Cat Carrier

Cat Carrier is anime style IllustriousXL model, Not NoobAI XL.

VAE is baked in.

Recommended Setting

• Steps: 25-30

• CFG scale: 5-7

• Sampler: Euler a, DPM++ 2M Karras

• Use ADetailer as needed

Positive Prompt

masterpiece, best quality, absurdres, very aesthetic, 

Negative Prompt

worst quality, bad quality, low quality, lowres, scan artifacts, jpeg artifacts, sketch, light particles, watermark,

Merge recipes (v9)

1. Merging the difference between Raehoshi illust XL v8.0 and stable-diffusion-xl-base-1.0, CyberRealistic XL v7.0 and stable-diffusion-xl-base-1.0 to Raehoshi illust XL v6.0 by "Perpendicular Component" and "Add Difference" (Checkpoint A)

2. Merging Checkpoint A to Raehoshi illust XL v6.0 by "SLERP", alpha = 0.5 (Checkpoint B)

3. Merging Checkpoint B to Raehoshi illust XL v6.0 by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)

4. Replacing the CLIP of Checkpoint C by Raehoshi illust XL v6.0 (Checkpoint D)

5. Finetuning Checkpoint D with anime style dataset(0.5k) by OFT method, dim = 4, alpha = 1e-3, learning rate = 2.5e-6, 15000 steps. And merging the OFT model to Raehoshi illust XL v8.0, ratio = 1.0 (Checkpoint E)

6. Merging Checkpoint E to Checkpoint D by "SLERP" with block (0,0,0,0,0,0,0,0,0,0,0,0.7,0.7,0.7,0.7,0,0,0,0,0) (Checkpoint F)

7. Merging the difference between Cat Carrier v8.0 and Raehoshi illust XL v7.1 to Checkpoint F by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) (Checkpoint G)

8. Merge the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint G by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) ) (Checkpoint H)

9. Merging the Checkpoint H by "Clamp", bounds = Checkpoint F, Cat Carrier v8.0, copycat-RouWei vpred-v0.42 (Checkpoint I)

10. Merging style LoRA to Checkpoint I (Cat Carrier v9.0)

Merge recipes (v8)

1. Merging the difference between RRoDIkeaV2 and Illustrious XL 1.0 to Raehoshi illust XL v7.1 by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) (Checkpoint A)

2. Merging the difference between Cat Carrier v6.0 and Raehoshi illust XL v5.1 to Checkpoint A by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) (Checkpoint B)

3. Merging the Checkpoint B by "Clamp", bounds = Raehoshi illust XL v7.1, RRoDIkeaV2, Cat Carrier v6.0 (Checkpoint C)

4. Merge the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Raehoshi illust XL v7.1 by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) ) (Checkpoint D)

5. Merging the Checkpoint D by "Clamp", bounds = Raehoshi illust XL v7.1, copycat-RouWei vpred-v0.42 (Checkpoint E)

6. Merging style LoRA to Checkpoint E (Checkpoint F)

7. Merging Checkpoint F to Checkpoint C by "Weighted Sum", alpha = 0.5 (Cat Carrier v8.0)

Merge recipes (v7)

1. Merging the difference between Raehoshi illust XL v6.0 and stable-diffusion-xl-base-1.0, CyberRealistic XL v7.0 and stable-diffusion-xl-base-1.0 to Raehoshi illust XL v6.0 by "Perpendicular Component" and "Add Difference" (Checkpoint A)

2. Merging Checkpoint A to Raehoshi illust XL v6.0 by "SLERP", alpha = 0.5 (Checkpoint B)

3. Merging Checkpoint B to Raehoshi illust XL v6.0 by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)

4. Replacing the CLIP of Checkpoint C by Raehoshi illust XL v6.0 (Checkpoint D)

5. Finetuning Checkpoint D with anime style dataset (0.5k) by full finetune, resolution = '1536,1536', optimizer = AdamW8bit, scheduler = warmup stable decay, learning rate = 1e-7, scheduler min lr ratio = 0.1, 62 epoch, unet only, gradient accumulation steps = 1, batch size = 3. (Checkpoint E)

6. Merging Checkpoint E to Raehoshi illust XL v6.0 by "Weighted Sum" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint F)

7. Merging the difference between LunarPeachMix v2.0 and Illustrious XL v1.1 to Checkpoint F by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0.3,0.3,0,0,0,0,0) (Checkpoint G)

8. Merge the difference between JANKU v4.0 and RouWei v0.7 epred to Checkpoint G by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0.3,0.3,0,0,0,0,0) (Checkpoint H)

9. Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint H by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.15,0.15,0.15,0.15,0,0,0,0,0) (Cat Carrier v7.0)

Merge recipes (v6)

1. Finetuning Raehoshi illust XL v5.1 with anime style dataset(0.5k) by OFT method, dim = 4, alpha = 1e-3, learning rate = 2.5e-6, 15000 steps. And merging the OFT model to Raehoshi illust XL v5.1, ratio = 1.0 (Checkpoint A)

2. Merge Checkpoint A to Raehoshi illust XL v5.1 by "Weighted Sum" with block (0,0,0,0,0,0,0,0,0,0,0,0.68,0.68,0.68,0.68,0,0,0,0,0) (Checkpoint B)c

3. Merge LunarPeachMix v2.0 to Checkpoint B by "SLERP" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0.3,0.3,0,0,0,0,0) (Checkpoint C)

4. Merge the difference between JANKU v4.0 and RouWei v0.7 epred to Checkpoint C by "Train Difference", alpha=0.25 (Cat Carrier v6.0)

Merge recipes (v5.1)

1. Merging plumMix v1.0 to Cat Carrier v5.0 by "SLERP", alpha=0.3 (Checkpoint A)

2. Merging LunarPeachMix v2.1 to Checkpoint A by "SLERP", alpha=0.5 (Cat Carrier v5.1)

Merge recipes (v5)

1. Merging the difference between CyberRealistic XL v5 and stable-diffusion-xl-base-1.0 to Illustrious-XL-v2.0-stable by "Add Difference" (Checkpoint A)

2. Replacing the CLIP of Checkpoint A by Illustrious-XL-v2.0-stable (Checkpoint B)

3. Finetuning Illustrious-XL-v2.0-stable with anime style dataset by OFT method, dim = 8, alpha = 1e-3, learning rate = 1e-4 and merging the OFT model to Checkpoint B, ratio = 1.0 (Checkpoint C)

4. Merge Checkpoint C to Checkpoint B by "Weighted Sum", alpha=0.3 (Checkpoint D)

5. Merge songMix v2.2 to Checkpoint D by "Weighted Sum" with block (0,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0,0,0,0,0) (Checkpoint E)

6. Merge LunarPeachMix v2.1 to Checkpoint E by "Weighted Sum", alpha=0.3 (Checkpoint F)

7. Merging style LoRA to Checkpoint F (Cat Carrier v5)

Merge recipes (v4)

1. Finetuning Illustrious XL v1.0 with anime style dataset by OFT method, dim = 8, alpha = 1e-3, learning rate = 1e-4

2. Merging the OFT model to Raehoshi illust XL v4.0, ratio = 1.0 (Checkpoint A)

3. Merge Checkpoint A to Raehoshi illust XL v4.0 by "Add Cosine B", alpha=0.25 (Checkpoint B)

4. Merging the difference between Cat Carrier v3.0 and Illustrious XL v1.1 to Checkpoint B by "Train Difference" with block (0,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0.2,0,0,0,0,0) (Checkpoint C)

5. Merging style LoRAs to Checkpoint C (Cat Carrier v4)

Merge recipes (v3)

1. Finetuning Illustrious XL v1.0 with anime style dataset by OFT method, dim = 8, alpha = 1e-3, learning rate = 1e-4

2. Merging the OFT model to Illustrious XL 1.0, ratio = 1.0 (Checkpoint A)

3. Merge Checkpoint A to Illustrious XL v1.0 by "Weighted Sum", alpha=0.75 (Checkpoint B)

4. Merging the difference between ExilluSPO Anime v1 and Illustrious XL v0.1 to Checkpoint B by "Train Difference", alpha = 0.2 (Checkpoint C)

5. Merging the difference between Catloaf and ponyDiffusionV6XL to Checkpoint C by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.4,0.7,0.7,0.7,0,0,0,0,0) (Checkpoint D)

6. Merge style LoRA to Checkpoint D (Cat Carrier v3)

Merge recipes (v2)

1. Finetuning Illustrious XL v0.1 with anime style dataset(0.5k) by OFT method, dim = 8, alpha = 1e-3, learning rate = 1e-4

2. Merging the OFT model to Raehoshi illust XL v1.0 spo edition, ratio = 1.0 (Checkpoint A)

3. Merging ExilluSPO Anime v1 to Checkpoint A, ratio = 0.2 (Checkpoint B)

4. Merging the difference between Catloaf and ponyDiffusionV6XL to Checkpoint B, Checkpoint B + (catloaf - ponyDiffusionV6XL) x alpha (0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.4,0.7,0.7,0.7,0.0,0.0,0.0,0.0,0.0,0), mode = Add difference, calcmode = trainDifference (Cat Carrier v2)

Merge recipes (v1)

1. Finetuning Illustrious XL v0.1 with anime style dataset(0.5k) by OFT method, dim = 8, alpha = 1e-3, learning rate = 1e-4

2. Merging the OFT model to Raehoshi illust XL, ratio = 1.0 (Checkpoint A)

3. Merging ExilluSPO Anime v1 to Checkpoint A, ratio = 0.2 (Checkpoint B)

4. Merging LoRA which is the difference between Catloaf and ponyDiffusionV6XL to Checkpoint B, ratio = 1.0 and lbw = [0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0,0,0,0] (Cat Carrier v1)

License

This model adheres to the Fair AI Public License 1.0-SD and is available for non-commercial use only. Commercial use is prohibited.