I contributed this collection containing 6 different ConvMixer models that were pre-trained on the ImageNet-1K dataset available for fine-tuning as well as image classification. Further, the models are also accompanied with a tutorial to help you get started in <5 minutes.
ConvMixer is a simple model that uses only standard convolutions to achieve the mixing steps. Despite its simplicity, ConvMixer outperforms ViT and MLP-Mixer. ConvMixer relies directly on patches as input, separates the mixing of spatial and channel dimensions and maintains equal size and resolution throughout the network.
https://tfhub.dev/rishit-dagli/collections/convmixer
The associated GitHub repo could be found here:
You might want to take a look at a ConvMixer implementation by @Sayak_Paul here and by @anon26514083 here.