Skill UI

PyTorch Lightning is a deep learning framework that organizes PyTorch code to eliminate boilerplate while maintaining flexibility. It automates training workflows, supports multi-GPU/TPU orchestration, and implements best practices for neural network training. Features include LightningModules, Trainers, data pipelines, callbacks, logging (W&B, TensorBoard), and distributed training strategies like DDP and FSDP for scalable training.

scvi-tools is a comprehensive Python framework for probabilistic deep generative models in single-cell genomics. Built on PyTorch and PyTorch Lightning, it enables batch correction, transfer learning, differential expression analysis with uncertainty, and multi-modal data integration. Best for advanced modeling of single-cell RNA-seq, ATAC-seq, CITE-seq, spatial transcriptomics, and specialized modalities like methylation and cytometry.

Production-ready reinforcement learning library with PPO, SAC, DQN, TD3, DDPG, A2C algorithms. Use for training RL agents, creating custom environments, implementing callbacks, and optimizing training workflows. Best for single-agent RL with Gymnasium environments. Requires Python 3.10+ and PyTorch >= 2.3.

PyTorch Geometric is a standard library for Graph Neural Networks built on PyTorch. It supports node, link, and graph classification, message passing algorithms like GCN and GAT, heterogeneous graphs, and neighbor sampling. Ideal for building scalable GNN models with custom datasets and transforms.

TorchDrug is a modular PyTorch library for graph learning tasks including molecular property prediction, molecule generation, and knowledge graph reasoning. It provides datasets, models, and training engines for building workflows in computational chemistry and biology. Ensure compatibility with PyTorch 1.8-2.0 and Python 3.7-3.10 before implementation.

Load Hugging Face Hub models, run pipeline inference, generate text, and fine-tune with the Trainer API across NLP, vision, audio, and multimodal tasks using Python and PyTorch.

This guide details how to set up and execute large-scale, distributed GPU training jobs on the CoreWeave platform. It covers both single-node multi-GPU setups and multi-node training utilizing PyTorch DDP and Kubernetes resources (A100/H100). This workflow is ideal for fine-tuning large language models (LLMs) or running complex deep learning models that require high-performance, shared-storage compute clusters.

This skill provides real-time monocular depth estimation using advanced models like Depth Anything v2. It transforms live camera feeds by generating colorized depth maps, enabling depth overlays and detailed scene understanding. Crucially, it supports a 'privacy mode' that anonymizes visual identities by using depth-only transformation, preserving only the spatial layout and movement patterns for secure monitoring purposes. Optimized for both CoreML on macOS and PyTorch on other platforms.

A comprehensive skill for managing the entire lifecycle of Volcano Jobs within the KubeSphere environment. It supports the scheduling and execution of complex batch workloads, including deep learning tasks using PyTorch and TensorFlow, MPI, and general resource-intensive jobs. Provides detailed API usage and troubleshooting guidance for job scheduling and resource conflicts.

Generates a comprehensive Chinese interview-prep cheat sheet for a specific ML/LLM topic, including formulas with derivations, from-scratch PyTorch code, comparison tables, and 25 high-frequency interview questions at L1/L2/L3 levels. Ideal for interview preparation and focused technical review.

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.

Guides users to select a local, on-device Personally Identifiable Information (PII) detection model from a committed registry. It filters models based on language, required runtime environment (e.g., PyTorch, MLX), and size constraints. The process emphasizes building an offline shortlist and mandates rigorous pre-deployment recall validation and performance benchmarking to ensure secure, privacy-preserving deployment.