Hugging Face

Durable memory layer for coding agents with local indexing.

Fine-tuned NeoMME for visual document retrieval tasks.

Multimodal-native multilingual encoder for efficient AI model training.

Library for optimized WebGPU kernels from Hugging Face Hub.

Library for loading and running optimized WebGPU kernels.

Graph-based AI workflow tool for building and deploying pipelines.

Mutable, non-versioned object-storage for AI data management.

Automatically classifies prompts into specific Arabic dialects.

Allows leaderboard ranking by specific Arabic dialects.

Run agent rollouts in remote Hugging Face sandboxes for scalability.

On-device AI model for tool use and instruction following.

Advanced multilingual retrieval model with superior generalization capabilities.

Multilingual retrieval model with 307M parameters for AI tasks.

High-accuracy encoder for long-context inference on CPU.

Fast, efficient encoder for long-context inference on CPU.

Toolkit for quantizing AI models with SVDQuant in Diffusers.

Integrates Nunchaku 4-bit inference into Hugging Face Diffusers.

Ultra-fast vLLM inference using transformers modeling backend.

Open source skills for AI model deployment on AWS.

Integrates Hugging Face Storage with SkyPilot for AI workloads.

Cloud training without local GPU requirement.

Deployment workflow with DAgger-style corrections.

CLI for running six new simulation benchmarks.

Zero-shot reward model using log-probability.

General-purpose reward model scoring task progress.