A CNN-Transformer hybrid tiny model combining Neural ODE and MHSA for modest FPGAs. Most of the model (except for pre- and post-processing layers) fits into on-chip SRAM of the FPGA. A related paper is available at arXiv.