In data_utils.py, the last incomplete batch from the next() function has wrong shape (real batch is padded with zero values). This will affect to training and testing results and is caused by:
index_array, current_batch_size = next(self.index_generator), self.batch_size
Now the current batch_size is always the predefined self.batch_size, right?
I think the current batch size should be instead defined like this:
current_batch_size = len(index_array)
In data_utils.py, the last incomplete batch from the next() function has wrong shape (real batch is padded with zero values). This will affect to training and testing results and is caused by:
index_array, current_batch_size = next(self.index_generator), self.batch_size
Now the current batch_size is always the predefined self.batch_size, right?I think the current batch size should be instead defined like this:
current_batch_size = len(index_array)