bhaskatripathi / pdfGPT

PDF GPT allows you to chat with the contents of your PDF file by using GPT capabilities. The most effective open source solution to turn your pdf files in a chatbot!
https://huggingface.co/spaces/bhaskartripathi/pdfGPT_Turbo
MIT License
6.91k stars 833 forks source link

text-DaVinci-003 vs gpt-4 #69

Closed ollieanwyll closed 1 year ago

ollieanwyll commented 1 year ago

Hey, awesome project.

just curious about the use of text-DaVinci-003 over gpt-4. would you still recommend using text-DaVinci-003 over gpt-4 if it’s available to the user? If so, would you be so kind as to explain why text-DaVinci-003 is a better choice? Thanks

bhaskatripathi commented 1 year ago

Hey @ollieanwyll , This was developed in Feb,23 when GPT4 wasn't launched. For some reason, I am still in the waitinglist of GPT4 API so I could not test the GPT4 model and compare its results to text-DaVinci-003. However, I am sure that GPT4 will give improved output and has longer context of 32K tokens compared to the 4096 tokens on text-DaVinci-003. Some contributor can submit a PR for this change.

Yiiipu commented 1 year ago

I tried 3.5 and 4, they work great. They feel more alive when giving responses. 3.5 is also faster and cheaper. I also manage to give them conversation memory by using 3.5 or 4. You can check out my repo: https://github.com/Yiiipu/SimplePDF

bhaskatripathi commented 1 year ago

@Yiiipu - This is great! Also, I have added support for all Turbo models now including GPT4: https://huggingface.co/spaces/bhaskartripathi/pdfGPT_Turbo