Optimizes LLM inference with NVIDIA TensorRT for maximum throughput and lowest latency. Use for production deployment on NVIDIA GPUs (A100/H100), when you need 10-100x faster inference than PyTorch, or for serving models with quantization (FP8/INT4), in-flight batching, and multi-GPU scaling.
Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.
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.
High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
Search and download GIFs from Tenor using curl. No dependencies beyond curl and jq. Useful for finding reaction GIFs, creating visual content, and sending GIFs in chat.
PyTorch library for audio generation including text-to-music (MusicGen) and text-to-sound (AudioGen). Use when you need to generate music from text descriptions, create sound effects, or perform melody-conditioned music generation.
Sets up Portless for a project to replace port numbers with stable named .localhost URLs. Use when configuring local development routing, fixing port conflicts, or setting up monorepo dev environments.
Use when refreshing package-lock.json files in ng-mocks.
Increment the version of NNCF in src/nncf/version.py.
Install and manage Python packages using uv pip. Use when a Python import fails with ModuleNotFoundError, user asks to install a package, or a script requires a missing dependency.
Create new applications for ryOS following established patterns and conventions. Use when building a new app, adding an application to the desktop, creating app components, or scaffolding app structures.
Design and style UI components for ryOS following the 4 OS themes (System 7, macOS Aqua, Windows XP, Windows 98). Use when creating UI components, styling elements, working with themes, adding visual effects, or implementing retro OS aesthetics.
Detect and configure a conda-compatible tool, create the CausalPy environment, and run commands inside it. Use before any task that requires Python execution.
Adds a new compiler option to Metro.
Use when writing, modifying, porting, or optimizing CuTe DSL GPU kernels in Python; reading CuTe DSL API reference material; integrating a CuTe DSL kernel into a project; or rewriting an existing CUDA or C++ operator into CuTe DSL while preserving correctness and performance expectations.
Use when doing operator migration or kernel migration for CUDA, Triton, or custom ops in cache-dit; porting kernels from nunchaku, deepcompressor, or other repos; designing operator registration and public wrappers; wiring build and packaging for optional extensions; or reviewing an operator migration plan. Guides survey, minimal-closure migration, API design, extension loading, packaging, and layered validation. Do not use for blind copy-paste ports.
Use when asked to release Rsdoctor packages for a specific version. All `@rsdoctor/*` packages (except `@rsdoctor/mcp-server`) are versioned together via changesets.
This skill defines the only right architecture for android projects. Activate when doing big tasks (related to several files and their interconnections) in this project
Use when creating or developing, before writing code or implementation plans - refines rough ideas into fully-formed designs through collaborative questioning, alternative exploration, and incremental validation. Don't use during clear 'mechanical' processes
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes - four-phase framework (root cause investigation, pattern analysis, hypothesis testing, implementation) that ensures understanding before attempting solutions
Golang data structures — slices (internals, capacity growth, preallocation, slices package), maps (internals, hash buckets, maps package), arrays, container/list/heap/ring, strings.Builder vs bytes.Buffer, generic collections, pointers (unsafe.Pointer, weak.Pointer), and copy semantics. Use when choosing or optimizing Go data structures, implementing generic containers, using container/ packages, unsafe or weak pointers, or questioning slice/map internals.
