Cat Tower is anime style NoobAI XL checkpoint.
VAE is baked in.
v-prediction versions are experimental models.
You need to use the webui that support v-prediction.
• ComfyUI (recommended)
• Forge NEO / Classic (recommended)
• reForge
• Forge (maybe)
• AUTOMATIC1111 (from v1.4, and you need to place the config file in the same folder as the checkpoint)
License
This model adheres to the NoobAI-XL License (Modified Fair AI Public License 1.0-SD) and is available for non-commercial use only. This license prohibits any form of commercialization, including but not limited to monetization or commercial use of the model, derivative models, or model-generated products.
Recommended Setting
• Steps: 25-30
• CFG scale: 5-7
• Sampler (vPred): I recommend using "Euler". Using "Euler a" sometimes generates colored skin and objects such as red, blue and so on.
• Sampler (Epsilon-pred): DPM++ 2M Karras, Euler a
• Scheduler (vPred) : Normal, Simple ("Automatic" produce noisy images.)
• Use ADetailer as needed
Positive Prompt
masterpiece, best quality
Negative Prompt
worst quality, bad quality, low quality, lowres, scan artifacts, jpeg artifacts, sketch, light particles
Merge recipes (ChenkinNoob v1.2)
• Full Finetuning ChenkinNoob-XL-V0.5 with anime style dataset(2.1k), optimizer = AdamW8bitKahan, scheduler = cosine, learning rate = 2.5e-6, 30 epoch, unet only, batch size = 2, gradient accumulation steps = 8. (Checkpoint A)
• Merging Cat Carrier v8.0 to Checkpoint A by "Weighted Sum" with block (0,0,0,0,0,0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0,0,0,0,0) (Checkpoint B)
• Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint B by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0,0,0,0,0,0,0) (Cat Tower ChenkinNoob v1.2)
Merge recipes (ChenkinNoob v1.1)
• Merging the difference between ChenkinNoob-XL-V0.2 and stable-diffusion-xl-base-1.0, CyberRealistic XL v7.0 and stable-diffusion-xl-base-1.0 to ChenkinNoob-XL-V0.2 by