Add some wonder to your generations. Trained on photos of large landscapes, garden photos, photos of wildflowers. Trained on tall portrait to wide landscape images. Consider 768x786 box and tall/short resolutions in that range (448x768, 768x448).

BOOscapes

unet: 0.0008 text encoder: 0.0001 scheduler: cosine_with_restarts optimizer: torch.optim.adamw.AdamW(weight_decay=0.01,betas=(0.9, 0.999)) network dim/rank: 16.0 alpha: 16.0 noise offset: 0.1 UNet attention weight average magnitude: 1.7017517682187613UNet attention weight average strength: 0.012998264675958538Text Encoder weight average magnitude: 1.7404300412039346Text Encoder weight average strength: 0.011727700822200437

Files with SHA256: 9c0569317bd70ad8676cf4ccdcef05bbe6307a2d643d958dc5c1031c859f1b99