Repolode
Submit
apple

apple/corenet

A PyTorch-based toolkit from Apple for training small and large deep neural networks, including CLIP, LLMs, classification, detection and segmentation models.

View on GitHub
Stars
7,013
Forks
542
Open issues
18
Last commit
last month

What is it?

CoreNet is a deep neural network toolkit from Apple for training standard and novel models, from small to large scale, across tasks such as foundation models (CLIP and LLMs), object classification, object detection, and semantic segmentation. It is built on PyTorch and is driven by YAML configuration files that select models, datasets and training pipelines from its registries.

The repository contains training and evaluation recipes, along with links to pretrained weights, for a set of published Apple research projects, including OpenELM, CatLIP, FastViT, MobileOne, MobileViT and KV Prediction. Tutorials cover training a model on a new dataset, multi-node SLURM training, CLIP, semantic segmentation and object detection. Examples for running selected models efficiently on Apple Silicon with MLX are also included.

CoreNet evolved from the earlier CVNets library and extends it beyond computer vision to language and multimodal models.

Key features

  • 01Training recipes and pretrained weights for published models such as OpenELM, CatLIP and MobileViT.
  • 02Model, dataset and loss registries configured through YAML files.
  • 03Supports vision classification, detection, segmentation, CLIP-style multimodal and language models.
  • 04Multi-node training guidance, including SLURM.
  • 05MLX examples for running selected models on Apple Silicon.

Use cases

  • Training a classification or detection model on a custom dataset with reproducible configs.
  • Reproducing or extending published Apple research such as MobileViT or CatLIP.
  • Pretraining or fine-tuning CLIP-style or small language models at scale.
  • Running efficient inference of selected models on Apple Silicon via MLX.

Quick start

  1. 1

    Install Git LFS (Linux, using apt)

    sudo apt install git-lfs
  2. 2

    Clone the repository

    git clone git@github.com:apple/corenet.git
  3. 3

    Enter the repo folder and fetch the LFS files

    git lfs pull
  4. 4

    Install CoreNet in editable mode with pip (optionally inside a venv)

    python3 -m pip install --editable .

Things to know

  • The license is not clearly identified (listed as NOASSERTION), so check the LICENSE file before use.
  • The README recommends Linux with Python 3.10+ and PyTorch 2.1.0 or newer, and Git LFS is required to get the repository contents.
  • Audio and video features need extra system packages such as sox and ffmpeg.

In collections

Computer vision libraries & tools

Topics

#deep-learning#pytorch#computer-vision#llm#clip#object-detection#semantic-segmentation#training

FAQ

What is corenet?
A PyTorch-based toolkit from Apple for training small and large deep neural networks, including CLIP, LLMs, classification, detection and segmentation models. CoreNet is a deep neural network toolkit from Apple for training standard and novel models, from small to large scale, across tasks such as foundation models (CLIP and LLMs), object classification, object detection, and semantic segmentation. It is built on PyTorch and is driven by YAML configuration files that select models, datasets and training pipelines from its registries.
Is corenet free to use? What license does it have?
corenet is open source under the Other license. Check the license terms before using it in commercial products.
Is corenet still maintained?
Updated within 6 months. The last commit to corenet was on Sep 25, 2026.
What can I use corenet for?
Training a classification or detection model on a custom dataset with reproducible configs. Reproducing or extending published Apple research such as MobileViT or CatLIP. Pretraining or fine-tuning CLIP-style or small language models at scale. Running efficient inference of selected models on Apple Silicon via MLX.
How do I install corenet?
Install Git LFS (Linux, using apt): sudo apt install git-lfs Clone the repository: git clone git@github.com:apple/corenet.git Enter the repo folder and fetch the LFS files: git lfs pull Install CoreNet in editable mode with pip (optionally inside a venv): python3 -m pip install --editable .
Where can I get corenet?
The source code and installation instructions are on GitHub: https://github.com/apple/corenet

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects