Debugging Bad AI Codes 🤣
HTML-код
- Опубликовано: 29 сен 2024
- URL of my website:
bracercom.neoc...
Please support my Patreon:
/ bracerjack
Paypal Donation:
bracercom@hotmail.com
Bitcoin Address Donation:
1698RW9rSJUXeSSbyNaQY7aNNbxQkLtjs5
===============================================================
Support me by reading my books !
Humpty Dumpty God
Google Play eBook:
play.google.co...
Kindle eBook:
www.amazon.com...
Paperback:
www.amazon.com...
===============================================================
===============================================================
Book 1 :: God Guardian: The Life Review Simulation
Google Play eBook:
play.google.co...
Kindle eBook:
www.amazon.com...
Paperback:
www.amazon.com...
===============================================================
===============================================================
Book 2 :: God Guardian: Incarnation Troll
Google Play eBook:
play.google.co...
Kindle eBook:
www.amazon.com...
Paperback:
www.amazon.com...
===============================================================
===============================================================
Book 3 :: God Guardian: Perpetual Slaves
Google Play eBook:
play.google.co...
Kindle eBook:
www.amazon.com...
Paperback:
www.amazon.com...
===============================================================
After Hours, Dark Forest, Pitter Patter, S’mores, Star Bright, Stranded by Purrple Cat | purrplecat.com
Music promoted by www.free-stock...
Creative Commons / Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0)
creativecommon...
Thanks for this excellent video, Jack. Good to see some others, like myself, think AI still need some work and time before it can start replacing coders.
Yeah I agree.
There are a number of videos talking about open source projects that uses A.I to create an app from start to finish.
It all sound very fictitious to me, at least for where the current tech is right now, there is the "trying to increase the share market" hype, and there is...reality.
The tech is amazing, but...it all just seems so...hyped ;p
I even tried crewA.I and as of 9 April 2024, it couldn't even initiate proper search most of the time, it is just...not there yet.
I would love to play with it more and record myself doing it as well.
What was context windows size of used LLM?
For this example and ONLY for this video, I am using Nous-Hermes-2-Mistral-7B-DPO since I am also teaching how to get started, in real life you would use the best model for the job, but circling back to your question, this is the data for Nous-Hermes-2-Mistral-7B-DPO:
Context Length: 32768
Model Max Length: 32768
Vocabulary Size: 32002
These are theoretical limits, GPT4All which is the one I used for the early half of the test in this video have its own context length limit, going to the settings and changing its context length didn't help for GPT4All, that's why for the later half of the video, I switched to Oobabooga where I can increase the context length for it to be able to comprehend an entire static site and it works.