Question regarding prompt token calculation


So I am observing that every time I make a request from playground, it takes entire text and calculates the token count. If there are repetitive requests, it will be an expensive in the long run. Is there a way to train a model to only take the user input as shown below in separator ("""") and generate output?

For example: here is my sample prompt:

Write an informative paragraph from below:

Custom t-shirt made with 100% cotton. Product is soft when wearing and made to last long time.

If you are looking for a custom shirt, this is a great product. The shirt is made of 100% cotton and is soft to wear. This shirt will last for a long time.

I feel like when using this in production, every request will be quite expensive.

Any suggestion?

1 Like

Thank you for taking time and explaining the process. So what is the setting I should adjust to get n > 1 responses?

1 Like