When I try to visualize the embeddings endpoint in 3D space as was done in this article: (Introducing Text and Code Embeddings in the OpenAI API) I am unable to use a JSONL file with more than around 38 examples, otherwise it times out and I get an error stating “RetryError: RetryError[<Future at 0x1fb37e201c0 state=finished raised InvalidRequestError>]”. I can get it to work on the limited examples, but does anyone know why this could be
1 Like
The rate limit on the embeddings endpoint is the same as the completions endpoint. 38 examples shouldn’t trigger it.
@curranjanssens Can you tell me a bit more about what you’re trying to do? I’m not sure what would be generating that error.
3 Likes
I found the issue, One of the lines had (“text”: “”) as one of the JSONL inputs, I didn’t know that having no character value would throw an issue. It would be nice to be able to categorize a lack of an input, but it makes sense. Really appreciate both your help, I should have been more diligent in my debugging. I love the outputs this code can make, very appreciative to Openai for including how to do this in their recent blog post. The different classifications I used were not super similar, and I was only concerned with how well the 5’s clumped together - you can see they did a pretty good job from this picture!
3 Likes