Open Iphykay opened 5 months ago
I found out what the issue was. I was down-sampling the data to the point the kernel can't do the convolution on the data. 3 layers is enough to prevent that error. Number of layers will depend on the grid size of the image.
You are correct! I think it can be closed @Iphykay.
Yeah sure thing.
Describe the bug I don't know why this happens but it does the first trial and then shows the error on the 2nd and 3rd trials. I don't know why it says the shape is changing. The input shape is not changing. This happens when it does the
tuner.search()
To Reproduce When you run the colab file, it runs well and sometimes it produces that error even when the input same is not changing. Unfortunately, I can't share my data, but i reproduced something close to it https://colab.research.google.com/drive/1RkpAoA-bunT0muFLYAtkKdM6F3WPp1YS?usp=sharing
Expected behavior No crash!
Additional context Tensorflow version: version 2.16.1 Keras version: 3.1.1 Keras-Tuner version: 1.4.7
Would you like to help us fix it? I don't know to be honest.