Number of shards for imagenet dataset?

[biocyberman]

I downloaded the original imagenet dataset

1d675b47d978889d74fa0da5fadfb00e ILSVRC2012_img_train.tar
ccaf1013018ac1037801578038d370da ILSVRC2012_img_train_t3.tar
29b22e2961454d5413ddabcf34fc5622 ILSVRC2012_img_val.tar
e1b8681fff3d63731c599df9b4b6fc02 ILSVRC2012_img_test_v10102019.tar

After unpacking, I ran ./run makeshards.
Number of shards is only 147 for train and 6 for val. I wonder why nshards is set so high in this line:
https://github.com/webdataset/webdataset-lightning/blob/7b98a6a4e9e8735973f9de29151e6215380e5c9d/run#L3

and this line:
https://github.com/webdataset/webdataset-lightning/blob/7b98a6a4e9e8735973f9de29151e6215380e5c9d/train.py#L116

What is the dataset it is expecting, or what is the correct size of the shards?