mojo kernels. CX Codex +1 2,600installs cuda mohitmishra786 Guides the agent through CUDA C/C++ work: thread hierarchy, shared memory, streams, nvcc flags and Thrust. CX Codex +1 320installs

From the same set178

tao-generate-image-grounding A two-pass pipeline where a vision model pulls referring phrases out of each caption and returns pixel bounding boxes for them. Useful when you need phrase-grounding training data. CX Codex +1 1,600installs tao-train-image-classification Configures and executes PyTorch image classification tasks in NVIDIA TAO. Useful for training ResNet or EfficientNet backbones, distilling models, and exporting to TensorRT. CX Codex +1 1,600installs tao-finetune-clip A skill that walks an agent through CLIP training, zero-shot classification, embedding extraction and ONNX or TensorRT export in NVIDIA TAO. CX Codex +1 1,600installs tao-finetune-cosmos-embed Configures Docker runs for Cosmos-Embed1 inside NVIDIA TAO, enabling text-to-video retrieval, model fine-tuning, evaluation, and artifact exports. CX Codex +1 1,600installs tilegym-converting-cutile-to-triton Turns cuTile GPU kernels into Triton ones through gated phases: analyze, convert, validate, test, benchmark. Debugging of failed ports included. CX Codex +1 1,500installs cuopt-multi-objective-exploration Traces Pareto frontiers for multi-objective optimization using NVIDIA cuOpt. It maps tradeoffs between competing goals when no single weighting exists. CX Codex +1 1,300installs jetson-diagnostic Collects a read-only Jetson snapshot covering device identity, memory, GPU, thermal state, power, storage, services and processes. Use it to inspect a live host. CX Codex +1 1,300installs jetson-headless-mode Plans a headless-mode switch on Jetson and applies only safe, reversible changes: drops the graphical target and non-essential daemons. CX Codex +1 1,200installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor