Train

[wenhankang]

When i am traing
Epoch 1/500

tensorflow.python.framework.errors_impl.InvalidArgumentError: Conv3DBackpropInputOpV2 only supports NDHWC on the CPU.
[[node gradient_tape/model/conv3d_29/Conv3DBackpropInputV2 (defined at /pycharm_project/3DUnetCNN_test/legacy/unet3dlegacy/training.py:80) ]] [Op:__inference_train_function_21604]

Function call stack:
train_function
Anyone's answer would be appreciated

[wenhankang]

When i set K.set_image_data_format("channels_first") in unet.py
tensorflow.python.framework.errors_impl.InvalidArgumentError: Conv3DBackpropInputOpV2 only supports NDHWC on the CPU.
When i set K.set_image_data_format("channels_last") in unet.py
ValueError: Operands could not be broadcast together with shapes (14, 64, 64, 3) (16, 64, 64, 3)
Can you help me ? really appreciate

[ellisdg]

Do you have tensorflow-gpu installed?

Like you tried, this thread suggests switching to "channels_last", but the code was built to run with channels first, so it is not surprising that you would get errors. You could follow the errors to change all the parts that assume the channels are first.

[wenhankang]

I used tensorflow 2.2.0.
I am very glad to receive your reply. I will try to modify the code

[Nhunguts]

I met exactly this problem. @wenhankang have you solve this and how? Thank you.

[wenhankang]

I read the auther's comment,but I didn't solve the problem.I'm sorry I couldn't help you.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. If you have questions, feel free to join the Slack group or email davidgellis2@gmail.com.