Menu
Skills in this repository
SkillsMP has collected 5,287 skills from TuYv/ccpm. Open a skill to review its source and details.
TuYv/ccpmShowing 40 of 5,287 collected skills.
Adds PyTorch FSDP2 (fully_shard) to training scripts with correct init, sharding, mixed precision/offload config, and distributed checkpointing. Use when models exceed single-GPU memory or when you need DTensor-based sharding with DeviceMesh.
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…
Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention framework. Use when conducting causal tracing, activation patching, interchange intervention training, or testing causal hypotheses about model…
Scalable data processing for ML workloads. Streaming execution across CPU/GPU, supports Parquet/CSV/JSON/images. Integrates with Ray Train, PyTorch, TensorFlow. Scales from single machine to 100s of nodes. Use for batch inference, data preprocessing,…
Distributed training orchestration across clusters. Scales PyTorch/TensorFlow/HuggingFace from laptop to 1000s of nodes. Built-in hyperparameter tuning with Ray Tune, fault tolerance, elastic scaling. Use when training massive models across multiple machines…
RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic…
Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory), deterministic vocabulary. Used by T5, ALBERT, XLNet, mBART. Train on raw text without pre-tokenization. Use…
Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs…
Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training…
Multi-cloud orchestration for ML workloads with automatic cost optimization. Use when you need to run training or batch jobs across multiple clouds, leverage spot instances with auto-recovery, or optimize GPU costs across providers.
Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling.
Provides guidance for experiment tracking with SwanLab. Use when you need open-source run tracking, local or self-hosted dashboards, and lightweight media logging for ML workflows.
Visualize training metrics, debug models with histograms, compare experiments, visualize model graphs, and profile performance with TensorBoard - Google's ML visualization toolkit
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),…
Provides guidance for PyTorch-native agentic RL using torchforge, Meta's library separating infra from algorithms. Use when you want clean RL abstractions, easy algorithm experimentation, or scalable training with Monarch and TorchTitan.
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.
Provides guidance for mechanistic interpretability research using TransformerLens to inspect and manipulate transformer internals via HookPoints and activation caching. Use when reverse-engineering model algorithms, studying attention patterns, or performing…
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works…
Expert guidance for fast fine-tuning with Unsloth - 2-5x faster training, 50-80% less memory, LoRA/QLoRA optimization
Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints,…
Track ML experiments with automatic logging, visualize training in real-time, optimize hyperparameters with sweeps, and manage model registry with W&B - collaborative MLOps platform
Update peon-ping configuration — volume, pack rotation, categories, active pack, and other settings. Use when user wants to change peon-ping settings like volume, enable round-robin, add packs to rotation, toggle sound categories, or adjust any config.
Log exercise reps for the Peon Trainer. Use when user says they did pushups, squats, or wants to log reps. Examples - "/peon-ping-log 25 pushups", "/peon-ping-log 30 squats", "log 50 pushups".
Rename the current Claude session for peon-ping notifications and terminal tab title. Use when user wants to give this session a custom name like "/peon-ping-rename Auth Refactor". Call with no argument to reset to auto-detect.
Toggle peon-ping sound notifications on/off. Use when user wants to mute, unmute, pause, or resume peon sounds during a Claude Code session. Also handles config changes like volume, pack rotation, categories — any peon-ping setting.
Set which voice pack (character voice) plays for the current chat session. Automatically enables session_override rotation mode if not already set. Use when user wants a specific character voice like GLaDOS, Peon, or Kerrigan for this conversation.
Remove backgrounds from images using local AI (rembg). Use when removing backgrounds from character art, mascot images, photos, or any image that needs a transparent background.
Edit images with precision — crop, resize, mirror, rotate, trim, and reframe. Use this skill whenever the user asks to crop, resize, trim, mirror, flip, rotate, reframe, or otherwise manipulate an image. Also use for creating square crops, portraits/headshots…
Generate character art and image variations using AI image generation (Google Gemini) with reference images for style and character consistency. Use this skill when the user asks to generate new character poses, mascot variations, art assets, illustrations,…
Update project documentation after code changes. Maintains the 4 core ai-context files (spec, project-structure, progress, deployment-infrastructure) and CLAUDE.md. Use after completing features, refactors, or any changes that affect project structure,…
Create and register new patches for tweakcc. Use when adding new customizations to Claude Code.
Interview the user about a plan or design until reaching shared understanding, resolving each branch of the decision tree. Use when user wants to stress-test a plan, get grilled on their design, or mentions "grill me".
Extract the JSDoc example nearest the active source selection or cursor into ./scratchpad as a TypeScript file. Use when the user asks to dump, copy, open, or try a source example in scratchpad.
Creates a Jobs to be Done canvas capturing the functional, emotional, and social dimensions of a customer job. Use when deeply understanding customer motivations, designing for jobs, or reframing product positioning.
Creates a clear problem framing document with user impact, business context, and success criteria. Use when starting a new initiative, realigning a drifted project, or communicating up to leadership.
Creates user-facing release notes that communicate new features, improvements, and fixes in clear, benefit-focused language. Use when shipping updates to communicate changes to users, customers, or stakeholders.
Creates an Architecture Decision Record following the Nygard format to document significant technical decisions, their context, and consequences. Use when making technical choices that affect system architecture, technology selection, or development patterns.
Documents the reasoning behind design decisions including alternatives considered, trade-offs evaluated, and principles applied. Use when making significant UX decisions, aligning with stakeholders on design direction, or preserving design context for future…