Purebasic and chat gpt (version 3.5)

[idle]

Hello idle,
Thank you for your answer.
I'll watch it tomorrow (it's 1:35 am in France and I'm going to bed).
Good day to you.
See you

bon nuit, a bientot

[jak64]

In French

[idle]

In French

Qui, je suis parle un pettie peu Francais

[idle]

I've changed it so the download is now on github
https://github.com/idle-PB/gpt4all_pb

[jak64]

Hello idle,
Thanks for putting the link but I don't know what to do when I click on it, there are lots of lines and I don't know what to download.

[jak64]

Hello idle,
Here is the contents of my directory

Code: Select all

    Répertoire : D:\Purebasic\ChatGPT\gpt4all_pb


Mode                 LastWriteTime         Length Name                                                                 
----                 -------------         ------ ----                                                                 
-a----        04/09/2023     00:42     4061641216 ggml-model-gpt4all-falcon-q4_0.bin                                   
-a----        04/09/2023     00:32       25881295 gpt4all-installer-win64.exe                                          
-a----        04/09/2023     01:41          10050 gpt4all.pbi                                                          
-a----        04/09/2023     23:37          82944 gpt4all_pb.exe                                                       
-a----        04/09/2023     00:34        2145792 libc++.dll                                                           
-a----        04/09/2023     00:34         681984 libgptj-avxonly.dll                                                  
-a----        04/09/2023     00:34         681984 libgptj-default.dll                                                  
-a----        04/09/2023     00:34         646656 libllama-230511-avxonly.dll                                          
-a----        04/09/2023     00:34         646656 libllama-230511-default.dll                                          
-a----        04/09/2023     00:34         740352 libllama-230519-avxonly.dll                                          
-a----        04/09/2023     00:34         740352 libllama-230519-default.dll                                          
-a----        04/09/2023     00:34         755200 libllama-mainline-avxonly.dll                                        
-a----        04/09/2023     00:34         755200 libllama-mainline-default.dll                                        
-a----        04/09/2023     00:34          99328 libllamamodel-230511-avxonly.dll                                     
-a----        04/09/2023     00:34          99328 libllamamodel-230511-default.dll                                     
-a----        04/09/2023     00:34         101888 libllamamodel-230519-avxonly.dll                                     
-a----        04/09/2023     00:34         101888 libllamamodel-230519-default.dll                                     
-a----        04/09/2023     00:34         101888 libllamamodel-mainline-avxonly.dll                                   
-a----        04/09/2023     00:34         101888 libllamamodel-mainline-default.dll                                   
-a----        04/09/2023     00:34         304128 libllmodel.dll                                                       
-a----        04/09/2023     00:34          47256 libllmodel.dll.a                                                     
-a----        04/09/2023     00:34         679936 libmpt-avxonly.dll                                                   
-a----        04/09/2023     00:34         679936 libmpt-default.dll                                                   
-a----        04/09/2023     00:34         207360 libunwind.dll                                                                                                           

[idle]

I just changed the location of the code so its hosted on github.

You still have the old version of the zip, No wonder it isn't working.

https://github.com/idle-PB/gpt4all_pb/a ... s/main.zip

[jak64]

Thanks idle,
I will download tomorrow, it's 2:26 a.m. in France...

See you

[jak64]

Hello idle,
I downloaded with the new link, copied the template to the same directory and launched the program "gpt4all.pbi".

I get the screen below...

What do I do next to ask Chargpt a question?

Thanks to you

[idle]

Goto http://127.0.0.1

[jak64]

Idle,
When I click on the link, I get the web page that says "Sorry, this page cannot be accessed"

[Caronte3D]

Idle,
When I click on the link, I get the web page that says "Sorry, this page cannot be accessed"

First run the "gpt4all.pbi" then go to that localhost, you will see the prompt input on your web browser.

[jak64]

OK,
Thank you Caronte3D but ....

When I ask a question:

He always tells me that he doesn't know the answer, for example:

"Easy MCQ on Napoleon with a question and 4 propositions. Give me the answer too". Moreover, it takes a long time before he answers me that.

With Chatgpt3.5 (directly on the WEB), it's faster and it finds answers almost instantly!!!



Hello to all my friends...

Can't we modify the Purebasic program so that chatgpt 3.5 and not Chatgpt 4 is used?

I recall my need:

I would like to write a small MCQ game program (all in Purebasic) which allows me to ask (only with a Purebasic form) the questions to ChagtGpt and to be able to save the answers of ChatGpt in a text file or a string variable.

This, to me, is not important if Chatgpt 3.5 stopped learning in 2021.

When I discovered Chatgpt 3.5, I thought that we could (like on many other WEB sites) drive it directly with Purebasic (requests and responses) either in a Purebasic variable, or in a text file that I know how to use with Purebasic.

However, it turned out (after discussions with all of you), that this was not possible so simply with ChatGpt.

That's why we had all these (very interesting) exchanges.

I don't know if Google translate will be able to translate my words correctly...

See you

[idle]

Yes the answer speed is slow it would greatly benefit from running on graphics cards but that will take some time.
It's easy to remove the webserver code.
Try different models the bigger ones need 16 gb of ram but it will still be slow

[Caronte3D]

When I ask a question:
He always tells me that he doesn't know the answer, for example...

...Moreover, it takes a long time before he answers me that...

As I was said some posts ago... You shouldn't expect results like ChatGPT either speed nor quality, this works on your cheap computer, not is super-computers with hundreds / thousands of processors in paralel.

If you want to use ChatGPT, you must pay for it (it's very cheap if you not abuse), and use her API. Don't treat to reinvent the wheel.