Test tune of ~17K images trained at 1024x1024 with buckets using captions from CogFlorence-2.2 at batch size 4 for 4 epochs.
All images were manually filtered to the best of my ability to avoid low quality/watermarks/AI/etc.
All images are originally at least 1024x1024 total pixels.
I wanted to throw some fairly high quality but "general use-case" images at a model to use as a base so it can be further trained on more specific tasks in the future.
This version does not contain much ??????, and may be biased towards women. I am working on increasing the dataset size and variety, and optimizing the training settings if needed.
There is a v1, but it was trained at batch size 1 for 1 epoch with the text encoder LR at half the amount v1.1 was trained at, so I don't feel it is worth publishing.
1. Dieses Modell dient nur Lernzwecken. Urheber- und Auslegungsrechte liegen beim Originalautor.
2. Bist du der Originalautor eines Modells, kontaktiere uns bitte zur Authentifizierung über unsere offiziellen Kanäle. Wir schützen die Rechte aller Schöpfer. Hier klicken, um es zu verifizieren.
