Hdf5 optimized using TensorRT

[ahnesathurai]

I am trying to use TensorRT to optimize a machine learning model. The file I have is in the format .hdf5, and I am successfully able to convert this to .pb. As I understand it, I need to then convert .pb to .onnx, then from .onnx I can use TensorRT. However, I am getting errors converting from protobuffer to onnx. I have looked at multiple example codes and it has not worked. Is anyone able to help me with this? I have attached the relevant .hdf5 file.
best_model.hdf5 (53.4 KB)

[AastaLLL]

Hi,

Suppose you can convert the hdf5 model into ONNX format directly.
Could you give it a try with the following tool?

GitHub - onnx/keras-onnx: Convert tf.keras/Keras models to ONNX

Thanks.

[ahnesathurai]

Hello, thanks for the reply. I tried converting the model, but I’m getting import errors. I tried fixing what I could, but I’m still getting an error importing tensorflow.python.keras.layers normalization_v2. Any ideas?

[AastaLLL]

Hi,

How do you install the TensorFlow package?

We just verified the Keras importing in our l4t-tensorflow:r32.7.1-tf2.7-py3 container on JetPack 4.6.1.
Maybe you can also give it a try.

ImportError: cannot import name 'dtensor'

Hi, Could you share how do you install the TensorFlow package? We confirm that importing Keras with the l4t-tensorflow:r32.7.1-tf2.7-py3 container is working. $ sudo docker run -it --rm --runtime nvidia --network host nvcr.io/nvidia/l4t-tensorflow:r32.7.1-tf2.7-py3 root@nvidia-desktop:/# python3 Python 3.6.9 (default, Dec 8 2021, 21:08:43) [GCC 8.4.0] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import keras >>> import tensorflow.keras >>> Thanks.

Thanks.