Skill UI

This skill automates the process of renting, managing, and terminating cloud GPU instances on vast.ai. It analyzes your AI/ML task—determining VRAM, required GPU count, and estimated runtime—and then searches for the most cost-effective and reliable GPU offers, presenting optimized options for your training workflow.

Collection of battle-tested PyTorch training recipes covering architecture selection, optimizers, LR scheduling, precision tips, and debugging to train or fine-tune LLMs, vision, diffusion, and biomedical models with high GPU throughput.

Utilize Modal, a serverless GPU cloud, for running complex machine learning workloads, including training, fine-tuning, and inference. This tool provides a zero-config environment, eliminating the need for SSH setup or Docker management. It features automatic scale-to-zero billing and local-first execution, making high-powered GPU compute accessible for debugging and small-scale experiments.

PathML is a comprehensive Python framework designed for computational pathology workflows. It supports end-to-end analysis of Whole Slide Images (WSI), including loading 160+ proprietary formats, performing advanced stain normalization and tissue segmentation, constructing spatial cell graphs, and training deep learning models (e.g., HoVer-Net) for multiplex immunofluorescence and spatial proteomics. Ideal for large-scale biomedical image analysis.

A comprehensive framework designed to structure and automate PyTorch deep learning workflows. It simplifies boilerplate code by organizing models into LightningModules, automating training loops via the Trainer, and providing robust support for multi-GPU/TPU scaling (DDP, FSDP, DeepSpeed), data handling (DataModules), and best-in-class experiment logging (W&B, MLflow). Ideal for professional, scalable deep learning project development.

A robust orchestration skill designed for managing large-scale, complex machine learning experiments on remote GPU clusters via SSH. It handles multi-seed sweeps, sequential wave transitions, and deep learning pipelines requiring state management. Features include OOM-aware automatic retries, cleanup of stale terminal sessions, and deterministic scheduling to prevent race conditions, making it ideal for grid computing and staged training.

This guide provides a deep dive into migrating machine learning workloads, including inference services and training pipelines, from major hyperscalers (AWS, GCP, Azure) to the CoreWeave GPU cloud. It covers cost comparisons, containerization, Kubernetes YAML adaptation, and a phased deployment strategy, helping users ensure a smooth, cost-effective, and optimized transition.

A comprehensive checklist for ensuring GPU workloads, such as inference services or model training pipelines, are fully prepared for production deployment on CoreWeave. It covers critical MLOps and DevOps aspects including autoscaling, resource limits, security policies, persistent storage validation, monitoring setup, and rollback procedures.

A comprehensive tool for creating high-quality training datasets for custom object detection and segmentation models. It supports multiple advanced annotation methods including BBox drawing, SAM2 pixel-perfect segmentation, and DINOv3 visual grounding. Features include AI auto-detection, object tracking, and standardized COCO format export, streamlining the entire data pipeline from capture to training.

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.

This skill simulates a Chief Data Officer's deep interrogation, pressure-testing any plan touching core data strategy. Use it before approving ML model training runs, selecting data infrastructure (warehouse vs. mesh), productizing data assets, or making major team hires. It ensures that all data decisions are grounded in clear business decisions, compliance, and structural necessity.

This skill manages the entire lifecycle of long-running deep learning training and inference jobs deployed on remote, ephemeral computing environments (e.g., AutoDL, RunPod, bare SSH, Slurm, K8s). It provides critical capabilities like resilient checkpointing, managing spot instance preemption, debugging resource failures (OOM, NaN), and ensuring safe billing teardown, making it essential for professional ML Ops on rented resources.