Sabitlenmiş Tweet
Vardhan
2.9K posts

Vardhan
@baaatli
seeking wisdom. Tech, Startups, Marketing, Psychology and Memes.
India Katılım Şubat 2019
477 Takip Edilen636 Takipçiler

@baaatli @Codie_Sanchez literally nothing. it's all empty. sometimes some very courageous guy opens a new store, eventually realizing that it was a mistake
English

LA is a mess of a dying city.
Homeless man trailed me 5 blocks screaming how he’s going to assault me.
Comedy club had a woman behind me threaten to hit me bc I laughed at a joke (albeit a very naughty one).
More stores boarded than open.
More zombies than functioning humans.
What a tragedy. Also fire EVERY single member of this government.
English

@Codie_Sanchez I suppose this is not just LA.
Even my much smaller hometown here in Germany dies out. The city center is very much dead.
More and more stores close, not even to mention the shopping malls.
And I can't blame the people, I don't go there myself as often as I used to.
English
Vardhan retweetledi

@shyamfoxface Ah, informational genre should be tough then.
Would prolly make sense for philosophy because maybe the human reviews could garnish great creativity to the model.
English

@baaatli Depends on the reviews and the use case.
I dont think reviews of a chem-engg book will have data on chemical reactions
English

Got a couple of good DMs.
Based on this, I understand an LLM can only be as smart as the RLHF dataset used to optimise it.
In order to make it learn from a book and make it useful, we need to have a {prompt} {completion} dataset from that book itself.
Sounds like a Chicken & Egg problem again😂
Shyam Arvadia@shyamfoxface
Dear AI coders, What’s the easiest way I can fine-tune a Llama 3 on a personal dataset of books? I would prefer a good GUI based approach rather than pure code.
English

@shyamfoxface @X Absolutely, Rabbit is a collector's item for sure. Humane pin subscription is a deal breaker for me.
English




