GPU Optimizer
Mathews-Tom/armory
GPU optimization for consumer NVIDIA GPUs (8-24GB VRAM) covering mixed precision, gradient checkpointing, XGBoost GPU, CuPy/cuDF migration, and torch.compile.
Touch-lists for common OCANNL extension tasks: adding a primitive operation, adding or extending a backend, extending shape inference, and diagnosing output differences between backends.
Install Extending Ocannl
$ npx skills add ahrefs/ocannl --skill extending-ocannl -a claude-codeProject install by default; add -g for ~/.claude/skills/.
$ gh skill install ahrefs/ocannl extending-ocannl --agent claude-codeProject scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).
$ git clone --depth 1 https://github.com/ahrefs/ocannl.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/extending-ocannl .claude/skills/extending-ocannl && rm -rf skills-srcUse ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.
Claude Code skills documentation · loads skills from .claude/skills/
Install the "extending-ocannl" agent skill from https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannl into .claude/skills/extending-ocannl/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "extending-ocannl", then confirm the skill loads.Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$skill-installer install https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannlType this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.
$ npx skills add ahrefs/ocannl --skill extending-ocannl -a codexProject install goes to .agents/skills/; add -g for ~/.codex/skills/.
$ gh skill install ahrefs/ocannl extending-ocannl --agent codexProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/ahrefs/ocannl.git skills-src && mkdir -p .agents/skills && cp -r skills-src/.claude/skills/extending-ocannl .agents/skills/extending-ocannl && rm -rf skills-srcUse ~/.agents/skills/ instead of .agents/skills for a personal install.
Codex skills documentation · loads skills from .agents/skills/
Install the "extending-ocannl" agent skill from https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannl into .agents/skills/extending-ocannl/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "extending-ocannl", then confirm the skill loads.Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add ahrefs/ocannl --skill extending-ocannl -a cursorProject install goes to .agents/skills/; add -g for ~/.cursor/skills/.
$ gh skill install ahrefs/ocannl extending-ocannl --agent cursorProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/ahrefs/ocannl.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/.claude/skills/extending-ocannl .cursor/skills/extending-ocannl && rm -rf skills-srcUse ~/.cursor/skills/ instead of .cursor/skills for a personal install.
Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/
Install the "extending-ocannl" agent skill from https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannl into .cursor/skills/extending-ocannl/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "extending-ocannl", then confirm the skill loads.Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gemini skills install https://github.com/ahrefs/ocannl.git --path .claude/skills/extending-ocannl--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.
$ npx skills add ahrefs/ocannl --skill extending-ocannl -a gemini-cliProject install goes to .agents/skills/; add -g for ~/.gemini/skills/.
$ gh skill install ahrefs/ocannl extending-ocannl --agent gemini-cliProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/ahrefs/ocannl.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/.claude/skills/extending-ocannl .gemini/skills/extending-ocannl && rm -rf skills-srcUse ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.
Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/
Install the "extending-ocannl" agent skill from https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannl into .gemini/skills/extending-ocannl/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "extending-ocannl", then confirm the skill loads.Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ gh skill install ahrefs/ocannl extending-ocannlInstalls for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).
$ npx skills add ahrefs/ocannl --skill extending-ocannl -a github-copilotProject install goes to .agents/skills/; add -g for ~/.copilot/skills/.
$ git clone --depth 1 https://github.com/ahrefs/ocannl.git skills-src && mkdir -p .github/skills && cp -r skills-src/.claude/skills/extending-ocannl .github/skills/extending-ocannl && rm -rf skills-srcUse ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.
GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/
Install the "extending-ocannl" agent skill from https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannl into .github/skills/extending-ocannl/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "extending-ocannl", then confirm the skill loads.GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
$ npx skills add ahrefs/ocannl --skill extending-ocannl -a opencodeOpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.
$ gh skill install ahrefs/ocannl extending-ocannl --agent opencodeProject scope by default (.agents/skills/); add --scope user for a personal install.
$ git clone --depth 1 https://github.com/ahrefs/ocannl.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/.claude/skills/extending-ocannl .opencode/skills/extending-ocannl && rm -rf skills-srcUse ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.
OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/
Install the "extending-ocannl" agent skill from https://github.com/ahrefs/ocannl/tree/master/.claude/skills/extending-ocannl into .opencode/skills/extending-ocannl/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "extending-ocannl", then confirm the skill loads.OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.
Facts
- Skill name
extending-ocannl- GitHub stars
- 118
- Token cost
- ~728 tokens
- SKILL.md length
- 287 words
- Files
- 1
- Skills in repo
- 3
- Repo updated
- First seen
- Licence
- BSD-2-Clause
At a glance
Touch-lists for common OCANNL extension tasks: adding a primitive operation, adding or extending a backend, extending shape inference, and diagnosing output differences between backends.
- Works in 5 steps: Add primitive operation to… → Implement interpretation in the same file → Add syntax support in tensor/ppx_*.ml if… → …
- Adding an op to ops.ml
- SKILL.md covers Adding New Primitive Operations, Debugging Backend Discrepancies, Backend Extensions and Shape Inference Extensions
- Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md
What it does
Extending Ocannl is an agent skill from ahrefs/ocannl. Touch-lists for common OCANNL extension tasks: adding a primitive operation, adding or extending a backend, extending shape inference, and diagnosing output differences between backends. Use when adding an op to ops.ml, implementing a new backend, changing projection/constraint logic, or when the same computation gives different results on cc vs cuda vs metal.
Its SKILL.md is about 730 tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.
It sits in AI & LLM Engineering, covering Deep learning. It works with CUDA. The repository describes itself as: OCANNL: OCaml Compiles Algorithms for Neural Networks Learning. The licence is BSD-2-Clause.
When your agent uses it
- Adding an op to ops.ml
- Implementing a new backend
- Changing projection/constraint logic
- The same computation gives different results on cc vs cuda vs metal
Example prompts
- “/extending-ocannl”
Workflow steps
5 steps, taken from the first numbered list in SKILL.md.
- Add primitive operation to arrayjit/lib/ops.ml
- Implement interpretation in the same file
- Add syntax support in tensor/ppx_*.ml if needed
- Add high-level wrappers in tensor/operation.ml
- For neural network blocks, see lib/nn_blocks.ml for patterns
What it can do on your machine
Read from SKILL.md and the folder at commit 003e177. It shows what the files ask for, not the result of running them.
Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
No scripts in the folder and no shell commands in SKILL.md.
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.
Context cost
Extending Ocannl loads about 728 tokens when it runs. Until then it costs about 95 tokens; SKILL.md has 287 words of instructions outside code blocks.
Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.
Safety
Auto-check passed
The automated check found no risky patterns in SKILL.md.
Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.
SKILL.md
The full file from ahrefs/ocannl at commit 003e177, republished under its BSD-2-Clause licence (© ahrefs). 287 words, ~728 tokens.
Download SKILL.mdSave it as
.claude/skills/extending-ocannl/SKILL.md (or your agent's skills folder).- name
- extending-ocannl
- description
- Touch-lists for common OCANNL extension tasks: adding a primitive operation, adding or extending a backend, extending shape inference, and diagnosing output differences between backends. Use when adding an op to ops.ml, implementing a new backend, changing projection/constraint logic, or when the same computation gives different results on cc vs cuda vs metal.
Extending OCANNL
Adding New Primitive Operations
- Add primitive operation to
arrayjit/lib/ops.ml - Implement interpretation in the same file
- Add syntax support in
tensor/ppx_*.mlif needed - Add high-level wrappers in
tensor/operation.ml - For neural network blocks, see
lib/nn_blocks.mlfor patterns
Debugging Backend Discrepancies
When outputs differ between backends:
- Compare runtime logs in
<backend>-<device>-<stream>.logfiles (might require minimizing test tensors) - Check generated code in
build_files/<exe-name>/*.cvs*.cu/*.metalfor differences - Common issues:
- Incorrect type conversion in
convert_precisionoverrides - Different numerical precision between CPU and GPU operations
- Incorrect type conversion in
Backend Extensions
- Implement device-specific module following
Backend_implsignatures; backends implement stream-based execution with FIFO queuing, plus events and synchronization between streams/devices - Add compilation logic in
arrayjit/lib/backends.ml - Handle memory management and synchronization
- Add configuration options in
ocannl_config.reference
Code generation architecture
c_syntax.mlprovides a functor with default C code generation patterns fromLow_level.tcc_backend.mluses the defaults with minimal overrides;cuda_backend.mloverrides more functions for CUDA-specific syntax (e.g.__float2half);metal_backend.mloverrides using MSL-specific syntax;hip_backend.mlmirrors the CUDA backend- Backends must provide
convert_precisionfor type conversions - Builtin functions (e.g. type conversions) must be implemented in the per-backend builtin modules prepended to generated code:
builtins_cc.mlfor the C backends,builtins_cuda.ml(CUDA),builtins_hip.ml(HIP),builtins_metal.ml(Metal).builtins.cprovides the host-side FFI stubs compiled into the library - When adding new precision types, ensure conversion functions exist in all backend builtins
- A change to a codegen default moves emitted text outside
test/too:dune build @test/operations/runtest-codegen_text_inventorylists every pinning file. A change toRegister_tile.default, for one, re-promotesarrayjit/test/test_vectorized_codegen, which atest/operations-only batch never reaches
Shape Inference Extensions
- Modify projection logic in
arrayjit/lib/indexing.ml - Update shape constraint generation in
tensor/shape.ml - Test with various einsum patterns in e.g.
test/einsum_trivia.ml
© ahrefs, BSD-2-Clause. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file
Files
Just SKILL.md in .claude/skills/extending-ocannl of ahrefs/ocannl.
Open the folder on GitHubat commit 003e177
Compare with similar skills
Extending Ocannl next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.
| Skill | Stars | Used in | Tokens | Auto-check | Licence | Repo updated |
|---|---|---|---|---|---|---|
| Extending Ocannl this skillahrefs/ocannl | 118 | — | ~728 | Automated safety check: Pass | BSD-2-Clause | |
| GPU OptimizerMathews-Tom/armory | 328 | — | ~3.5k | Automated safety check: Notes | MIT | |
| The Art of Debuggingstas00/the-art-of-debugging | 1.7k | — | ~6.1k | Automated safety check: Notes | CC-BY-SA-4.0 | |
| Cuda Cpp Kernelvipshop/cache-dit | 1.3k | — | ~2.3k | Automated safety check: Pass | Apache-2.0 | |
| Debug Cuda Crashsgl-project/sglang | 37k | 2 repos | ~4.9k | Automated safety check: Pass | Apache-2.0 | |
| ONNX Runtime CUDA Attention Patternsmicrosoft/onnxruntime | 22k | — | ~6.5k | Automated safety check: Pass | MIT |
Similar skills
The Art of Debugging
stas00/the-art-of-debugging
Condensed debugging method and tool recipes for Unix, Python and PyTorch programs: crashes, hangs, segfaults, wrong output, CUDA OOM, NaN values and slowness.
Cuda Cpp Kernel
vipshop/cache-dit
A skill your agent uses when writing, debugging, porting, reviewing, or optimizing CUDA C++ or PTX kernels; investigating CUDA Runtime or Driver API behavior; profiling kernels with Nsight Systems…
Debug Cuda Crash
sgl-project/sglang
Call this skill when you need to debug CUDA crashes in SGLang using kernel API logging
- Official
ONNX Runtime CUDA Attention Patterns
microsoft/onnxruntime
Patterns and pitfalls for the ONNX-domain Attention operator's CUDA implementation in ONNX Runtime: dispatch cascade, eligibility limits, mask and bias kernels, and test routing.
Aoti Debug
pytorch/pytorch
Debug AOTInductor (AOTI) errors and crashes. An agent skill from pytorch/pytorch.
More from ahrefs/ocannl
Slipshow
ahrefs/ocannl
Writing Slipshow presentations, a text-based engine that compiles Markdown to interactive scrolling HTML presentations.
Ocannl Debug Tracing
ahrefs/ocannl
How to get debug output out of OCANNL: kernel/routine runtime logs, generated code artifacts, and ppxminidebug tracing of library internals such as the shape/row solver.
Related
Works with
Questions about Extending Ocannl
What does Extending Ocannl do?
Touch-lists for common OCANNL extension tasks: adding a primitive operation, adding or extending a backend, extending shape inference, and diagnosing output differences between backends. Extending Ocannl is an agent skill from ahrefs/ocannl. Touch-lists for common OCANNL extension tasks: adding a primitive operation, adding or extending a backend, extending shape inference, and diagnosing output differences between backends.
When should I use Extending Ocannl?
Extending Ocannl fits situations like: adding an op to ops.ml; implementing a new backend; changing projection/constraint logic; the same computation gives different results on cc vs cuda vs metal.
How do I install Extending Ocannl in Claude Code?
Run `npx skills add ahrefs/ocannl --skill extending-ocannl -a claude-code`. Or copy the skill folder (.claude/skills/extending-ocannl in ahrefs/ocannl) into .claude/skills/extending-ocannl in your project. Claude Code loads it when a task matches its description.
How do I install Extending Ocannl in Codex?
Run `npx skills add ahrefs/ocannl --skill extending-ocannl -a codex`. Or copy the skill folder (.claude/skills/extending-ocannl in ahrefs/ocannl) into .agents/skills/extending-ocannl in your project. Codex loads it when a task matches its description.
Can I use Extending Ocannl in Cursor, Gemini CLI or GitHub Copilot?
Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add ahrefs/ocannl --skill extending-ocannl -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/extending-ocannl, .gemini/skills/extending-ocannl, .github/skills/extending-ocannl and .opencode/skills/extending-ocannl in your project.
What does Extending Ocannl need to run?
SKILL.md names no scripts, command-line tools or credentials: Extending Ocannl is instructions for the agent only.
Does Extending Ocannl access the network?
SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.
Is Extending Ocannl safe to install?
Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.
What licence does Extending Ocannl use?
Extending Ocannl is published under the BSD-2-Clause licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.
How many tokens does Extending Ocannl use?
About 728 tokens (SKILL.md is roughly 2.9k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.
What are the alternatives to Extending Ocannl?
Skills that share tags, products or a category with Extending Ocannl: GPU Optimizer (Mathews-Tom/armory, 328 stars), The Art of Debugging (stas00/the-art-of-debugging, 1.7k stars), Cuda Cpp Kernel (vipshop/cache-dit, 1.3k stars) and Debug Cuda Crash (sgl-project/sglang, 37k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.
Who maintains Extending Ocannl?
ahrefs (a GitHub organization) maintains it in ahrefs/ocannl, which has 118 GitHub stars. The repository holds 3 skills in this directory. The repository was last updated on October 7, 2026.
Source: ahrefs/ocannl on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.