open-mmlab/mmpretrain
PyTorch-based toolbox for pre-training and benchmarking vision models including vision transformers, MAE, CLIP, and self-supervised learning methods.
Velocity · 7d
+1.8
★ / day
Trend
→steady
star history
This repository is a pre-training toolbox and benchmark from the OpenMMLab ecosystem focused on computer vision models. It provides implementations of vision transformers (ViT, Swin, ConvNeXt), self-supervised learning methods (MAE, MoCo, BEiT), and multimodal models like CLIP. The toolbox supports training, evaluation, and fine-tuning of pretrained models for image classification tasks.