How to expand Gemini output window

Gemini API
,
1

Is there a feature in Gemini similar to GPT4 “Continue generating” option?

When maximum output size is reached, is it because of computing costs or is there a practical obstacle in the model that does not allow it to generate more?

What would happen if maximum output tokens would be set to higher number?

I want to use it to translate longer texts and want the model to remember the settings it used to translate the first 8k tokens.

Thank you in beforehand for your advice.

1 Like
2

There are some technical issues about the output token size, and some of them are also aimed a limiting potential cases where it can get into an output token loop.

There is no way to increase the max tokens above the model limit, which is 8k tokens for the Gemini 1.5 models.

As with GPT, sending “continue” will usually have it continue where it left off.

1 Like
4

Thank you. However, sendfing “continue” does not seem to work with Gemini API.

po 7. 10. 2024 o 13:48 Allen Prisoner Firstenberg via Build with Google AI “Modified by moderator”

5

Hi @mam

You can implement prompt chaining by passing the output of one prompt as input to the next. In your case, once the token limit is reached, you can feed the previous output into a new prompt to continue the process. Once refer to this cookbook link

Thanks

7

Thank you. I tried to supply Gemini with its previous translation and the full text and asked it to continue translating, but it translated the same part of the text again.

ut 8. 10. 2024 o 7:16 Susarla Sai Manoj Kumar via Build with Google AI “Modified by moderator”

8

Need assistance how to stop

10

Hi @mam

Could you please share the sample code so I can understand the issue and help fix it? Sharing the code will help me reproduce the problem, find the cause, and provide a better solution.

Thanks!