It looks promising!!! I’ll do that tonite ty sir
fab_space
joined 1 year ago
I ❤️ this answer.
Tried most of them absolutely useless.
I want to share my test with u for reviewing, and hopefully, integration.
how it sounds?
There are 4 ways to achieve the mission as far I understood:
- privacy way: use open source stuff (LLmstack, DocsGPT) yourself at home or lab
- build the same from scratch
- use OpenAI or others AI APIs
- use find in excel since the most is to find a single information in the ocean of data
😂
how to contribute?
I’m newbie and i have no GPU at home but i dig into the topic at many layer.. experimenting and so on with also paid or free APIs and all home lab tools the dream rush is giving out there and it’s quite engaging..
after several iterations (2 months full free time spent on that) i’m back to create proper datasets.
Datasets are the most engaging where any model then can fit flawlessly later on. Data interpretation is the king.
Please make me wrong at this, yet another chance to learn in the topic.
In my experiment since i deliver blacklists i’m playing with such flow:
- getting data in an ethical way
- processing data
- train data
- rank domains
the application part can be:
- browser extension
- blacklists improvements
- domain rankings
the required in this context is a proper made dataset more than the best GPU powered model 💻
Sorry if this is quite off 🦙
PS: i’m not dev and i use gpt to code some, the best iteration and lesson learned:
k now please make this function parallel, up to 32 concurrent
and i learned concurrent processing of python 🐍🙏
i second your high times find me somewhere score them all and save the world
❤️
Hi monsters, can u please gimme some pathway to reduce the expected 23 lifes to learn more on fine tuning in the LLM for code context?
I already tested pretrained models at home and started to create datasets and crazy experiments my self but of course, some guidance can help me to focus and learn important things i’m of course missing.
One of the goals is to understand how fine tune more than training works for code oriented models.
I also checked code oriented datasets on huggingface and so on.
Any advice is really appreciated.
dnscontrol it via git actions and u will never go to gui :)
no it’s dnscontrol the dns way. sir
view more: next ›
ty for this additional suggestion!!
I started years ago with emby and youtube-dl crons but i’m happy to test all recently updated alternatives to find the one which fit best for my needs.