Huggingface GitHub Repositories (2026) · Open Source Tools

Find why PyTorch training is slow, compare runs, and catch performance regressions in CI.

AI Commits for IntelliJ based IDEs/Android Studio.

Interactive 3D visualization platform for exploring transformer architectures, tensors, and real-time LLM inference.

⚡ Pure-Rust WebGPU inference engine — OpenAI-API compatible, GGUF native, runs on any GPU. No Python. No llama.cpp. Single binary.

Declarative AI pipelines in one YAML file. Tokens, audio chunks, and video frames flow between isolated components. Compose 100+ components for models, agents, speech, vision, and live broadcast. Run local models, cloud APIs, or both. Inspired by docker-compose.

LakeSoul is an end-to-end, realtime cloud-native Lakehouse framework for fast data ingestion, concurrent updates, incremental analytics, multimodal data processing and vector search — powering next-generation BI and AI workloads.

A library for training and deploying machine learning models on Amazon SageMaker

A CLI to estimate inference memory requirements for Hugging Face models, written in Python.

An Open-source, self-hosted AI model hub with Hugging Face compatibility, accelerating vLLM/SGLang performance.

Self-hosted artifact registry and pull-through cache in one binary. 17 formats: Docker/OCI, Maven, npm, PyPI, Cargo, Go, Raw, RubyGems, Terraform, Ansible Galaxy, NuGet, Pub, Conan, CPAN, RPM, Debian/APT, Hugging Face. No database, about 15 MB RAM idle, supply-chain curation built in.

Self-hosted artifact repository manager written in Rust. Hosts, proxies and groups 30+ formats: Docker/OCI, Maven, npm, PyPI, Cargo, Go, Helm, NuGet, Hugging Face, Terraform and more. PostgreSQL for metadata, blob stores on disk or S3. An open-source alternative to Nexus and Artifactory.

a self-hosted webui for 30+ generative ai

A full-stack AI helpdesk platform that uses machine learning, NLP, and OCR to automatically analyze support requests, detect similar incidents, and help teams resolve technical issues faster.

A speech-to-text system you can train for $25. A frozen speech encoder and a Qwen LLM joined by a small trained projector, with word timestamps, speaker diarization, a batched HTTP server, a live demo, and a free course.

Universal Python SDK to run AI workloads on Kubernetes

Laplace's Scissors: protected readout rotations for open LLMs. Diagnose weak output directions, repair answers, preserve successful generations, and analyze scaling deviations with Rust and Triton on CUDA/HIP.

From PyTorch model to end-to-end TensorRT inference experience in two commands—AI-native, cross-platform, and built for the best possible user experience.