Tensorflow question (MaxPool3d, MaxPool3D, MaxPooling3D)!)

General Discussion
,
#1

Hi all,
I am trying to run an image against an existing model created with Python 3.7 / Tensorflow 2.7
(https://github.com/pboutinaud/SHIVA_PVS). There seems to be some confusion between MaxPool3d (in torch), MaxPool3D (in keras) and possibly MaxPooling3d.

I get this error and can not seem to solve it.

tensorflow.python.framework.errors_impl.InvalidArgumentError: No OpKernel was registered to support Op ‘MaxPool3D’ used by {{node model/Enc_Max_D7/MaxPool3D}} with these attrs: [ksize=[1, 2, 2, 2, 1], padding=“VALID”, T=DT_HALF, data_format=“NDHWC”, strides=[1, 2, 2, 2, 1]]
Registered devices: [CPU]
Registered kernels:
device=‘XLA_CPU_JIT’; T in [DT_FLOAT, DT_BFLOAT16, DT_HALF]
device=‘CPU’; T in [DT_FLOAT]

If anybody has any ideas on how to troubleshoot this I would be very grateful!

Roger

#3

On the CPU, you can try using data type float32. Thank you.

#5

Hi Roger, welcome to the forum!

I didn’t understand some details,

Are you converting the model from PT to TF? or is the model already converted?
or did you train the model from scratch?

#6

Hi all,
Thank you for your replies!
According to the developer, “the MaxPool3D layer is not correctly implemented for the CPU version of tensorflow 2.7, I couldn’t find an easy way to fix that and I am currently waiting for tensorflow to fix that problem in a future version.”
Alas!
R

#8

Can you try with the latest version TF2.10 and let us know if the issue still continues? Thank you!