openai-python
openai-python copied to clipboard
Completion api injecting restart text after every prompt, no option to switch it off like in playground
I'm not sure this is an issue with the python api per se. When using the completion api to generate the next line of a poem using davinci, the api always starts the text with "\n\n". This can't be filtered with the logit parameter as the phrase "\n\n" is actually 4 tokens according to the tokenizer: [59, 77, 59, 77]
What is the correct workaround? Is there any plan to add the option "Inject restart text" like in the playground? The restart text adds 4 tokens to each request, which adds up fast...
Hi @drisme, thanks for writing in.
I'm not entirely sure I understand what the issue here is. Can you elaborate a bit more on the problem and the behavior you'd expect?
I'm going to close this due to inactivity. Feel free to reopen.