replace fp32 model with one with attention layers

onnx model with attention

rename q8 model

add quantized models

onnx model

initial commit