Deberta does work quite well for seq cls, but Mistral should generally outperform it. However in this case the target modules do not cover enough (it should be all linear layers in the main backbone) so there's *much* fewer trainable params
AI
-
Best practices for converting API documentation to OpenAI specifications
By
–
What’s the best way to convert any API docs to an OpenAI spec
-
Custom Instructions Cause AI Agent Web Browsing Regression
By
–
Yeah, I had to add a custom instruction and tell it to never browse the internet. Big regression
-
Dell Tech Forum explores generative AI phenomenon
By
–
Venez en savoir plus sur le phénomène de l'#intelligenceartificielle générative en participant au #DellTechForum (au carroussel du Louvre ou en Livestream)
— Nicolas Babin (@Nicochan33) 10 novembre 2023
Plus d'informations et inscriptions :https://t.co/f59WzMw0R0#sponsorisé #innovation @DellTechFrance @3itcom @Shi4Tech pic.twitter.com/3TGTAQPENrVenez en savoir plus sur le phénomène de l'#intelligenceartificielle générative en participant au #DellTechForum (au carroussel du Louvre ou en Livestream)
Plus d'informations et inscriptions : https://
dell.com/dt/forum-fr?dg
c=sm&cid=dtf23_sm_EMEA_FR&lid=sm_ChoYou_Influenceur&RefID=sm_EMEA_FR_sm_ChoYou_Influenceur
…
#sponsorisé #innovation @DellTechFrance @3itcom @Shi4Tech -

Selected as Speaker at OpenSSF Day Japan 2023
By
–
I got selected as a in-person speaker at @openssf Day Japan 2023 VISA Tickets Hotels Also going to attend Open Source Summit Japan this year. I will be in Japan from Nov 29 to Dec 7. I was working for this day from last two years and it is finally happening.
-
GPT-4 Router: Smart Tool Selection for AI Tasks
By
–
So there is some little helper, some part of the GPT4 interface that decides what underlying version of GPT4 is going to be used. If it should browse the web, use code-interpreter or DALLE-3 something that dictates what tools best used for the job, and currently that alignment
-
Open-Access Deep Learning Course: Years of Teaching Refinement
By
–
The full story of the open-access Deep Learning course that Alfredo Canziani and I have been teaching together and refining over the last several years.
-
Human Translation Services as Exclusive Trusted Market Moat
By
–
Adapt the tools, if you are providing human powered translation / interpretation you need to move up the ladder, it will be seen as a very exclusive service to use, and potentially have moat in that it is a trusted human that does the work. I think the world will be in need of
-

AI Existential Risk Fears Based on Flawed Ideas
By
–
The fears of AI-fueled existential risks are based on flawed ideas.
A WSJ piece by Princeton's @random_walker and @sayashk . -
ChatGPT Playground vs UI: Dispatcher Differences Explained
By
–
To do this, only option now is the playground, since we can no longer switch in the ChatGPT UI. There are differences between the playground and ChatGPT. Mainly the dispatcher.
