Nerogar / OneTrainer

OneTrainer is a one-stop solution for all your stable diffusion training needs.
GNU Affero General Public License v3.0
1.81k stars 153 forks source link

Add cache-info.py utility #587

Open ppbrown opened 2 days ago

ppbrown commented 2 days ago

This shows association between cache directory, and backend data directory

pbrownrobo commented 2 days ago

FYI, output looks like:

python cache-info.py /data/OneTrainer/workspace-cache/run/image/6041f3ef3e050533dd0a7104500abeca0dc1751a98005abfcf06ccd7bc61994f |head -4 Unique directories found: /data2/CC/CC8M/CC8m-woman-40k/00000 /data2/CC/CC8M/CC8m-woman-40k/00001 /data2/CC/CC8M/CC8m-woman-40k/00002

Nerogar commented 1 day ago

There are quite a few issues:

ppbrown commented 1 day ago

interesting. What actual functionality does the script_imports()

call add?

On Sun, Nov 24, 2024 at 2:11 PM Nerogar @.***> wrote:

There are quite a few issues:

  • The script doesn't follow the usual import structure. It should start with.

    from util.import_util import script_imports script_imports()

  • Parameters should be parsed using a subclass of BaseArgs (see TrainArgs.py for an example usage), not by accessing sys.argv

  • You can't assume that image_path always exists in the cache dict. The cache structure depends on the model, and future supported models (like text2audio, text2video or something entirely different) won't have that key.

— Reply to this email directly, view it on GitHub https://github.com/Nerogar/OneTrainer/pull/587#issuecomment-2496262587, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANEV6MTHSXLBQNCSIOT2U32CJFJZAVCNFSM6AAAAABSLPOSJ2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIOJWGI3DENJYG4 . You are receiving this because you authored the thread.Message ID: @.***>