PaLM 2 chat-bison structured

Can someone help me to understand how make a structured dataset for chat-bison@2?

I read a lot of about https://cloud.google.com/vertex-ai/docs/generative-ai/model-reference/text-chat

but 1st not working

 

 

{"input_text": "sample input", "output_text": "sample ouput"}
{"input_text": "sample input", "output_text": "sample ouput"}
{"input_text": "sample input", "output_text": "sample ouput"}

 

 

 

2nd also not working

 

 

[
{"input_text": "sample input", "output_text": "sample ouput"},
{"input_text": "sample input", "output_text": "sample ouput"},
{"input_text": "sample input", "output_text": "sample ouput"}
]

 

 

3d also

 

 

{
  "context": "Your context here",
  "messages": [
    {
      "author": "user",
      "content": "User's message"
    },
    {
      "author": "assistant",
      "content": "Assistant's response"
    }
    // ... additional messages ...
  ]
}

 

 

Please send me format jsonl data for chat-bison. I found a lot of for text-bison but I want tune mode for chat-bison

0 1 409
1 REPLY 1

The dataset must be at jsonl format 

 
You can view some examples in the link provided.