5,250,968 AI resources indexedWorking toward the world’s largest AI resource index.
Reusable instructions for compatible AI agents.
Field counts come from the latest full-catalog snapshot. A resource can belong to several fields and platforms.
Systematically identify whether a GPU kernel is compute-bound, memory-bound, or latency-bound using roofline analysis, occupancy calculations, compute/load ratio per tile, and SASS instruction inspection. Produces a decision matrix for optimization strategy selection (cp.async, warp interleaving,...
Drive an agentic loop that iteratively optimizes a GPU kernel for maximum speedup. Use this skill whenever the user wants to optimize / speed up / benchmark a GPU kernel (CUDA, Triton, TileLang, C++, Python), mentions AKO / AKO4ALL / AKO4X / agentic kernel optimization, asks to "make this kernel ...
Code instrumentation for timing workloads. Two scenarios: (1) Training loop — inject manual timing to report per-iteration latency, throughput (samples/sec), and data load time. (2) Standalone kernel/op — write CUDA event timing code with warmup, per-iteration statistics, and anti-pattern avoidan...
CUDA/GPU computing guardrails, patterns, and best practices for AI-assisted development. Use when working with CUDA files (.cu, .cuh), or when the user mentions CUDA/GPU programming. Provides kernel design patterns, memory hierarchy guidelines, and occupancy optimization specific to this project'...
LLM inference in C/C++ with Python bindings. GPU acceleration via CUDA/Metal/Vulkan, 2-8 bit quantization (GGUF), KV cache, and grammar-based sampling. Run Llama, Mistral, Gemma, Phi locally.
Apply software pipelining (double-buffering) to a tiled GPU kernel to overlap global memory loads with Tensor Core computation. Covers prologue/loop/epilogue restructuring, LDG-register vs cp.async (LDGSTS) variant selection based on compute/load ratio, shared memory budget verification against a...
Use when adding, modifying, optimizing, or debugging CuTile autotuning code. Trigger signals: `exhaustive_search` / `replace_hints` / `hints_fn` / `cuda.tile.tune` in code, `autotune` in filenames, or correctness/performance issues in autotuned CuTile kernels. Covers: tune-once/cache/launch patte...
Optimizes PyTorch workloads using CUDA Graphs to enhance GPU performance and reduce overhead in training and inference processes.
Optimizes high-throughput LLM serving with efficient memory management and OpenAI-compatible API for production workloads.
Enables users to train a small GPT model from scratch, providing hands-on experience with LLM architecture and machine learning techniques.
Assists in building, training, and deploying neural networks using PyTorch for various applications like computer vision and NLP.
Enables real-time face swapping and deepfake video creation using a single image, ideal for applications in video effects and virtual try-ons.
Try a broader search or clear some filters.
Platform labels describe intended use. Read each resource's author credits, requirements and available access methods.