Trouble running the exact same code in PyCharm

Course Q&A TensorFlow Developer Professional Introduction to TF for Artificial Intelligence ...
1

Hi guys,

Thank you for the great course content. Parallel to doing the exercises on Google Colab, I am running the exercises on my computer. I have bumped into a very odd error when I run the convolutional neural network exercise on my computer. The error reads thusly: “Error when checking input: expected conv2d_input to have shape (28, 28, 1) but got array with shape (60000, 28, 28)”.

I have no idea why the exact same code runs without a problem on google colab notebook but not in PyCharm.
Could you guys by any chance help me with this?
The code is in the third block of C1_W3_Lab_1_improving_accuracy_using_convolutions.ipynb notebook.

Thanks in advance.
Mahmood

2

You should reshape the training data to (60000, 28, 28, 1). While greyscale is obvious to us, the TF needs to see it as a separate channel. Use x = x.reshape(60000, 28, 28, 1)

3 Likes
3

Thanks balaji! That indeed solved the problem.

4

That may have solved the runtime problem, but it didn’t answer the question asked, which was why it runs on colab but not in a PyCharm editor. Is it a package version difference? Seems like kind of a big change.