Loading...
Loading...
Found 424 Skills
Read and write large cuPyNumeric arrays to HDF5 with Legate's parallel, distributed HDF5 I/O (legate.io.hdf5: to_file, from_file, from_file_batched). Use when a developer needs to save a cuPyNumeric array to an .h5/.hdf5 file, load an HDF5 dataset into a distributed cuPyNumeric array, read a large HDF5 dataset in chunks, hand arrays to an HPC pipeline as a single file, or accelerate HDF5 disk I/O with GPUDirect Storage (GDS). Do not use it for Parquet/cuDF/raw-binary or other sharded/custom layouts (see the cupynumeric-parallel-data-load skill), Zarr or object-store/S3 output, .npz or pickled archives, plain h5py without cuPyNumeric, or pure array compute such as FFT, matmul, or reductions.
Read-only Jetson health snapshot for identity, memory, GPU, thermal, power, storage, services, and top processes.
Activation-aware weight quantization for 4-bit LLM compression with 3x speedup and minimal accuracy loss. Use when deploying large models (7B-70B) on limited GPU memory, when you need faster inference than GPTQ with better accuracy preservation, or for instruction-tuned and multimodal models. MLSys 2024 Best Paper Award winner.
Layout and styling in GPUI. Use when styling components, layout systems, or CSS-like properties.
Global state management in GPUI. Use when implementing global state, app-wide configuration, or shared resources.
Composable transformations of Python+NumPy programs. Differentiate, vectorize, JIT-compile to GPU/TPU. Built for high-performance machine learning research and complex scientific simulations. Use for automatic differentiation, GPU/TPU acceleration, higher-order derivatives, physics-informed machine learning, differentiable simulations, and automatic vectorization.
Write, explain, and debug Metal GPU code including shaders (vertex, fragment, compute), render pipelines, compute pipelines, buffer/texture management, and Metal 4 APIs. Use this skill whenever the user mentions Metal, GPU programming, shaders, MSL (Metal Shading Language), render passes, compute kernels, MTLDevice, MTLCommandBuffer, MTLRenderPipelineState, or any Apple GPU/graphics programming topic. Also trigger when the user wants to do parallel computation on Apple devices, write GPU-accelerated code, or work with Metal Performance Shaders, MetalFX, MetalKit, or Compositor Services. Covers iOS, macOS, iPadOS, tvOS, and visionOS.
Iteratively optimize cuTile kernel performance through systematic profiling, bottleneck analysis, IR comparison, and targeted tuning. Covers tile sizes, occupancy, autotune configs, TMA, latency hints, persistent scheduling, num_ctas, flush_to_zero, and IR-level debugging. Use when asked to "optimize cutile kernel", "improve kernel perf", "tune cutile performance", "make kernel faster", or iteratively benchmark and refine a cuTile GPU kernel in the TileGym project.
DGX Cloud Lepton managed GPU compute platform with run/status/cancel interface. Use when submitting TAO jobs to DGX Cloud, dispatching training/eval/inference to Lepton GPU resources, or managing Lepton workspace deployments. Trigger phrases include "run on Lepton", "submit to DGX Cloud", "Lepton job", "managed GPU on DGX Cloud".
Auto-scale LLM inference clusters on Kubernetes using KEDA, custom GPU metrics, and horizontal pod autoscaling. Handle traffic spikes, implement queue-based scaling, and optimize cost with spot instances for AI workloads.
NVIDIA CUDA GPU 並列コンピューティングリファレンス。 CUDA C++ / CUDA Python, kernel, nvcc, Unified Memory, CUDA Graphs, Cooperative Groups, Driver API, マルチ GPU。 PTX ISA 命令セット, state space, MMA 命令。 Blackwell チューニング, Streaming Multiprocessor, NVLink。 CUTLASS / CuTe DSL / CuTe C++ GEMM, Tensor Core, tcgen05, Operator API, nsight-compute, compute-sanitizer。
Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.