Not able to find the piece of code where fixed point computation is happening.

[Hanumanth004]

Hi, thanks for open sourcing this. I went through your paper, as well as the implementation in detail. I want to train the network for activation and weights at 8-bits. I have gone through your code in detail, but I do not find where fixed-point computation is happening. Based on your presentation and also your paper, I should be able to see int8 output (for my case). So, activations output after quantization should output int8; however, I have looked at the code as well as the tensorboard, but I see float32. Please let me know if I am missing something. Thank you for taking time and helping with this.