prompt tokens length compute result is different from openai.ChatCompletion return which "name" in messages"

[TTurn]

I have problem in file "examples/How_to_format_inputs_to_ChatGPT_models.ipynb",

I use the method "num_tokens_from_messages" in file "examples/How_to_format_inputs_to_ChatGPT_models.ipynb" to calculate the length of prompt tokens，the result is different from the openai.ChatCompletion.create return when there have "function" role in messages, and the model is "gpt-3.5-turbo-0613".

When a "function" role appears, the length will differ by 2,

I think it's because of the value of the parameter "tokens_per_name" is set to -1 instead of 1.

example：
message = [{'role': 'function',"name":"get_info_from_web", 'content': '87°F'}, {'role': 'user', 'content': 'What is the weather like in Hangzhou today?'}]

the prompt_tokens in response of openai.ChatCompletion.create is 26,

but use num_tokens_from_messages the result is 28.

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 10 days.

[github-actions]

This issue was closed because it has been stalled for 10 days with no activity.