Изображения
diffusers
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.
★ 34,6 тыс. звёздApache-2.0
Смотреть руководство →Transformers — проект с открытым исходным кодом из категории «Инструменты разработчика» от huggingface (huggingface/transformers на GitHub). У него 166 942 звезды на GitHub и 34 748 форков, он распространяется по лицензии Apache-2.0 и последний раз обновлялся на GitHub 4 октября 2026.
The model-definition framework behind modern AI — thousands of pretrained transformers for text, vision, and audio.
Text, image, audio, or multimodal — find the pipeline that matches what you want to build.
One line downloads a pretrained checkpoint from the Hugging Face Hub.
Call the pipeline on your input; batching and device placement are handled for you.
Use the Trainer API for custom training, then export for serving.
One API across BERT, GPT, Llama, Whisper, ViT, and thousands more — swap checkpoints without rewriting code.
Task-based one-liners for classification, generation, transcription, and vision — from prototype to demo in minutes.
PyTorch, TensorFlow, and JAX backends; export to ONNX and TorchScript for production serving.
A full training loop with mixed precision, distributed training, and experiment tracking built in.
Blazing-fast Rust tokenizers with Python bindings — correct encoding for every model family.
The default dependency of the open AI world: tutorials, integrations, and answers for everything.
Pip-install the library plus your framework of choice.
pip install transformers torchChoose the task; the docs list every supported pipeline with examples.
from transformers import pipelineLoad any Hub model by name and run it locally or on your GPU.
pipe = pipeline('summarization'); print(pipe(long_text))From the tooling end of the shelf, Pacwich handles monorepos that run on bun and pnpm — the unglamorous plumbing larger projects end up needing. Readers working through the Dev Tools shelf also stop at @shadcn/lint, an agent-first linter for Tailwind design systems: you define what is allowed, and errors explain the fix in terms of your components. Also on this shelf: phoenix — a code editor designed to make coding as fun as playing a video game.
It is an open-source project at huggingface/transformers with 34,742 forks. Check the repo's contributors and recent commits to see how active maintenance is.
Yes — Apache-2.0. Individual model checkpoints carry their own licenses, so check each one.
Inference on small models works on CPU; training and large models want a GPU.
Both are supported; PyTorch is the community default today.
The Hugging Face Hub hosts hundreds of thousands of checkpoints loadable by name.
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.
Build resilient agents
Run and fine-tune LLMs and diffusion models locally — faster training, smaller memory footprint.
Google's cross-platform framework for on-device ML perception: vision, text and audio pipelines
Adaptive web scraping framework in Python — from a single request to a full-scale crawl
Google's Python library for extracting structured data from unstructured text with LLMs, with grounding
Другие проекты из категории «Инструменты разработчика» с наибольшим числом звёзд в каталоге LibreAI — реалистичный список альтернатив, если Transformers вам не подходит.
The open source coding agent
Anthropic's agentic coding tool that lives in your terminal and works across your whole codebase
Cross-platform desktop all-in-one manager for Claude Code, Codex, OpenCode and other coding agents
OpenAI's lightweight coding agent that runs locally in your terminal
Turn any codebase, docs and schemas into a queryable knowledge graph for your coding assistant
Соседние проекты из категории «Инструменты разработчика» в каталоге, ближайшие по числу звёзд к Transformers — продолжайте изучать полку по обе стороны от него.
The open source coding agent
Anthropic's agentic coding tool that lives in your terminal and works across your whole codebase
Extract documentation for AI agents from any site with llms.txt support.
Cross-platform desktop all-in-one manager for Claude Code, Codex, OpenCode and other coding agents
A cost-effective terminal AI coding assistant in Go.
OpenAI's lightweight coding agent that runs locally in your terminal
Бесплатно и открыто — забирайте исходный код, ставьте звезду репозиторию и начинайте создавать.