batch size issue

[tina5218]

Hi Author:
I recently tried to run your program with my own data set and ran into a problem,
An error occurs when neither my training data nor my validation data is divisible by the batch size.
The test data is also.
At present, the number of my data is respectively, train: 10742, validation: 3200, test: 1579.
Setting batch_size to 32 will be wrong, but setting 2 is ok. The train model is ok, but the test model will be wrong.

Do you have any suggestions?

[rladomingo]

Hello. Were you able to fix the problem? I am also planning to use my own dataset, but the train, validation, and test sets will not be divisible by 32. I cannot change the partitioning since I need it for consistency in my experiments. Looking forward to your response. Thank you.