Skills for your AI assistant
Add a skill and Claude Code, Codex, Cursor or a similar assistant learns one more job.
GPU acceleration
17 skillsCUDA kernels, compilers and communication libraries.
- cudaq-guide2.5k installsGuides CUDA-Q setup, simulation targets, QPU access, and writing quantum kernels.
- compute-sanitizer-diagnose-cuda-bugs105 installsFinds CUDA memory errors, data races, uninitialized reads, and sync bugs and explains how to read the report.
- cluster-network-analysisEstimates cross-server traffic, network bottlenecks and sync waits for multi-node training or inference.
- kernel-runtime-analysisDiagnoses GPU kernel, fusion, compilation and runtime submission issues to find what sits on the request critical path.
- cudaq-importing47 installsPorts quantum circuits from frameworks like Qiskit into CUDA-Q while keeping the algorithm and validation intact.
- genomics-workflow-acceleration10 installsSpeeds up existing genomics pipelines by adding optional GPU steps with NVIDIA Parabricks, keeping CPU fallbacks.
- cuequivariance4 installsDefines cuEquivariance groups and irreps, builds equivariant tensor products, and validates polynomials with NumPy.
- cuequivariance-torch4 installsBuilds and debugs cuEquivariance PyTorch tensor products and equivariant graph neural networks.
- cuequivariance-jax3 installsWrites and debugs cuEquivariance JAX polynomials, irrep dictionaries, and Flax NNX layers.
- compileiq-bootstrap2 installsChecks that CUDA, ptxas, GPU access, and the CompileIQ Python package are set up before any search runs.
- compileiq-author-objective1 installsWrites the objective function that scores compiler configurations during a CompileIQ search.
- compileiq-booster-pack1 installsApplies ready-made compiler optimization candidates one by one and keeps only those that compile, stay correct, and run faster.
- compileiq-debug1 installsDiagnoses CompileIQ problems like hangs, invalid scores, flat results, and ptxas errors from symptoms.
- compileiq-run-search1 installsSets up and launches a CompileIQ search, choosing worker type, sizing, checkpointing, and tracking.
- compileiq-validate-result1 installsRe-measures top CompileIQ candidates against the baseline and filters out false-positive speedups before shipping.
- parallelism-planningCompares data, tensor, sequence, context, pipeline and expert parallelism by capacity and communication cost.
- compileiq-search-spacePicks the right search space for a CompileIQ run, including version pinning and attention-kernel variants.