I am working on the “Data Pipelines with TensorFlow Data Services” , week 2 assignment and the kernel keeps crashing in the Split the dataset
I suspect the issue is that the runtime is running a different version that the what is expected in the notebook.
Hi, and thank you for reporting! Will look into this and update here asap.
Hi. I cannot replicate the issue so it might be an error in your code. Please see the hints in the exercise cell. It might help guide you to the correct solution:
# Remember to use `cats_vs_dogs:4.*.*`
# https://www.tensorflow.org/datasets/catalog/cats_vs_dogs
# It has been downloaded for you so use the data_dir parameter (use 'data_dir' if running on Coursera, otherwise skip that parameter)
# else it will try to download the dataset and give you an error here
That means being mindful of the dataset version, and using the data_dir parameter.
We will modify the markdown in Step 0 nonetheless to avoid confusion. Hope this helps. Thanks!
Hi Chris,
I am using the data_dir with the correct version (attached photo). The kernel still crashes.
(Attachment PastedGraphic-1.tiff is missing)
Hi. Please send your notebook file to me via direct message. You can click my name then the blur Message button. I have to delete your other post with solution to follow the Forum guidelines. Will wait for your message. Thanks.
Resolved via direct message:
The error is from the data split. Only 20% of the images should be loaded. As per the instructions, you will extract it from the entire dataset as such:
You will split it up to get
The first 10% as the 'new' training set
The last 10% as the new validation and test sets, split down the middle
i.e. the first half of the last 10% is validation (first 5%)
the second half is test (last 5%)
If you load the entire set, the kernel will crash.