Why h5ad files directly downloaded using wget, can't be identified by abc_cache.get_data_path() function?

[zyll123]

I am downloading the WMB dataset. Because the abc_cache.get_data_path function is too slow, I used wget instead.
I put the wget files using the same folder patterns asked by abc_atlas_access.
After several days I downloaded data like WMB-10Xv3-MY-raw.h5ad and etc. However, when I used abc_cache.get_data_path to get the path, it just started to download all the files again, except for those downloaded by the function before. Why this happen? Can the function identify files that are manually downloaded?

[morriscb]

Hi, sorry the download took so long for you.

The cache object writes the path of the files it's downloaded to a file _downloaded_data.json. If they path isn't there the code will attempt to redownload the file. You can avoid this by using AbcProjectCache.from_local_cache(<path to your cache>) assuming the data you've downloaded matches the same format of folders as they are hosted on S3.