The value of run.usage.completion_tokens is 10% to 20% higher than the number of tokens shown in the OpenAI Dashboard. I don’t know where the bug is … in my code or in OpenAI Dashboard ? Which one is the real number?
The value of run.usage.completion_tokens is 10% to 20% higher than the number of tokens shown in the OpenAI Dashboard. I don’t know where the bug is … in my code or in OpenAI Dashboard ? Which one is the real number?