FPGA-aware automatic acceleration framework for vision transformer with mixed-scheme quantization: late breaking results