12:16:30

has anyone tried training VQGAN with a larger transformer? i know the compute increases quadratically but wouldn't it increase coherence to have more parameters

🗨️ 2 ♺ 0 🤍 5


12:21:12

although deepmind's chinchilla paper suggests that the transformer might not even be compute optimal unless it has the right amount of data anyway?

🗨️ 1 ♺ 0 🤍 4


12:26:30

Women are like this. Men are like that

🗨️ 30 ♺ 14 🤍 274


16:38:00

What happened to political assassinations? I feel like those really fell off

🗨️ 17 ♺ 4 🤍 70


17:12:06

I am going to eat Twitter

🗨️ 3 ♺ 0 🤍 24


17:47:21

Students are unforgivable

🗨️ 0 ♺ 0 🤍 12


18:03:07

If only everyone would see reason, accept enlightened truths and overthrow the dysfunctional state of things

🗨️ 5 ♺ 1 🤍 11


18:27:13

Thinking about becoming one of those clout-chasing engagement bait accounts. Which discourse should I develop inflammatory bad-faith takes about?

🗨️ 10 ♺ 1 🤍 24


18:29:12

I'm open to other suggestions as well, these just seem like the targets people usually go for

🗨️ 5 ♺ 0 🤍 8


18:38:34

import this https://t.co/zcGFBRk4j1

image from twitter

🗨️ 1 ♺ 2 🤍 35


20:07:20

the world if https://t.co/1jubipm9BN

image from twitter

🗨️ 3 ♺ 2 🤍 41


21:09:21

I'm cool with spiders

🗨️ 9 ♺ 2 🤍 36