Open navotera opened 7 months ago
Hey @navotera this issue you are experiencing I believe is not really because of a problem with the UI, this is the normal behaviour when using the model called gpt-4
.
There are many questions online about that. For example: gpt-4-through-api-says-its-gpt-3 , anyone-getting-gpt-3-responses-when-calling-api-for-gpt-4-model, api-model-set-to-gpt-4-but-seems-to-respond-as-gpt-3
To see the expected result (the model saying it is based on chat GPT 4) use the model: gpt-4-turbo
for example.
Showing this behaviour locally:
Model gpt-4
:
Model gpt-4-turbo
:
Thank you for your response. It should be noted in the readme to avoid such confusion.
@navotera LLMs are generally not introspective, because their own parameters (model name, cutoff date, etc.) are not generally present in training data. there are many such tripping hazards for new users of LLMs, and i think this kind of knowledge is unfortunately out of scope for this repo. they would be better suited to tutorials or cheatsheets providing general information.
I have use the paid version API and when I change the version it does not change based on the ChatGPT respond.
Thank you for any of the support.