Skip to content

AGENTS.md

Latest commit

 

History

History
96 lines (82 loc) · 4.95 KB

AGENTS.md

File metadata and controls

96 lines (82 loc) · 4.95 KB
Raw

Agent Instructions

Repository Target

  • Treat GitHub as the active repository for this project: https://github.com/NVIDIA/TensorRT-Model-Connect.git.
  • Use the local github remote for fetch, push, PR, and CI operations.
  • Do not push project changes to any non-GitHub legacy remote unless the user explicitly asks for that repository.

Branch And PR Flow

  • The GitHub default branch is main.
  • Do not push directly to GitHub main.
  • Start new work from github/main on a short-lived branch.
  • Push the branch to the GitHub remote and open a pull request targeting main.
  • Wait for GitHub CI before merging.
  • Merge with squash or rebase, matching the repository ruleset.
  • Sign off every commit introduced by a pull request with DCO using git commit --signoff; ensure the Signed-off-by email matches the commit author.
  • Preserve or re-add valid sign-offs when amending, rebasing, or cherry-picking commits, and never sign off on another author's behalf.
  • Avoid commit messages containing Claude; the GitHub ruleset rejects them.

GitHub Pages

  • Keep GitHub Pages dedicated to the documentation website.
  • Do not publish CI reports to GitHub Pages unless the user explicitly changes that decision.

Model-Family Ownership And Execution Philosophy

  • The model family is the unit of ownership, fault isolation, and horizontal scale. One team or agent must be able to implement, validate, change, and revert one family without changing or coordinating with another.
  • Each model family owns all model-specific code: configuration and weights, mathematical topology, engine composition, runtime orchestration, and model validation.
  • Duplicate model-specific code by design. Duplication keeps model-specific defects, merge conflicts, and rollbacks family-local; code similarity never justifies a cross-family abstraction.
  • Shared code is limited to model-agnostic contracts and mechanics. It contains no model topology, model semantics, model orchestration, model-specific validation logic or evidence, or family-specific behavior.
  • Generic family code defines only the mathematical computation graph and runtime orchestration. TensorRT owns all lowering to GPU execution, including fusion, tactic and kernel selection, scheduling, code generation, and hardware or version adaptation.
  • Generic family semantics do not depend on GPU, SM, CUDA, driver, or TensorRT version. Platform-specific failures are project topology or orchestration defects, or upstream TensorRT issues.
  • Platform specialization is limited to complete network offload, such as TensorRT Edge-LLM, and TVM-FFI BYOK kernel bindings. Target-specific plans, timing caches, and compiled kernels do not make the model family specialized.
  • Existing family-owned GPU helper kernels outside TVM-FFI BYOK violate this architecture and are temporary migration debt, not a supported specialization.

Dos And Don'ts

  • Do keep validation criteria meaningful and aligned with the behavior under test.
  • Write repository documentation, code comments, user-facing messages, and PR text in English. Preserve non-English model inputs, tokenizer data, and test fixtures when they are required to validate multilingual functionality.
  • Never change the test passing criteria for the purpose of passing CI. If you believe the test is faulty, escalate to a human

Repo Skills

  • Codex skills packaged for this repo are registered through .agents/plugins/marketplace.json.
  • The trtmc-agent-skills plugin is marked INSTALLED_BY_DEFAULT and exposes repo-local skills from plugins/trtmc-agent-skills/skills/.
  • If a skill below is not listed in the active runtime skills, load plugins/trtmc-agent-skills/skills/<name>/SKILL.md directly and follow it. Registration is through the Codex plugin path, so other agent runtimes do not list these; an empty runtime skill list is not evidence that no skill covers the task.
  • Consult the skill that matches the task before starting it:
Task Skill
Onboard a Hugging Face model, or extend a family, to produce a bundle transform-model
TensorRT output disagrees with the reference, or validation fails numerically debug-trt-mismatch
FP16/BF16 dtype threading and explicit FP32 boundaries fp16-trt-network
Evaluate a precision or quantization format for a model optimize-model-precision
Diagnose runtime cost, or produce performance evidence profile-model
Prepare a build or validation environment on an unfamiliar host setup-trtmc-environment
Publish a change as a pull request submit-github-pr
Monitor PR CI, diagnose failed checks, rebase onto github/main pr-babysitter
Turn a finding into a GitHub issue submit-github-bug-issue
Draft or review commit messages, PR titles, PR descriptions, merge messages write-git-messages
Keep the website, skills, commands, and architecture docs aligned doc-sync