I have a few questions on effectivity of fine tuning to create customized chat bot

Below are some of my questions for the case where I create a fine tuned model by plugging my training dataset which is in Hindi language

  1. Can I make model to override its original trained context and push my context in case of answering? How bigger training dataset is required to achieve this? Any additional customization is needed?
  2. How good is Chat GPT model in understanding context from Indian Languages, specifically hindi?