Accessing Arrow dataset cache_files

[Mehrad0711]

Describe the bug

In datasets 1.5.0 the following code snippet would have printed the cache_files:

train_data = load_dataset('conll2003', split='train', cache_dir='data')
print(train_data.cache_files[0]['filename'])

However, in the newest release (1.6.1), it prints an empty list.

I also tried loading the dataset with keep_in_memory=True argument but still cache_files is empty.

Was wondering if this is a bug or I need to pass additional arguments so I can access the cache_files.

[Mehrad0711]

Thanks @bhavitvyamalik for referencing the workaround. Setting keep_in_memory=False is working.