README.md

metadata

license: creativeml-openrail-m

NAIの含まれる各種モデル(any系)をACertaintyベースで再現しようという試みです。
レシピに表記されているチェックポイントにNAIが含まれていなければNAIリークフリーのモデルになります。
マージはよく分からないので適当にマージしてください。
蒸留画像は使用していません。下記のレポジトリからデータセットのキャプションのみダウンロードできます。
DataSet: https://huggingface.co/datasets/paimonimpact/ONN

ONN_anyV3.fp16.safetensors

▼Use Models

• ACertainty.ckpt
• bp_1024_e10.ckpt

ACertaintyにany_A ~ FのデータセットでDB。

Model: A	Model: B	Weight	Base alpha	Merge Name
A	C	-	0.3	AC
B	E	-	0.5	BE
D	F	-	0.5	DF
BE	DF	0.0,0.0,0.1,0.3,0.5,0.7,1.0,0.7,0.5,0.3,0.1,0.0,0.0,0.0,0.1,0.3,0.5,0.7,0.5,0.3,0.1,0.0,0.0,0.0,0.0	0.3	BEDF
BEDF	AC	1.0,0.9,0.7,0.5,0.3,0.1,0.3,0.3,0.3,0.3,0.3,0.3,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.1,0.3,0.5,0.7,0.9,1.0	0.3	BEDFAC
BEDFAC	bp_1024_e10	-	0.15	ONN_anyV3

ONN_AOM2.fp16.safetensors

▼Use Models

• ONN_anyV3.fp16.safetensors
• Bra6-2(beta).safetensors
• instagram-latest-plus-clip-v6e1_50000.safetensors
• dreamshaper_631BakedVae.safetensors

Model: A	Model: B	Weight	Base alpha	Merge Name
instagram-latest-plus-clip-v6e1_50000	Bra6-2(beta)	0,0.0,0.0,0.0,0.0,0.1,0.3,0.5,0.7,0.5,0.3,0.1,0.0,0.0,0.1,0.3,0.5,0.7,0.9,0.7,0.5,0.3,0.1,0.1,0.0,0.0	0(cosineB)	insta_bra

Model: A	Model: B	Model: C	Interpolation Method	Merge Name
insta_bra	dreamshaper_631BakedVae	v1-5-pruned	Add Difference @ 0.7	onn_real

Model: A	Model: B	Weight	Base alpha	Merge Name
ONN_anyV3	onn_real	0,1.0,0.9,0.7,0.5,0.3,0.1,1.0,1.0,1.0,1.0,1.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.1,0.3,0.5,0.7,0.9,1.0	0	ONN_AOM2

ONN_anyV4.fp16.safetensors

▼Use Models

• ACertainty.ckpt
• ONN_AOM2.fp16.safetensors

1. ACertaintyをanyV4.zipのデータセットでFT = FT_ACertainty
2. ONN_AOM2と単純マージ

   Model: A	Model: B	Base alpha	Merge Name
   FT_ACertainty	ONN_AOM2	0.5	ONN_anyV4

ONN_pastel.fp16.safetensors

▼Use Models

• bp_1024_e10.ckpt
• ONN_AOM2.fp16.safetensors
• onn_real.fp16.safetensors

1. pastel_A ~ EのデータセットでLoRaを5つ作成 = onnpastelLoRaA~E

2. ONN_AOM2とpastelLoRAをマージ

   Model	Lora	Weight	Merge Name
   ONN_AOM2	onnpastelLoRaA	0.8	onnpastel_baseA

3. onnpastel_baseAとbp_1024_e10.ckptをマージ

4. Model: A	Model: B	Base alpha	Merge Name
   onnpastel_baseA	baseAとbp_1024_e1	0.5	onnpastel_baseB

5. ACertaintyをany_C.zipのデータセットでFT = onnpastel_baseC

6. onnpastel_baseBとonnpastel_baseCをマージ

   Model: A	Model: B	Base alpha	Merge Name
   onnpastel_baseB	onnpastel_baseC	0.5	onnpastel_baseBC

7. onnpastel_baseBCとonn_realをマージ

   Model: A	Model: B	Weight	Base alpha	Merge Name
   onnpastel_baseBC	onn_real	1.0,1.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0	0	onnpastel

8. 各種LoRAをマージ

   Model	Lora	Weight	Merge Name
   onnpastel	onnpastelLoRaB	0.2	onnpastel-1
   onnpastel-1	onnpastelLoRaC	0.3	onnpastel-2
   onnpastel-2	onnpastelLoRaD	0.5	onnpastel-3
   onnpastel-3	onnpastelLoRaE	0.6	onnpastel-4
   onnpastel-4	onnpastelLoRaA	0.2	onnpastel-5

9. 再度マージして微調整

   Model: A	Model: B	Weight	Base alpha	Merge Name
   onnpastel-5	onnpastel_baseC	0.7,0.5,0.3,0.1,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.1,0.3,0.5,0.7	0	ONN_pastel