EasyDict allows accessing dict values as attributes (works
recursively). It provides Javascript-like properties dot notation
for Python dicts.

Python library for reading and writing ML models in the GGUF (GGML
Universal File) format used by GGML-based inference engines such as
llama.cpp. It also ships command-line tools to dump metadata, edit
metadata values, copy a GGUF file with new metadata, and convert
endianness.

Kornia is a differentiable computer vision library built on PyTorch.

Python bindings for kornia-rs, a low-level computer vision library written in Rust.

LangSmith helps your team debug, evaluate, and monitor your language models and
intelligent agents. It works with any LLM Application, including a native
integration with the LangChain Python and LangChain JS open source libraries.

ModelScope is a Model-as-a-Service SDK for browsing, downloading and
running models from ModelScope Hub. This package ships the hub/library
core only: the cv, nlp, audio and related extras are not required.
The modelscope and ms commands delegate to python-modelscope-hub.

Python SDK and CLI to download, upload and manage models, datasets,
Studio spaces, skills and MCP servers on ModelScope Hub. Provides a
HubApi class and the modelscope-hub and ms-hub commands.

SGLang is a fast serving framework for large language models and
vision language models.

This is a CPU build: inference uses PyTorch's native CPU operators.
The Rust PyO3 extensions (multimodal preprocess, native gRPC) are
built. The embedded Rust server is not (dynamo-protocols 5.3.1
API drift). CUDA kernels, sgl-kernel and flashinfer are not.

Spandrel loads pre-trained PyTorch image models and auto-detects architecture from .pth / safetensors files.

Stable Diffusion is a latent text-to-image diffusion model. Thanks to a generous compute donation from Stability AI and support from LAION, we were able to train a Latent Diffusion Model on 512x512 images from a subset of the LAION-5B database. Similar to Google's Imagen, this model uses a frozen CLIP ViT-L/14 text encoder to condition the model on text prompts. With its 860M UNet and 123M text encoder, the model is relatively lightweight and runs on a GPU with at least 10GB VRAM. See this section below and the model card.

PyTorch Image Models (timm) is a collection of image models, layers, utilities, optimizers, schedulers, data-loaders / augmentations, and reference training / validation scripts that aim to pull together a wide variety of SOTA models with ability to reproduce ImageNet training results.

TorchAO is a PyTorch-native library for quantization and sparsity of model weights and activations, for both training and inference. CPU-only build for the SGLang cone.

TorchAudio applies PyTorch to the audio domain with autograd-friendly transforms and C++ extensions.

PyTorch implementation of differentiable SDE solvers with GPU support and efficient backpropagation.

The torchvision package consists of popular datasets, model architectures, and common image transformations for computer vision.

Simple and tiny yield-based trampoline implementation for Python.

Transformers provides thousands of pretrained models to perform tasks on
different modalities such as text, vision, and audio.

These models can be applied on text, for tasks like text classification,
information extraction, question answering, summarization, translation
and text generation; on images, for tasks like image classification,
object detection and segmentation; and on audio, for tasks like speech
recognition and audio classification.

The library is designed with two strong goals in mind: be as easy and
fast to use as possible, and provide state-of-the-art models with a
unified API.