PyTorch

Momentum

Each signal is ranked as a percentile among the technologies that have it. A signal without enough history is not counted, and never shown as zero. How momentum works

Official repository stars

Collecting

Star growth needs repository snapshots 30 days apart.103,616 stars now

Member repository stars

Collecting

Star growth needs repository snapshots 30 days apart.0 new members in 30 days

Package adoption

—Not published by its registries

Its packages’ registries publish no adoption figure, so there is nothing to grow.

Open models

—No official models

No official model on Hugging Face, so this signal does not apply.

News

Collecting

News counts once there are 3 events across the last 60 days; it has 2.2 in the last 30 days · 0 in the 30 before

Adoption by registry

The growth of PyTorch’s own packages, one row per registry in that registry’s own unit. Registries count different things over different windows, so they are never added together.

No registry has two observations of PyTorch’s own packages about 30 days apart yet, so there is no growth to show. Nothing is estimated in its place.

News activity

Events · 7 days

2published events that mention it

Events · 30 days

2published events that mention it

Previous 30 days

0events in the 30 days before

Momentum history

Momentum history starts OCT 8, 2026. A chart appears once two daily snapshots carry a score; none does so far.

Repositories

Official 1

• pytorch/pytorch

  Tensors and Dynamic neural networks in Python with strong GPU acceleration

  Official repositoryPython

  103.6K stars

Members 794repositories tagged with one of its aliases, most-starred first

• THUDM/SwissArmyTransformer

  SwissArmyTransformer is a flexible and powerful library to develop your own Transformer variants.

  Tagged pytorchPython

  1.1K stars

• datadreamer-dev/DataDreamer

  DataDreamer: Prompt. Generate Synthetic Data. Train & Align Models.   🤖💤

  Tagged pytorchPython

  1.1K stars

• microsoft/Windows-Machine-Learning

  Samples and Tools for Windows ML.

  Tagged pytorchC++

  1.1K stars

• AdaptiveMotorControlLab/CEBRA

  Learnable latent embeddings for joint behavioral and neural analysis - Official implementation of CEBRA

  Tagged pytorchPython

  1.1K stars

• bearpaw/pytorch-pose

  A PyTorch toolkit for 2D Human Pose Estimation.

  Tagged pytorchPython

  1.1K stars

• ndif-team/nnsight

  The nnsight package enables interpreting and manipulating the internals of deep learned models.

  Tagged pytorchPython

  1.1K stars

• ryfeus/lambda-packs

  Precompiled packages for AWS Lambda

  Tagged pytorchPython

  1.1K stars

• vietnh1009/Super-mario-bros-A3C-pytorch

  Asynchronous Advantage Actor-Critic (A3C) algorithm for Super Mario Bros

  Tagged pytorchPython

  1.1K stars

• gordicaleksa/pytorch-original-transformer

  My implementation of the original transformer model (Vaswani et al.). I've additionally included the playground.py file for visualizing otherwise seemingly hard concepts. Currently included IWSLT pretrained models.

  Tagged pytorchJupyter Notebook

  1.1K stars

• snap-research/EfficientFormer

  EfficientFormerV2 [ICCV 2023] & EfficientFormer [NeurIPs 2022]

  Tagged pytorchPython

  1.1K stars

• Res2Net/Res2Net-PretrainedModels

  (ImageNet pretrained models) The official pytorch implemention of the TPAMI paper "Res2Net: A New Multi-scale Backbone Architecture"

  Tagged pytorchPython

  1.1K stars

• mit-han-lab/gan-compression

  [CVPR 2020] GAN Compression: Efficient Architectures for Interactive Conditional GANs

  Tagged pytorchPython

  1.1K stars

Packages

“Is” packages are PyTorch itself; ecosystem packages depend on them. The two are counted separately and never added together. Adoption is each registry’s own figure, with its percentile within that registry.

Is 1the technology itself, most-adopted first

• torch

  Tensors and Dynamic neural networks in Python with strong GPU acceleration

  Listed in the catalogPyPI

  —Not published by PyPI

Ecosystem 582packages that depend on it, most-adopted first

• fedml

  A research and production integrated edge-cloud library for federated/distributed machine learning at anywhere at any scale.

  Depends on torchPyPI

  —Not published by PyPI
• flagembedding

  FlagEmbedding

  Depends on torchPyPI

  —Not published by PyPI
• flair

  A very simple framework for state-of-the-art NLP

  Depends on torchPyPI

  —Not published by PyPI
• flair-pos

  A very simple framework for state-of-the-art NLP

  Depends on torchPyPI

  —Not published by PyPI
• flash-attn

  Flash Attention: Fast and Memory-Efficient Exact Attention

  Depends on torchPyPI

  —Not published by PyPI
• flash-attn-4

  Flash Attention CUTE (CUDA Template Engine) implementation

  Depends on torchPyPI

  —Not published by PyPI
• flashinfer-python

  FlashInfer: Kernel Library for LLM Serving

  Depends on torchPyPI

  —Not published by PyPI
• flashrag-dev

  A library for efficient Retrieval-Augmented Generation research

  Depends on torchPyPI

  —Not published by PyPI
• flow-matching

  Flow Matching for Generative Modeling

  Depends on torchPyPI

  —Not published by PyPI
• freetoken

  Local MoE-offload LLM inference runtime with OpenAI- and Anthropic-compatible APIs

  Depends on torchPyPI

  —Not published by PyPI
• galore-torch

  GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection

  Depends on torchPyPI

  —Not published by PyPI
• gandlf

  PyTorch-based framework that handles segmentation/regression/classification using various DL architectures for medical imaging.

  Depends on torchPyPI

  —Not published by PyPI

Latest news the 3 latest events that mention it

• Announcing New vLLM and SGLang Inference Engines on TPU1 source
  22h agoOCT 24, 2026
• Integrating TorchTPU into TorchTitan and Optimizing Models for TPU2 sources
  2d agoOCT 22, 2026
• Profiling in PyTorch (Part 3): Attention is all you profile1 source
  4mo agoJUL 10, 2026