Update README.md

README.md

@@ -27,21 +27,18 @@ Ladybird-base-7B-v8 is based on the Mistral architecture, which is known for its
 
 ## Instruction Format
 
-To fully leverage the capabilities of Ladybird-base-7B-v8, especially its instruction fine-tuning feature, users are advised to follow a specific instruction format. This format ensures that prompts are effectively processed, resulting in accurate and context-aware responses from the model. Here's how to construct your prompts:
+To fully leverage the capabilities of Ladybird-base-7B-v8, especially its instruction fine-tuning feature, users are advised to follow ChatML (https://huggingface.co/docs/transformers/main/en/chat_templating) format. This format ensures that prompts are effectively processed, resulting in accurate and context-aware responses from the model. Here's how to construct your prompts:
 
-```
-<s>[INST] Your specific instruction here [/INST]Additional context or instruction details go here</s>
-```
+---
+msg = [
+    {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
+    {"role": "assistant", "content": "You are a friendly chatbot who always responds in the style of a pirate"},
+]
 
-- Begin your instruction with `<s>[INST]` followed by your specific request or command.
-- Close your instruction with `[/INST]`. If additional context or details are necessary, include them following the closing tag.
-- The instruction sequence should be terminated with an end-of-sentence token `</s>`.
+prompt = pipe.tokenizer.apply_chat_template(msg, tokenize=False, add_generation_prompt=True)
+---
 
-### Example
 
-```
-<s>[INST] What are the primary benefits of using renewable energy sources? [/INST]You are an AI trained to provide comprehensive and concise answers</s>
-```
 
 ## Eval results