The split of the femnist dataset train and test data in the file (femnist/split_data.py)

[Yupeng-mango]

Hi authors,

When I running ./data/femnist/preprocess.sh to get the femnist dataset, I met a problem about the split of train and test data.
Is there a small mistake in the file (femnist/split_data.py) ? The line 118 of the file: ./data/femnist/split_data.py should be test_indices = list(set(worker_indices) - set(train_indices)) instead of test_indices = list(set(indices) - set(train_indices)).

Thanks.