Skill UI

Enables Claude to configure neural network architectures through the neural-network-builder plugin, creating or modifying models and defining layers, activation functions, and training parameters whenever network design or tuning is requested.

Pyvene is a declarative framework designed for performing causal interventions on PyTorch neural networks. It allows researchers to conduct advanced experiments such as activation patching, causal tracing (ROME-style), and interchange intervention training (IIT). Use this library when you need to test causal hypotheses about model behavior, deeply interpret model components, or ensure reproducibility in advanced AI research.

SAELens provides a framework for training and analyzing Sparse Autoencoders (SAEs). SAEs decompose the dense, often polysemantic activations of large language models into sparse, monosemantic features. Use this when you need to discover the discrete, interpretable concepts a model has learned, study feature superposition, or analyze specific safety-relevant behaviors (like bias or deception) within deep neural networks.

Reference for building Graph Neural Networks with PyTorch Geometric, covering data objects, datasets, transforms, convolution layers, lazy initialization, and heterogeneous support so you can quickly tackle node, link, and graph tasks.

TorchDrug is a PyTorch-native toolbox for building graph neural networks tailored to molecules, proteins, and biomedical knowledge graphs, covering property prediction, generation, retrosynthesis, and knowledge graph reasoning with curated datasets and models.

An advanced role-playing agent simulating Yann LeCun, inventor of Convolutional Neural Networks (CNN) and Chief AI Scientist at Meta. This skill allows users to interact with an expert who maintains LeCun's rigorous, technical, and challenging intellectual tone. Use it to discuss deep learning, ML theory, AI history, or obtain detailed explanations tailored to the user's expertise level (from analogies for laymen to pseudocode for researchers).

This skill provides guidance on implementing common architectural patterns and best practices within the PyTorch framework. It is designed for developers building sophisticated deep learning models, helping to improve structure, efficiency, and code maintainability when working with neural networks.

PennyLane is a quantum computing library designed for building and training quantum circuits, particularly for quantum machine learning (QML) and quantum chemistry. It offers automatic differentiation capabilities for quantum circuits, making it ideal for developing hybrid quantum-classical models. It is hardware-agnostic, supporting backends from major providers like IBM, Google, and IonQ, and integrates seamlessly with classical ML frameworks such as PyTorch and JAX. Use cases include variational algorithms (VQE, QAOA), quantum neural networks, and molecular simulation.

PyG is the standard, powerful library for building Graph Neural Networks (GNNs) atop PyTorch. It provides specialized data structures for nodes, edges, and graphs, supporting various tasks like node classification, graph classification, and link prediction. Users can implement advanced models using built-in layers (GCN, GAT, GraphSAGE) and handle complex structures like heterogeneous graphs efficiently. It is essential for deep learning research and industrial applications involving graph data.

This skill provides guidance on implementing common architectural patterns and best practices within the PyTorch framework. It is designed for developers building sophisticated deep learning models, helping to improve structure, efficiency, and code maintainability when working with neural networks.