Update README.md

README.md

@@ -11,8 +11,23 @@ language:
 # Model Card for Model ID
 
 <!-- Provide a quick summary of what the model is/does. -->
+*実験モデルです / This is an experimental model.*  
+[lightblue/suzume-llama-3-8B-japanese](https://huggingface.co/lightblue/suzume-llama-3-8B-japanese)と、  
+[meta-llama/Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)の差分をchat-vectorアプローチで抽出し、  
+[meta-llama/Meta-Llama-3-70B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-70B-Instruct)に適用しました  
 
+## 手順/procedure
+- ja
+1. `meta-llama/Meta-Llama-3-8B-Instruct`と`lightblue/suzume-llama-3-8B-japanese`の差分を作成
+2. shapeが異なるので、差分をmeta-llama/Meta-Llama-3-70B-Instruct用にアップサンプリング
+3. 前から 8-layer、最後から8-layerはそのまま適用
+4. 中間layerを引き延ばして適用
 
+- en
+1. Create the difference between `meta-llama/Meta-Llama-3-8B-Instruct` and `lightblue/zume-llama-3-8B-japanese`
+2. Since the shapes are different, the difference is upsampled for meta-llama/Meta-Llama-3-70B-Instruct
+3. Apply the 8 layers from the front and 8 layers from the end as they are.
+4. Continue applying the middle layer
 
 ## Model Details