Token length exceeds during running simulation about drivelikeahuman

[zijinoier]

Hi Haowei:
Thanks for your interest of our project!

It appears that the issue you are facing is due to the high number of tools and ideas that need to be invoked in the decision-making process, causing the token limit of the OpenAI model to be exceeded before a decision can be made. Based on your error message, we guess you're using a gpt-3.5-turbo model with only 4k token contexts. However, we recommend using GPT models with at least 8k contexts.

For the OpenAI API, we suggest using either the gpt-3.5-turbo-16k-0613 or gpt-3.5-turbo-16k model, or any GPT4 model. You can refer to the following link for more details: https://platform.openai.com/docs/models/overview

Specifically, you can modify line 36 on HELLM.py with the following code:

llm = ChatOpenAI(
    temperature=0.,
    model_name='gpt-3.5-turbo-16k-0613', # or any other model with 8k+ context
    max_tokens=1024
)

from drivelikeahuman.

[miracelplus]

Thanks a lot! That solves my question.