Jon Deaton

Research Scientist at @GoogleDeepMind

[Jon Deaton @deaton_jon]

@littmath You should read Against Method by Paul Feyerabend

[Jon Deaton @deaton_jon]

@yoavgo its called "confirmation bias"

[Jon Deaton @deaton_jon]

@tszzl @mrkocnnll @longstosee gerardian

[Jon Deaton @deaton_jon]

@luusssso im proud of this one

[Jon Deaton @deaton_jon]

@kvfrans pallas is the way to get what you want when xla doesn't do it

[Jon Deaton @deaton_jon]

@francoisfleuret vae is an adversarial training dynamic are you expecting different curves

[Jon Deaton @deaton_jon]

@peterrhague another lagrange multiplier for the new constraint

[Jon Deaton @deaton_jon]

@snowclipsed take a look at alpha fold 3

[Jon Deaton @deaton_jon]

@francoisfleuret I thought this was going to be a big problem especially in your sigma gpt but I found it doesn't matter

[Jon Deaton @deaton_jon]

@owl_posting discworld probably fits

[Jon Deaton @deaton_jon]

@liujc1998 can we call them eigentokens

[Jon Deaton @deaton_jon]

@francoisfleuret more specifically clamp * tanh(x/clamp)

[Jon Deaton @deaton_jon]

@francoisfleuret tanh

[Jon Deaton @deaton_jon]

@fermatslibrary best pythag triple is (1, i, 0)

[Jon Deaton @deaton_jon]

@ebetica I did that on my own home directory once

[Jon Deaton @deaton_jon]

@pdhsu Out of mana

[Jon Deaton @deaton_jon]

@francoisfleuret been there

[Jon Deaton @deaton_jon]

@thisismadani thats a neat trick you did to preserve the token positions of the span while doing fill-in-the middle. seems effective

[Jon Deaton @deaton_jon]

@sea_snell have you heard of mup

[Jon Deaton @deaton_jon]

@flynnslick The yin and yang of Gerry Lopez