interrupted download results in stuck corrupted file

[CookiePPP]

https://github.com/XPixelGroup/BasicSR/blob/ce5c55aec0d14a399998a8c63fc650fdce3eb519/basicsr/utils/download_util.py#L94-L97

If the process stops while downloading a file, then the file will exist but be incomplete and the corrupted file will be returned the next time the function is called.

Solution:

tmp_file = cached_file + '.tmp'
download_url_to_file(url, tmp_file, hash_prefix=None, progress=progress)
os.replace(tmp_file, cached_file)

Download to a temporary path then move/rename to the output file.

---

https://pytorch.org/docs/stable/hub.html#torch.hub.download_url_to_file

The documentation also suggests using temporary files for this reason.

---

I'd make the PR myself but my clone is failing because of my messed up OS. Sorry for any inconvenience and thank you for reading/looking into this issue.