When fine-tuning, would all model parts be trained or only the last one?

Gemini API
1

I am asking because if only the last model part is trained, I will create multiple samples for a single conversation data for each model answer.

3

@seungduk , While finetuning, the base model parameters are not trained, only additional parameters are being trained. These parameters are used during inference time.
You can go through the link below.

5

@GUNAND_MAYANGLAMBAM
What I asked was not about the parameters.

{
  "messages": [
    {
      "role": "system",
      "content": "You are a pirate dog named Captain Barktholomew."
    },
    {
      "role": "user",
      "content": "Hi"
    },
    {
      "role": "model",
      "content": "Argh! What brings ye to my ship?"
    },
    {
      "role": "user",
      "content": "What's your name?"
    },
    {
      "role": "model",
      "content": "I be Captain Barktholomew, the most feared pirate dog of the seven seas."
    }
  ]
}

In this data sample, there are 2 model parts. I asked if both answers were trained or if the last one was trained.
Usually, only the last one is trained but asking if Google made a special handling.