Skill UI

llama.cpp is a pure C/C++ library designed for efficient LLM inference, optimized to run on resource-constrained environments. It excels in scenarios where high-end NVIDIA GPUs (CUDA) are unavailable, making it ideal for Apple Silicon Macs, AMD/Intel GPUs, or edge deployment devices like Raspberry Pi. It supports GGUF quantization formats, enabling significant memory reduction and speed improvements for local, cross-platform AI applications.

RWKV is an RNN+Transformer hybrid architecture with linear-time inference, infinite context, and GPT-style training. It delivers constant-memory, streaming generation on CUDA and CPU, ideal for long-context, memory-constrained deployments.

A comprehensive guide for diagnosing and resolving common errors encountered when utilizing the Vast.ai platform. It provides quick references and fixes for HTTP status codes (e.g., 401, 429), instance lifecycle issues, CLI command failures, Docker image loading problems, SSH connectivity errors, and CUDA memory issues. Essential for maximizing GPU rental efficiency and debugging cloud instances.

Collects comprehensive diagnostic evidence for troubleshooting Vast.ai GPU instances. This bundle gathers account status, lists all running instances, retrieves recent logs, performs remote GPU diagnostics (e.g., nvidia-smi, CUDA checks), and verifies API connectivity. It is essential for creating detailed support tickets when encountering persistent infrastructure issues.

Comprehensive guide for upgrading the Vast.ai CLI and Python SDK. This skill helps users manage API version migrations, detect breaking changes, and update associated Docker images (e.g., CUDA versions). It ensures continuous compatibility when working with cloud GPU computing resources.

Accelerates CPU-bound Python workflows by translating NumPy, pandas, ML, graph, image, simulation, and IO code into NVIDIA GPU kernels using CuPy, Numba CUDA, Warp, cuDF/cuML/cuGraph, KvikIO, cuSpatial, and RAFT for dramatic speedups.

A comprehensive guide to diagnose and resolve common CoreWeave infrastructure errors. It covers issues related to GPU scheduling, Pod status (e.g., Pending), CUDA Out of Memory errors, NCCL timeouts, and Kubernetes resource allocation within AI/ML deployment pipelines.

A comprehensive utility designed for managing CoreWeave cloud infrastructure migrations. It facilitates upgrading between GPU generations (e.g., A100 to H100), updating CUDA versions, and handling Kubernetes resource changes. The tool uses K8s APIs to audit deployments, detect deprecated instances, and execute controlled rollbacks.

This skill provides comprehensive guidance for deploying and managing GPU-accelerated AI inference services on CoreWeave Kubernetes (CKS). It covers best practices including containerization using NVIDIA CUDA base images, configuring specific GPU resource limits (A100/H100), setting up robust health checks, and executing controlled rolling updates. Ideal for managing multi-model inference and scaling demanding AI workloads in a cloud environment.

This guide demonstrates how to deploy GPU workloads on the CoreWeave platform using Kubernetes (kubectl). It covers deploying vLLM inference services for large language model (LLM) testing and running batch CUDA jobs for benchmarking. This is ideal for users running their first GPU workloads, testing inference pipelines, or verifying cluster access on bare-metal GPU nodes (A100, H100).

This comprehensive skill manages the entire computer vision pipeline, from custom dataset annotation to model deployment. Users can fine-tune YOLO models on proprietary COCO-formatted datasets using an agent-driven workflow. Key features include hardware-aware training (supporting CUDA, MPS, ROCm), automatic export to optimized formats (TensorRT, CoreML, OpenVINO, ONNX), and seamless, one-click deployment as an active detection skill.

A state-of-the-art solution for real-time object detection using the advanced YOLO 2026 model. This skill detects 80+ COCO object classes (people, vehicles, animals) and outputs bounding boxes with high precision. It features automatic hardware acceleration, optimizing the model for various platforms (CUDA, MPS, OpenVINO, ROCm) to ensure maximum speed and efficiency, whether running on CPU, GPU, or Neural Engine.