Skill UI

This comprehensive Python SDK serves as a client library for managing the full lifecycle of Machine Learning resources on Azure Machine Learning. It allows users to programmatically handle workspaces, register data and models, provision compute clusters, define complex training jobs, and build reproducible, end-to-end MLOps pipelines.

This comprehensive Python SDK serves as a client library for managing the full lifecycle of Machine Learning resources on Azure Machine Learning. It allows users to programmatically handle workspaces, register data and models, provision compute clusters, define complex training jobs, and build reproducible, end-to-end MLOps pipelines.

Bridges frozen vision encoders and LLMs for high-quality image captioning, VQA, retrieval, and multimodal chat, enabling zero-shot reasoning without fine-tuning and offering efficient Q-Former training.

Comprehensive skill set covering the full lifecycle of advanced visual AI systems. Expertise includes object detection (YOLO, Faster R-CNN), instance and semantic segmentation (Mask R-CNN, SAM), model training using PyTorch, and production deployment optimization using ONNX and TensorRT for edge or cloud targets.

This comprehensive workflow guides users through the end-to-end process of utilizing cloud-based GPU computing resources on Vast.ai. It covers searching for ideal GPU offers based on specs and price, provisioning an instance, running complex training or inference jobs via SSH, collecting model artifacts, and crucially, destroying the instance to prevent unexpected billing. Ideal for deep learning training and research.

A comprehensive guide for deploying Machine Learning training jobs and inference services on the Vast.ai GPU cloud. This skill covers the entire deployment lifecycle, including optimizing Docker images for performance, creating automated provisioning scripts using the Vast.ai CLI, implementing robust data transfer strategies (SCP, rsync), and performing post-deployment health checks.

A comprehensive skill for fine-tuning and training large language and vision models using advanced techniques like TRL (SFT, DPO, GRPO) or Unsloth. This process runs entirely on managed Hugging Face cloud GPUs, eliminating the need for local GPU setup. It supports modern training workflows, model conversion (GGUF for local deployment), and ensures seamless artifact storage to the Hugging Face Hub.

This skill provides a managed cloud environment for training and fine-tuning advanced vision models. Users can build and customize models for object detection (e.g., DETR, YOLOS), image classification, and segmentation (SAM/SAM2). By leveraging Hugging Face Jobs, training is executed on powerful cloud GPUs without requiring local GPU setup, and all resulting models are automatically saved to the Hugging Face Hub.

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 comprehensive guide for hardening deep learning papers, particularly in Computer Vision (CVPR). It details best practices for achieving rigorous reproducibility by utilizing the Compute Reporting Form (CRF), maintaining a detailed recipe ledger, ensuring benchmark split hygiene, and meticulously documenting all training hyperparameters, seeds, and variance analyses for academic publication.

A comprehensive guide for authors submitting computer vision papers (e.g., to ECCV). It outlines best practices for ensuring the full reproducibility of AI research, covering everything from detailed training recipes and hyperparameter logging to pinning foundation model dependencies and reporting variance honesty. Essential for academic publication integrity.