TF Garden Post-Training Quantization Edge TPU

Hi community, I need help urgently. Thank you in advance!

Background:
I followed the TF Garden Tutorial for Object Detection (Model Garden overview  |  TensorFlow Core).

Model: ‘retinanet_mobile_coco’

Training and everything works. Also the model export is fine.

However when I want to do post training quantization my session crashes (Google Colab with 50GB RAM).

I get a bunch of these messages when loading model: WARNING:absl:Importing a function (__inference_internal_grad_fn_707881) with ops with unsaved custom gradients. Will likely fail if a gradient is requested.

And my Quantization cell using test data is based on Google Colab

Any help greatly appreciated!!!