Index of every page on this site: llms.txt
All skills
chelis-lang avatar

/backend-numerics

@f7af876
Run it yourself

Use for Chelis IR, backend, and autodiff work where numerical correctness matters. Focuses on evaluator-vs-backend agreement, compile-run harnesses, tolerances, and finite-difference checks.

  • 1 file
  • 1.3 KB
  • Updated 8 months ago
  • GitHub
Use this Skill: https://skilld.dev/gh/chelis-lang/chelis/backend-numerics

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈52 tokens for metadata: the name and description. ≈262 when used: this file.

Description uses 2.6% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈209 characters, or 2.6%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Backend Numerics

Use this skill for IR lowering, backend codegen, numerical kernels, and autodiff work.

Core Principle

A green string-emission test is not enough. Numerical paths must be validated with execution.

Validation Surfaces

  • evaluator vs generated C agreement
  • compile-and-run helpers
  • finite-difference comparisons for gradients
  • shape, stride, and reduction edge cases
  • BLAS and non-BLAS agreement where both paths exist

Chelis Expectations

  • prefer real numeric assertions over structural-only checks
  • include multi-op pipelines, not just single primitives
  • exercise scalar, small tensor, and larger tensor cases where threading matters
  • document any manual long-running numerical gate separately from default CI
  • On the current local machine, HIP numerical/manual checks are available: rocminfo reports Radeon 8060S Graphics with ISA gfx1100, and hipcc is installed. Use the real GPU path rather than downgrading to evaluator-only evidence when the gate specifically concerns HIP correctness.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at f7af876. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub last week.

Activeupdated 8 months ago

README badge

README badge for chelis-lang/chelis/backend-numerics