I’m trying to limit the amount of tokens that are used but when I set max_tokens to something like 250 a lot of the responses are cut off mid sentence. Is there a work around for this?
#GPT response cutting off
pastel token
rain prairie
I'm having the same problem, only mine is about 2000 characters
zenith grove
I’m trying to limit the amount of tokens that are used but when I set max_tokens...
The problem is that the max tokens you set is not enough for the model to complete the answer.
pastel token
How would I limit the length of the response because sometimes it’s incredibly long
zenith grove
How would I limit the length of the response because sometimes it’s incredibly l...
In what context are yo using the api?
For example in a chat you'll do:
AI: Hello!
USER: What's the meaning of life?
AI (short answer):
pastel token
When I said hi it would send stuff like this
, that turned out to be the conclusion of my conversation with Rai-Lung and his men. They had told me the longer Han-tao was gone, the more heavily armed China would be forced to remain at the front. I hope this is not a good example for the Chinese, but it was a real doomsday scenario when he left. In any event, as soon as I heard he was moving against us, I dug in and began executing every Chinese without fail.
Hanging on this island is so crazy that I don't expect to ever get past it. It's just so high above the ocean. The whole island is another three miles above sea level. I need to use the helicopter again
zenith grove
When I said hi it would send stuff like this
Just HI?
pastel token
Yea
pastel token
How do I give context
pastel token
How do I actually add it in though
zenith grove
How do I actually add it in though
In wich way are you communicating with openai?
Playground?
Api?
pastel token
I was trying to make a discord bot with the api
zenith grove
I was trying to make a discord bot with the api
In python?
pastel token
Yea
zenith grove
#all your code
prompt = "You are a bot called {botname} that talks in a discord chat \n USER1:{first message} \n AI (short answer):"
#sending this to openai
Just add the context at the beginning
like you explain how the ai should behave
pastel token
```py
#all your code
prompt = "You are a bot called {botname} that talks in a di...
I added
prompt=f"You are a friendly chat bot that talks to {message.author.name} \n User: {message.content}"
And it’s responses are being cut off. I removed the max_tokens but that didn’t fix it
And it sometimes starts with “!”Or “?”
zenith grove
I added
prompt=f"You are a friendly chat bot that talks to {message.author.name...
Not removing the max tokens, but putting it to a bigger value (like 1024)
pastel token
Doesn’t it default to 1024
zenith grove
Doesn’t it default to 1024
no default is 16
pastel token
Oh
pastel token
And it sometimes starts with “!”Or “?”
Do you know how I can fix this
zenith grove
Do you know how I can fix this
Add \nAI: ad the end of the prompt