TopFormer introduces a lightweight token–pyramid transformer that progressively merges local and global representations, achieving strong accuracy–efficiency trade-offs for mobile and edge vision tasks.

Original paper: TopFormer: Token Pyramid Transformer for Mobile Vision

This model uses the TopFormer-Base variant, which balances representational capacity and computational efficiency through a hierarchical token pyramid. It is well suited for on-device image classification and as an efficient backbone for downstream tasks where low latency and power efficiency are critical.

Model Configuration: