Why you can’t give an absolute dataset download link there? Why you must download the dataset by a python command? If I want to download through a single link and figure out what the link is, I should go and check the source code?
hacking the source is not that hard… here is the link https://router.sneakywines.me/minerl-v1/data_texture_0_low_res.tar.gz
How would you expect a link? It would be simple enough to have a data.url object with the link pointing to the latest dataset version.
Given the large size of the data, the download script was originally intended to use binary diffs to reduce the download time if errors were found or new data was to be added but the team has been hard at work on more critical features so this hasn’t been implemented yet!
True, it is really much faster than wget the link, and I know your team’s hardworking, thanks, but the python script can cause error, I don’t know why, I just can’t change the download directory, and cause the same error over and over again.
In this case, I think people need an alternative way to get the dataset, even take the cost of a longer download time.
Yes, thanks, can I ask where you track that link? I know the lib/python3.6/site-packages/minerl/data/download.py but even though I pip3 install --upgrade minerl first, what I can hack is also the https://router.sneakywines.me/minerl/data_texture_0_low_res.tar.gz not minerl-v1
Try using pip install - -upgrade minerl
Simply call pdb after line urls = [mirror + filename for mirror in mirrors] and print the urls. Btw, my minerl version is 0.2.2 but I think it is the same in 0.2.3