Google has been changing what GPUs are available on #Colab literally every other day.

if you're trying to guess parameter size of GPT2-chatbot based on the speed of the replies (tok/sec), know that that varies based on a lot of things besides the size of the model

such as which GPUs they allocated to it, how many GPUs, if they INTENTIONALLY slow it down to manage…

hot take: chat is actually a great ui (for a lot of cases)

if you type slow just say that

There are around 30 trillion cells in the human body

Credit me!

Made it to the top of a subreddit on AI with 2M members

Luckily they labeled it properly

12000 applicants for a $20/hr SWE internship

Fellow CS majors, we missed the boat.

1) The model seems quite capable:

twitter.com/itsandrewgao/s…

if anyone didn't like their rabbit and wants to sell, i'll buy it for $120

for those curious about how i learned to tweet good, in high school i did a marketing internship for a few months for a summer program called LaunchX (idk if any of y'all did it?) and lowkey learned a lot about social media which helps to this day, i also managed ad campaigns for…

hey guys, i wanted to come on here and apologize for the misinformation. it was an accident and i regret the error.

here's the correction:
GPT2 will have 100 quintillion parameters

ten tweets over 10k today, a personal record

silence, nontechnical

or AI at Meta
llama 3 8b just blew away mistral

although not fully open source

gpt4 but truthmaxxing and not rlhf'd an insane amount

the great hiatus...