Proposal discussion

[Sarthak5598]

Using OpenAI API can cost money ,should I add this in my proposal ?
@DonnieBLT @AtmegaBuzz

[AtmegaBuzz]

Ig it's ok to use it, it's not that costly if we compare it to training our LLM and deploying it

[Sarthak5598]

okay!! thank you

[HanilJain]

@AtmegaBuzz well it depends what kind of dataset we are training for.
In my proposal I was suggesting to train our model on lighter datasets and improvising it in future.
is it okay ??

[AtmegaBuzz]

training the AI model and deploying it takes more money than simply paying for monthly API subscription

[AtmegaBuzz]

if you want, we can still look into it. you can take opensource llm models and train it for our usecase.

[HanilJain]

I have added it in my proposal with a flex ability to API model integration I'll show you once it is completed

[Sarthak5598]

Any suggestions ?

[DonnieBLT]

Looks good!

[Sarthak5598]

Should add more projects to it ?

[DonnieBLT]

This looks good. I think this is enough for one gsoc

[Sarthak5598]

Okay , thank you

[HanilJain]

@DonnieBLT can we include open PR in proposal as I found this as important feature to be implemented
and also including in it google cloud storage #1737 if we can

[Sarthak5598]

how does this look ?

[Sarthak5598]

So, just after the slack bot or should I add the bacon crypto at the start

[DonnieBLT]

The start

[Sarthak5598]

Sir, Can I just use a cryptocurrency like Solana and build our own crypto over it ?

[DonnieBLT]

Yes, we decided on Peercoin because it is a currently maintained PoS coin.

[Sarthak5598]

Do you have any source to learn more about it because I went through a video on how to make crypto through solana but I cant find any proper source for using peercoin

[Sarthak5598]

@AtmegaBuzz @DonnieBLT I was thinking of making a NER dataset(NLP) to train a NER model for automating labels for Bug report , Do you think its worth it because it may take a lot of time to perfect it and we have a simple alternative to it (just giving user option to add label) what do you think

[AtmegaBuzz]

Automating labels based on user inputs sounds interesting to me, yeah training our own model would take time and then running it, can you look into chat gpt apis, i think they added new options to create your own gpt's by pre training it on the chat gpt platform. Lets test it on local first then if it works move it to prod or look for options like training our own model.

I'm open to training our own model if it doesn't take much time and cost than using cgpt for 1 year.
What do you think @DonnieBLT

[Sarthak5598]

Thats a great idea I think using chatgpt will make process much faster as we dont have to use tenserflow to train a ml model. I dont think it would take much time to train chatgpt

[Sarthak5598]

We have already thought about using chatgpt's new feature for slack bot we can use it here too

[ATHARVA-GAWAS]

Inputs please..