search

fastmachinelearning / hls4ml

Machine learning on FPGAs using HLS
https://fastmachinelearning.org/hls4ml
Apache License 2.0
1.22k stars 396 forks source link

How to match predictions in keras with predictions in Vivado HLS simulations #701

Closed riku238 closed 1 year ago

riku238 commented 1 year ago

I followed the hls4ml tutorial to create and synthesize a model. I have a question about the subsequent simulation in vibado hls. I input one part of the test data into "tb_input_features.dat" and ran the simulation using "myproject_test.cpp". The output of the simulation is quantized estimation, what does this mean? This quantized estimation does not match the value of the output predicted using "model.predict()" on the python notebook. What should I do?

thanks for watching!