intel

Create a CUDA-to-XPU migration assessment for an existing AI repo. Identify CUDA-specific assumptions, route to the right XPU skills, produce a migration report. Use when the user has a CUDA repo, notebook, Dockerfile, launch script, HF / vLLM / SGLang…

Passing dpnp arrays to and from other Python libraries on Intel CPUs and GPUs. Use when dpnp numeric work has to feed pandas, scikit-learn, PyTorch, or TensorFlow, when one of those libraries raises a type error on a dpnp array, when a pipeline mixes device…

Reading and writing files from dpnp code on Intel CPUs and GPUs. Use when the user needs to load an array into dpnp or save a dpnp result — .npy, .npz, HDF5 via h5py, Zarr, CSV or plain text — when a file is larger than device memory and has to be read in…

Linear algebra and FFT with dpnp on Intel CPUs and GPUs, backed by oneMKL. Use when a matrix multiply, solve, decomposition, eigenvalue problem, or Fourier transform is the hot part of NumPy code, when the user asks whether dpnp covers a linalg or FFT call,…

Device memory management for dpnp arrays on Intel CPUs and GPUs. Use when a dpnp script grows in memory until it fails, when a dataset does not fit in device memory, when an array turns out to be on a different device than expected, or when a loop allocates a…

Porting an existing NumPy or CuPy program to dpnp on Intel CPUs and GPUs. Use when deciding whether a codebase can run on dpnp at all, when a call raises NotImplementedError or AttributeError after the import was swapped, when the user asks whether dpnp…

NumPy-compatible array operations optimized for Intel hardware. Use when the user wants to migrate or port NumPy code to dpnp, asks whether a NumPy hot path can run on an Intel CPU or GPU, needs to check dpnp installation or SYCL device selection with dpctl,…

Random number generation with dpnp on Intel CPUs and GPUs, backed by oneMKL. Use when NumPy random calls move to dpnp, when a seeded dpnp run does not reproduce a NumPy sequence, when a distribution turns out not to be implemented, when results have to be…

Use when asked to fix a GitHub issue end-to-end, run the agent pipeline on an issue, or process an `agent:active` / batch tracking issue. Orchestrates the full pipeline: `issue-triage` → `fix-reproduce` → `fix-root-cause` → `fix-implement` → `fix-verify` →…

Use when asked to implement a fix for a root-caused failure, apply a proposed patch, or produce a staged code change from triage output. Takes fix-root-cause's output and edits code; leaves the change staged (uncommitted) for fix-verify to check. Does NOT run…

Use when asked to verify a fix works, confirm a staged patch resolves a failure, or produce a before/after summary of a fix. Runs the test command against a source build with the fix applied and reports PASSED / FAILED / CANNOT_VERIFY. Called by both…

Use when asked to fix nightly CI failures, analyze a nightly failure report, debug XPU tests from a CI run, or process a batch of failing tests emailed from the nightly job. Runs the same leaf pipeline as `issue-handler` but batches multiple failures using a…

Shallow, text-only triage of a GitHub issue on pytorch or torch-xpu-ops. Classifies the issue and returns a report (markdown table + JSON) for the caller to act on. Read-only; the skill itself does not comment on or label the issue.

How to create a pull request for the intel/torch-xpu-ops repository. Use when an agent has finished implementing a fix or feature and needs to prepare a branch and PR description that satisfies CI and review requirements.

Find upstream PyTorch behavior or fixes that may require XPU parity work, validate them on XPU, and produce independently reviewed evidence. Use for time-window alignment scans or targeted upstream-to-XPU investigations; not for implementing the resulting…

Analyze a failure and determine root cause, fix strategy, target repo, domain, and verdict (IMPLEMENTING or NEEDS_HUMAN). Analysis-only — no code changes. Used by both issue-handler and xpu-nightly-ci-fix orchestrators.

Recommend how to configure vLLM-XPU for a Hugging Face decoder-only LLM on Intel Arc B-series GPUs: choose quantization, KV dtype, DP/TP layout, max concurrency, and max context using roofline math against published hardware specs. Use when the user asks "How…

Run an arbitrary Hugging Face safetensors model on an Intel GPU using **upstream PyTorch** (>= 2.8) with the built-in `torch.xpu` device. Covers loading from the Hub, picking the right dtype, autocast, multi-GPU with accelerate's `device_map`, and the CUDA ->…

Benchmark a **running vLLM-XPU OpenAI-compatible server** on an Intel GPU using `vllm bench`. Measures TTFT (time-to-first-token), TPOT (time-per-output-token), ITL (inter-token latency), end-to-end latency, and throughput under concurrency. Covers online…

Profile a running vLLM-XPU server with torch.profiler around a window of real requests, either via /start_profile and /stop_profile HTTP endpoints or via vllm bench --profile for offline runs. Use to find the dominant op under real concurrent traffic. Not for…

Serve a Hugging Face safetensors model on an Intel GPU with upstream vLLM-XPU's OpenAI-compatible API, or check whether a model or architecture is currently documented on XPU. Covers live support lookup, image choice, container launch, known serve-flag…

Launch a Docker container with Intel GPU access on Linux. Encodes the correct combination of `--device /dev/dri`, render-group access, `--ipc=host`, `ZE_AFFINITY_MASK` pinning, Hugging Face cache mount, and `--entrypoint /bin/bash` for interactive use. Use…

Plan an end-to-end Intel XPU model deployment by chaining existing skills. Calls xpu-runtime-preflight (readiness), model-can-it-fit (sizing), model-config-recommend (flags), and the selected runtime skill (vllm-xpu-run / sglang-xpu-run / torch-xpu-run), then…

Profile Intel-XPU workloads at the SYCL / Level Zero kernel level via Intel pti-gpu's unitrace. Captures per-API-call and per-kernel timing, memory transfers, oneCCL / MPI events, and hardware counters PyTorch-level profilers cannot see. Use when a hot op is…

Review or prepare a pull request for the AutoRound repository — checks registration points for new data types/backends/VLMs, validates Chinese translation parity for modified markdown files, verifies quantization numerical stability (scale overflow, STE…

Adapt AutoRound to support a new LLM architecture that doesn't work out-of-the-box. Use when quantization fails for a new model type, block detection doesn't find layers, MoE models need unfusing, custom forward passes are needed, or non-standard linear layer…

Adapt AutoRound to support a new diffusion model architecture (DiT, UNet, hybrid AR+DiT). Use when a new diffusion model fails quantization, needs custom output configs, requires a custom pipeline function, or is a hybrid architecture with both autoregressive…

Add support for a new Vision-Language Model (VLM) to AutoRound, including multimodal block handler, calibration dataset template, and special model handling. Use when integrating a new VLM like LLaVA, Qwen2-VL, GLM-Image, Phi-Vision, or similar multi-modal…

Add a new hardware inference backend to AutoRound for deploying quantized models (e.g., CUDA/Marlin, Triton, CPU, HPU, ARK). Use when implementing QuantLinear kernels, registering backend capabilities, or enabling quantized model inference on a new hardware…

Add a new model export format to AutoRound (e.g., auto_round, auto_gptq, auto_awq, gguf, llm_compressor). Use when implementing a new quantized model serialization format, adding a new packing method, or extending export compatibility for deployment…

Add a new quantization data type to AutoRound (e.g., INT, FP8, MXFP, NVFP, GGUF variants). Use when implementing a new weight/activation quantization scheme, registering a new quant function, or extending the data_type registry.

Check TD Runtime Environment

Get TDVM event log

Get TDVM Quote Information

Detect and fix x86/C/C++ performance patterns from source code or profiling output (perf, VTune, flamegraphs). Invoke when the user asks to optimize, review for performance, or write new SIMD/vectorized code — even without profiling data. Trigger on: serial…

Profile and fix Linux performance problems using `perf`. Workflows: (A) hardware counters -- IPC, cache-miss, branch mispredictions; (B) hotspot profiling -- which functions and source lines consume CPU, with SIMD and accumulator detection; (C) cache-line…

Install, run, parse, and optimize benchmarks from the Phoronix Test Suite (PTS). Use this skill whenever the user mentions "phoronix", "pts/", or "phoronix-test-suite", or asks to run, measure, improve, or optimize a PTS test — e.g., "run pts/mt-dgemm",…

Use this skill when running functional tests to validate PerfSpect code changes, when the user says "run functional tests", "test my changes", "check for regressions", or when verifying a code change did not break existing functionality.

Use when writing, editing, reviewing, or debugging synthesizable SystemC code for Intel SystemC Compiler (ICSC) and SingleSource library. Covers rules for module hierarchy, channels and ports, process declarations, reset behavior, sensitivity lists, SystemC…

Translate a Triton kernel from the deprecated block-pointer API (tl.make_block_ptr / tl.advance / tl.load(boundary_check=...)) into an equivalent kernel using the modern device-side tensor-descriptor API (tl.make_tensor_descriptor / desc.load / desc.store)…

Deploy, tune, validate, and benchmark vLLM on Intel Xeon CPUs (CPU-only inference, no GPU). USE FOR: serving and performance optimizing LLMs on Intel Xeon, vLLM CPU install, CPU inference tuning, AMX bfloat16 setup, NUMA pinning, VLLM_CPU_KVCACHE_SPACE,…

Expert x86 Linux kernel crash/oops analyser. Trigger whenever a kernel crash is involved in any form: (1) inline — keywords "BUG:", "Oops:", "WARNING:", "Kernel panic", "Call Trace", "paging request"; (2) in a file — "I saved the crash/oops/dmesg in <file>",…