AI Skills for Hugging Face

Enables zero-shot image segmentation using Meta AI's Segment Anything Model, allowing flexible object detection without training.

Enables creation and management of datasets on Hugging Face Hub with SQL querying and transformation capabilities.

Facilitates structured evaluation management for Hugging Face model cards, enabling efficient data extraction and custom evaluations.

Tracks ML experiments using Trackio for logging, alerts, and metric retrieval, enhancing monitoring and diagnostics.

Enables training and fine-tuning of vision models on Hugging Face for tasks like detection and segmentation without local GPU setup.

Evaluates LLMs using 60+ benchmarks for model quality assessment, widely adopted in academic and industry settings.

Enables zero-shot image segmentation using Meta AI's Segment Anything Model, allowing flexible object detection in various domains.

Enables fast, calibration-free quantization of LLMs to 4/3/2-bit precision, optimizing memory and inference efficiency.

Generates high-quality images from text prompts using Stable Diffusion models, enabling creative workflows and image transformations.

LlamaGuard provides advanced content moderation for LLMs, ensuring safe interactions by filtering harmful inputs and outputs with high accuracy.

Mamba architecture offers a state-space model with O(n) complexity for efficient sequence modeling, enabling faster inference and long context handling.

Facilitates Hugging Face Hub operations via CLI for managing models, datasets, and repositories efficiently.

Connects AI assistants to Hugging Face Hub for model searches, dataset retrieval, and running compute jobs efficiently.

Enables users to search, download, and manage datasets from various repositories for machine learning projects.

Facilitates efficient uploads of Harbor parity experiment results to Hugging Face datasets using optimized git commands and LFS tracking.

Enables developers to search resources across GitHub, Stack Overflow, Hacker News, and HuggingFace for code examples and models.

Evaluates LLMs using 60+ benchmarks for model quality assessment and comparison, widely adopted in academic and industry settings.

Enables fast, calibration-free quantization of LLMs to 4/3/2-bit precision, optimizing model performance without dataset requirements.

Generates high-quality images from text prompts using Stable Diffusion models, enabling creative workflows and image transformations.

Optimizes reasoning efficiency in R1-style LLMs by reducing token usage during inference while maintaining accuracy.