pytorch

Triages GitHub issues by routing to oncall teams, applying labels, and closing questions. Use when processing new PyTorch issues or when asked to triage an issue.

Review PyTorch pull requests for code quality, security, and backward compatibility. Use when reviewing PRs, when asked to review code changes, or when the user mentions "review PR", "code review", or "check this PR".

Sub-triages issues in the oncall:distributed queue by assigning distributed module labels, routing to sub-oncalls, and marking triaged. Use when an issue has been routed to oncall:distributed and needs second-level triage.

Query PyTorch CI, GitHub Actions, HUD, Grafana, and infrastructure metrics. Use when users ask about CI duration, job failures, queue times, workflow trends, runner health, dashboard data, or PyTorch infrastructure metrics.

Deprecate, freeze, or remove PyTorch APIs; choose a removal version, handle won't-fix frozen components, or audit overdue removals.

Checklist for adding a new overload (e.g. `mm.dtype`, `addmm.compute_mode`) to an existing ATen op in native_functions.yaml. Use before the first rebuild when adding an overload or new kwarg variant of an existing operator, so every rebuild-triggering edit…

Manage tensor lifetimes across CUDA streams and decide whether record_stream is avoidable. Use when using a tensor on a stream other than the one it was allocated on, adding or reviewing Tensor.record_stream / CUDACachingAllocator::recordStream, side streams,…

Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to operators, kernels, or when user mentions enabling unsigned types, barebones unsigned types, or uint…

Review a pytorch/pytorch pull request's changes and decide whether they are safe to land. Emits a single machine-readable verdict (LAND or NO_LAND) for the greenlight auto-land gate.

Root-cause a vLLM torch-nightly CI regression report. Reads the pre-fetched, ANSI-stripped cluster logs from the triage artifact, identifies the real exception per regressed cluster, groups clusters by shared root cause, routes each cause to pytorch/pytorch,…

Review vLLM-routed torch-nightly root causes against existing upstream vLLM issues using read-only search results, and emit a validated upstream-checks artifact for the filer. Use in the vLLM torch-nightly triage workflow after root-cause analysis.

Load this FIRST whenever working with PyTorch CI data (any pytorch/* org repo), the torchci/HUD codebase, or the PyTorch HUD ClickHouse database. Covers the two MCP servers that reach this data — pytorch-hud (curated no-SQL triage plus ~145 pre-tuned named…

Fallback loader for the canonical PyTorch ci-metrics skill. Use for CI duration, failures, queue times, workflow trends, runner health, or dashboard data only when the canonical instructions are not already loaded and no accessible pytorch/pytorch checkout…

Find reverts that landed on pytorch/pytorch main but are missing from a release branch (release/X.Y) because the reverted commit was already shipped in a release candidate, and open one cherry-pick PR per missing revert against the release branch (on a fork…

Pin (or re-pin) the Linux manywheel builder docker images used by the nightly/release binary build workflows in pytorch/pytorch to a fixed .ci/docker build. Takes the release version (e.g. 2.13) as input. Updates DOCKER_IMAGE_PIN in…

Generate (and optionally open) a PyTorch release tracker / cherry-pick tracking issue from a release announcement, like https://github.com/pytorch/pytorch/issues/180506. Triggered by mentions of "release tracker", "release tracking issue", "cherry-pick…

Develop, debug, or explain the Arm backend in backends/arm/ and its community-facing examples in examples/arm/. Covers partitioning, TOSA lowering, VGF, Ethos-U, quantization, graph passes, memory layout, and node visitors. Not for development of other…

Review ExecuTorch pull requests for issues CI cannot check. Use when reviewing PRs, when asked to review code changes, or when the user mentions "review PR", "code review", or "check this PR".

Build ExecuTorch from source — Python package, C++ runtime, runners, cross-compilation, and backend-specific builds. Use when compiling anything in the ExecuTorch repo, diagnosing build failures, or setting up platform-specific builds.

Build and configure ExecuTorch as a Zephyr RTOS module for embedded boards. Use when setting up a Zephyr workspace with ET, adding board support (overlays, confs, memory layout), building with west, or debugging linker memory overflow.

Build, test, or develop the Cortex-M (CMSIS-NN) backend. Use when working on backends/cortex_m/, running Cortex-M tests, or exporting models for Cortex-M targets.

Build, test, or develop the QNN (Qualcomm AI Engine Direct) backend. Use when working on backends/qualcomm/, building QNN (use backends/qualcomm/scripts/build.sh), adding new ops or passes, running QNN delegate tests, or exporting models for Qualcomm HTP/GPU…

Search the ExecuTorch tribal knowledge base covering QNN, XNNPACK, Vulkan, CoreML, Arm, and Cadence backends, quantization recipes, export pitfalls, runtime errors, and SoC compatibility. Use when debugging ExecuTorch errors, choosing quantization configs,…

Analyze and reduce ExecuTorch binary size. Use when investigating binary size, running size tests, or optimizing the runtime for size-constrained deployments.

Write a torchtitan PR description that convinces a human reviewer -- concrete problem with a simple example, what changed, user-facing before/after, numerics, testing. Use when opening or updating a PR, or when asked to write or improve a PR summary.

Review torchtitan pull requests the way the maintainers do -- question the problem and design first, push for minimal code, explain the change with a concrete example, then check correctness. Use when reviewing a PR or branch, when asked to "review", "review…

Make a torchtitan model train faster with `local_compile` regions and compile-friendly rewrites: measure, find the slow glue, rewrite it, verify. Use when asked to speed up a model, layer or op with torch.compile, or to review such a change.

Add a new HF MoE model to the transformers_modeling_backend experiment. Use when the user provides a HuggingFace model ID and wants to integrate it with Titan's MoE, or invokes /add_moe_model.

Run TorchTitan with fully fake communication or real pipeline communication plus fake SPMD axes, especially for logical-rank memory debugging. Use for fake process groups, real-PP/fake-SPMD runs, large-model dry runs, or comparing rank-local CUDA memory…

Capture and compare per-op activations between two TorchTitan runs to spot numerics divergence (eager vs aot_fx_trace, FSDP vs no-FSDP, before vs after a refactor). Use when the user wants to debug bitwise / numeric drift in training, or invokes…

Profiler-driven hill-climbing to close the inference throughput gap between TorchTitan's unified model (running inside vLLM) and vLLM's native model. Benchmark with generate.py --benchmark, climb optimization rungs (compile / cudagraph / fused kernels),…

Make a CUDA-graph PyTorch profiler trace readable in Perfetto by packing replay streams into semantic lanes, and optionally merge per-rank pipeline-parallel traces into one aligned timeline with send/recv flow arrows. Use when a trace captured with CUDA…

OSDC deployment workflow, just recipes, base deploy order, module deploy order, clusters.yaml configuration, Terraform architecture, smoke tests, justfile conventions. Applies to ~/meta/ci-infra/osdc. Load when deploying, adding modules, or modifying deploy…

OSDC runners, NodePools, BuildKit, GitHub Actions constraints, EKS node taints, image mirroring, and the runner/nodepool change checklist. Applies to ~/meta/ci-infra/osdc. Load when modifying runners, nodepools, BuildKit, or node configurations.

Harbor pull-through cache configuration, Helm chart gotchas (inconsistent value paths, per-component image overrides, taint tolerations), and image mirroring strategy. Applies to ~/meta/ci-infra/osdc. Load when working on Harbor, image mirroring, or container…

How PyTorch CI jobs in .github/workflows/ resolve to a specific runner, AND which infrastructure that runner lives on: the two infras (legacy ALI Lambda-autoscaled EC2 vs current OSDC ARC-on-EKS) and the dot-vs-dash prefix split (linux./lf./c. = old EC2;…

OSDC observability: monitoring (metrics) and logging pipelines, three-Alloy architecture, Grafana Cloud Loki + Mimir queries, label strategy, credential setup, and troubleshooting. Applies to ~/meta/ci-infra/osdc. Load when working on monitoring, logging,…

Read-only CLI commands for debugging OSDC clusters: kubectl, aws, helm, tofu. Includes command references, safety boundaries, and the deploy-audit ConfigMaps in osdc-system (what version is running where, who deployed it, when, did it succeed, is a deploy…

OSDC platform architecture, project organization, directory tree, key design decisions, external knowledge base, key files, and docs index. Applies to ~/meta/ci-infra/osdc. Load when working on OSDC to understand project layout and architecture.

OSDC NodeLocal DNSCache (NLD) base component — per-node CoreDNS DaemonSet under base/kubernetes/nodelocaldns/. Covers iptables-mode rationale (vs IPVS-mode), dynamic kube-dns ClusterIP substitution via deploy.sh (validates the kube-dns ClusterIP is IPv6 and…

Verify changes to the Helion CuTe backend by running both test suites, lint, and a code review subagent — then fix any issues found.

Iteratively optimize a Helion kernel under HELION_BACKEND=cute until it matches or beats the best available baseline — via general backend/autotuner changes (knobs and seeds, not kernel-specific hacks) — then clean up history and report before/after numbers.

Address CI failures and unresolved review comments on a Helion pull request. Auto-activate when the user mentions a URL like https://github.com/pytorch/helion/pull/<number>.

Address CI failures and unresolved review comments on every PR in a stack-pr stack, amending each fix into its PR's commit, then rebase the stack onto origin/main and resolve conflicts. Use when the user asks to fix a stack of Helion PRs.

Context for working on Helion's TPU/Pallas backend. Auto-activate when the user mentions TPU or Pallas in the context of Helion development.

Investigate a Helion benchmark dashboard regression (helionlang.com/dashboard) — find the cause and classify it. Auto-activate when the user reports a perf drop/spike on a dashboard platform (e.g. b200 cute) around a given date.

Auto-activate when the user mentions runpod or scripts/runpod.py

Analyze torch-tensorrt local test results and drive failures to a fix. Use when the user pastes a test report / summary, asks why tests failed, asks to triage or fix failing tests, or mentions the JUnit/test-summary output from `just tests-report` / `just…

Build torch-tensorrt locally — install/pin the matching PyTorch nightly, drive Bazel through setup.py, do a clean rebuild after libtorch ABI changes, and recover from common build failures (undefined symbol, stale _C.so, libtorchtrt.so missing). Invoke…

Review PyTorch tutorials pull requests for content quality, code correctness, build compatibility, and style. Use when reviewing PRs, when asked to review code changes, or when the user mentions "review PR", "code review", or "check this PR".

Update the third_party/kineto submodule in PyTorch to the latest commit from this kineto repo and commit the change. Use when updating the kineto submodule hash for a release.

Replace duplicate documentation directories with tiny HTML redirect stubs to shrink GitHub Pages artifacts. Use when a versioned-docs site (Sphinx/MkDocs/Jekyll) ships byte-identical copies of the same content under multiple URL aliases (e.g., stable/…