Open mikes-bowden opened 1 year ago
@thiswinex Would you mind helping me answer a question about these two models (gpt-3.5-turbo
& gpt-3.5-turbo-0301
)?
I chose 0301 version just because this one is newer. 🤪 Maybe it's more advanced I think.
As for choosing model, I will take a deeper look into these models and then have a discussion with you, OK?
No worries at all, and very much appreciated.
@mikes-bowden You can get more infomation about these two models from this link: https://platform.openai.com/docs/models/gpt-3-5
@mikes-bowden You can get more infomation about these two models from this link: https://platform.openai.com/docs/models/gpt-3-5
Not what I was looking for. I'd like to know why that one selected was selected. The information on that page is lackluster at best.
@thiswinex thanks for your information.
@mikes-bowden I did not analyze carefully which model to use before, but after reading the documentation above, I believe that using gpt-3.5-turbo
is a more reasonable choice.
gpt-3.5-turbo-0301
is just a snapshot ofgpt-3.5-turbo
and it won't be updated.
Fair enough. Thanks!
Any plans to support the others?
I think it's possible to support other models. I'd like to test it in v1.3.0, and if nothing goes wrong it will be a new config key for you to select model.
AWESOME!
Hey @mikes-bowden, it's been a long time since our last conversation. I'd like to update some info about model selection:
We have released a lot of interesting new features. I hope you like them.
Awesome!
I'll get it updated today and test it out. Thanks!
First of all, fantastic work. This thing is by far the best I've used. Not sure how you were able to push specific tasks to the background and still communicate with it, but I love it!
I did want to ask about the models. I noticed you were using gpt-3.5-turbo-0301. I'm not familiar with it. Can you explain why that one and not the normal 3.5 turbo?
Also, I tried swapping to text-davinci-003 and 002, but I got a URL error. I assume it just means the wrong syntax is being used to call the API. Is there any way to set it up to utilize older models? I've found that 002 works much better with coding than the newer ones.