Update README.md

README.md

@@ -1,15 +1,17 @@
 ---
 license: creativeml-openrail-m
 ---
-ACertaintyベースのモデル群です。  
+NAIの含まれる各種モデルを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。  
-DataSet(キャプションのみ): [https://huggingface.co/datasets/paimonimpact/ONN](https://huggingface.co/datasets/paimonimpact/ONN)    
+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](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。  
+ACertaintyにany_A ~ FのデータセットでDB。  
 
 | Model: A     | Model: B | Weight                                                                | Base alpha | Merge Name          |
 | ------------ | -------- | --------------------------------------------------------------------- | ---------- | ------------------- |
@@ -37,4 +39,54 @@ ACertaintyにany_A~FのデータセットでDB。
 
 | 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_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 |