Trained at 640, batch 4 X gradient accumulation 3, separately on SD1.5 and MeinaHentai then merged 70-30. 44 large images with automatic tags plus "microbangs" prepended.
Trained at 640, batch 4 X gradient accumulation 3, separately on SD1.5 and MeinaHentai then merged 70-30. 44 large images with automatic tags plus "microbangs" prepended.