vram-gpu-oom
GPU VRAM management patterns for sharing memory across services (Ollama, Whisper, ComfyUI). OOM retry logic, auto-unload on idle, and service signaling protocol.
发现 tools framework internals 相关 Skills,比较来源仓库和用途,查看适合你工作流的技能说明。
GPU VRAM management patterns for sharing memory across services (Ollama, Whisper, ComfyUI). OOM retry logic, auto-unload on idle, and service signaling protocol.
Gradient-based approximation to activation patching for scalable circuit analysis. Use when activation patching is too slow or when analyzing many components simultaneously.
Write correct, idiomatic Apple MLX code for Apple Silicon ML. Use when working with MLX arrays, neural networks, training loops, lazy evaluation, unified memory, mx.eval, mx.compile, Metal GPU, memory optimizat...
Expert knowledge of Godot performance optimization, profiling, bottleneck identification, and optimization techniques. Use when helping improve game performance or analyzing performance issues.
AI-supervised development workflow. Use when running prep, cook, serve, tidy, or work commands, managing beads issues during sessions, or following prep→cook→serve→tidy cycle. Covers workflow orchestration, gua...
Modern Python language features and typing patterns. Use when writing type hints, using generics, implementing pattern matching, working with async/await, or leveraging Python 3.10+ features.
GreyCat C API and GCL Standard Library reference. Use for: (1) Native C development with gc_machine_t context, tensors, objects, memory management, HTTP, crypto, I/O; (2) GCL Standard Library modules - std::cor...
UV/UVX/Ruff toolchain for DiscoHy Thread Operad with Python packaging and linting
Enzyme.jl Automatic Differentiation Skill
HVM Runtime Skill
DiscoPy Operads Skill
DisCoPy categorical color streams via Hy with 3 parallel TAP streams and 7 operad variants
oapply: Operad algebra evaluation via colimits. Composes machines/resource sharers.
WebGPU tile rendering for Rio Terminal via wgpu and sugarloaf. Extends
Reverse Engineering Skill
MLX on Apple Silicon with JAX-style SplitMix64 PRNG. Deterministic color generation with GPU acceleration.
Unified homoiconic graph rewriting - λ-calculus, interaction nets, ACSets, CUDA parallelism
Unified Specter/Navigator/3-MATCH architecture with bidirectional path compilation
Overcoming fuzzing obstacles and improving fuzzer effectiveness.
Specter-style bidirectional navigation for Julia Collections, S-expressions, and ACSets with inline caching
Self-inverse derivation patterns where ι∘ι = id for frame-invariant self
cat Skill: Derivational Pipe Chaining
ASI skill integrating polynomial functors, free monad/cofree comonad
UV/UVX/Ruff toolchain for DiscoHy Thread Operad with Python packaging and linting
默认按已收录数据中的 GitHub 仓库 Stars 排序。Stars 属于来源仓库,同一仓库的多个技能可能共享相同数值;它反映仓库关注度,不代表单个技能的安装量或质量评分。