Keras Inceptionv3 doesn't work on TensorRT (TX2)

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Hi,

My model involves keras inceptionv3 model as well as a few layers after inceptionv3. I followed the TensorRT workflow and converted my model to a uff. The uff works on tensorRT when running on GTX1080ti, but when i tried the uff on the tx2, it failed. (It keeps producing ~1 confidence score for the same class)

I understand that [url]https://github.com/NVIDIA-Jetson/tf_to_trt_image_classification[/url] has an inception model as well, but i realised that the model does not have the exact same layers as Keras’ inceptionv3 model.

Can someone shed some light as to why Keras’ Inceptionv3 model isn’t working on TX2 only? Thank you!