This project is for personal learning and use, based on MMPreTrain. It aims to track and reproduce the latest backbone models since MMPreTrain hasn’t released any new models for over a year.MMPreTrain is an open source pre-training toolbox based on PyTorch. It is a part of the OpenMMLab project.
The main branch works with PyTorch 1.8+.
🌟 (2024.12.5) Support StartNet Rewrite the Stars.
- More details StartNet.
🌟 Support CAS-ViT.
- More details CAS-ViT.
Below are quick steps for installation:
conda create -n BackboneHub python=3.8 pytorch==1.10.1 torchvision==0.11.2 cudatoolkit=11.3 -c pytorch -y
conda activate BackboneHub
pip install openmim
git clone https://github.com/bear-coder-9527/BackboneHub.git
cd BackboneHub
mim install -e .Please refer to installation documentation for more detailed installation and dataset preparation.
For multi-modality models support, please install the extra dependencies by:
mim install -e ".[multimodal]"The basic usage is same as MMPreTrain for new users:
For more information, please refer to our documentation.
Results and models are available in the model zoo.
This project is based on MMPreTrain. Thanks for their public repository and excellent contributions!
