Skill UI

This skill covers deploying and managing Azure Virtual Machines using Azure CLI. It includes creating Linux and Windows VMs, configuring availability sets, managing managed disks, and generating custom images. It also provides guidance on selecting appropriate VM sizes for different workloads like general purpose, memory-intensive, or GPU tasks. Suitable for DevOps engineers and cloud administrators managing compute resources on Azure.

Automates installation, configuration, verification, repair, update, and uninstallation of Hyprland on Fedora Linux, with GPU-aware detection for NVIDIA, AMD, and Intel. Includes safety measures like backups and explicit user consent.

This skill covers deploying and managing virtual machines on Google Cloud Platform using Compute Engine. It includes creating instance templates, configuring managed instance groups for auto-scaling, and utilizing preemptible or spot VMs for cost optimization. Users will learn to provision GPU instances, manage snapshots, and apply startup scripts via gcloud CLI commands for production workloads.

Learn to operate GPU-backed Kubernetes clusters for AI inference and training workloads. Covers NVIDIA GPU Operator installation, device plugin configuration, MIG partitioning for resource sharing, and time-slicing for non-MIG GPUs. Includes GPU-aware autoscaling, health monitoring with DCGM and Prometheus, and troubleshooting scheduling or driver issues. Essential for production AI infrastructure management.

Provision, configure, and monitor NVIDIA GPU servers for AI inference and training workloads, including driver/CUDA installation, Docker GPU support, MIG partitioning, and health monitoring.

This skill enables AI assistants to interact with the Hugging Face Hub via Model Context Protocol (MCP). Users can search for models, datasets, and papers, retrieve repository details, fetch documentation, and execute compute jobs on GPU or CPU. It also allows leveraging Gradio Spaces as functional AI tools. Ideal for developers and researchers needing direct access to the HF ecosystem for model discovery, training, and deployment workflows.

Run evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware. Handles backend selection, local GPU evals, and choosing between vLLM, Transformers, and accelerate.

Run Python workloads on Hugging Face Jobs with managed CPUs, GPUs, TPUs, secrets, and Hub persistence. Execute batch inference, training, or recurring tasks without local infrastructure.

Train or fine-tune language and vision models using TRL or Unsloth on Hugging Face Jobs infrastructure. Supports SFT, DPO, GRPO methods and GGUF conversion for local deployment. Requires Pro account and HF_TOKEN for Hub saving. Optimized for cloud GPU training without local setup.

This skill helps users find the best AI models for specific tasks by querying Hugging Face benchmark leaderboards. It analyzes hardware constraints like GPU VRAM or CPU memory to filter compatible models. The output includes a comparison table with benchmark scores, parameter sizes, licenses, and device compatibility status. It also suggests whether models can run locally or require API access.

Build, deploy, and maintain machine-learning applications on Hugging Face Spaces using Gradio, Docker, or Static SDKs, with ZeroGPU and dedicated hardware, model loading, debugging, and inference providers.

Rules and patterns for building AI demos on Hugging Face Spaces with ZeroGPU hardware, covering @spaces.GPU, quota tuning, CUDA constraints, and concurrency safety.