Set `add_special_tokens` inference parameter to `true`

README.md

@@ -61,6 +61,7 @@ widget:
       <|im_start|>assistant
 inference:
   parameters:
+    add_special_tokens: true
     max_new_tokens: 250
     penalty_alpha: 0.5
     top_k: 5
@@ -94,6 +95,7 @@ The recommended prompt format is as follows:
 To get the best results, prefer using [contrastive search](https://huggingface.co/docs/transformers/main/en/generation_strategies#contrastive-search) for inference:
 
 ```yml
+add_special_tokens: true
 penalty_alpha: 0.5
 top_k: 5
 ```