Details about ThreadDataLoader,CacheDataset,SmartCachedataset

[lanslotttTT]

Monai is really helpful for AI research in medical images, I was a little confused when I used ThreadDataLoader,CacheDataset,SmartCachedataset.My question is that:
1:How ThreadDataLoader it works? The parameter buffe_size is the size of the wait queue? what is the parameter repeats
used for? Why repeats data in this dataloader ？
2: CacheDataset ，the parameter num_worker is related to my cpu cores? How do I specify cache_rate in order to not fill my memory?
3:Does I need update data before every traning step when using SmartCachedataset? Since it only have a part of dataset.

[Nic-Ma]

Hi @lanslotttTT ,

Thanks for your interest here.
For your 3 questions, you can find the details in the doc-string:

1. https://github.com/Project-MONAI/MONAI/blob/dev/monai/data/thread_buffer.py#L82
2. https://github.com/Project-MONAI/MONAI/blob/dev/monai/data/dataset.py#L690, you need to try suitable cache_rate.
3. https://github.com/Project-MONAI/MONAI/blob/dev/monai/data/dataset.py#L836

Thanks.