Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
Captain Foo
May 11, 2004

we vibin'
we slidin'
we breathin'
we dyin'

Sagebrush posted:

GearVR! Try it on your face!

*makes jerking off motion* Eventually!

Adbot
ADBOT LOVES YOU

Carthag Tuek
Oct 15, 2005

Tider skal komme,
tider skal henrulle,
slægt skal følge slægters gang



Captain Foo posted:

*makes jerking off motion* Eventually!

*claps yosposly*

SmokaDustbowl
Feb 12, 2001

by vyelkin
Fun Shoe

Snapchat A Titty posted:

*claps yosposly*

flakeloaf
Feb 26, 2003

Still better than android clock

Captain Foo posted:

*makes jerking off motion* Eventually!

raej
Sep 25, 2003

"Being drunk is the worst feeling of all. Except for all those other feelings."
https://twitter.com/Markov_Carmack/status/782856521088573440

Markov Carmack getting spookier!

PS, I added his .plan files from 1996-2010

my girlfriend is Legos
Apr 24, 2013
https://twitter.com/WYR_bot/status/783003788307202053

if only.........

WrenP-Complete
Jul 27, 2012


I love this one.

Sagebrush
Feb 26, 2012

i'll give up the internet for life. when i can fly, i'll be so rich i'll hire someone else to do the internet for me

Trig Discipline
Jun 3, 2008

Please leave the room if you think this might offend you.
Grimey Drawer
"convert today's top ten trending gifs to bluray and have them in the player before I get in tomorrow, charles "

Sagebrush
Feb 26, 2012

or just, like, download them

no one says you have to give up computers

WrenP-Complete
Jul 27, 2012

today my bot said "i was just raped"
and i made an empathy noise automatically.

theflyingexecutive
Apr 22, 2007

WrenP-Complete posted:

today my bot said "i was just raped"
and i made an empathy noise automatically.

the gently caress seed file are you working from?

WrenP-Complete
Jul 27, 2012

theflyingexecutive posted:

the gently caress seed file are you working from?

i scraped posts of a libertarian who used to post here.

Trig Discipline
Jun 3, 2008

Please leave the room if you think this might offend you.
Grimey Drawer

Sagebrush posted:

or just, like, download them

no one says you have to give up computers

i suppose

although honestly it's gotten to the point where if the internet goes out i just look at my computer and go "what the gently caress does this thing even do"

Crusader
Apr 11, 2002

just noting that it's definitely worth re-training your RNN if you're using a dataset you can refine and expand; it's really weird seeing what appear to be minor incremental improvements become relatively large increases in certainty/"understanding" over time. originally sampling runs with the word-rnn temperature parameter set to something low like "0.5" would yield "would you rather yes or no" over and over, but now there's some nice variety - still safe/somewhat boring output, but the network is definitely getting a better handle on what sort of word sequences i consider good.

Crusader fucked around with this message at 18:05 on Oct 4, 2016

NoneMoreNegative
Jul 20, 2000
GOTH FASCISTIC
PAIN
MASTER




shit wizard dad

theflyingexecutive posted:

the gently caress seed

modés

atomicthumbs
Dec 26, 2010


We're in the business of extending man's senses.

Crusader posted:

just noting that it's definitely worth re-training your RNN if you're using a dataset you can refine and expand; it's really weird seeing what appear to be minor incremental improvements become relatively large increases in certainty/"understanding" over time. originally sampling runs with the word-rnn temperature parameter set to something low like "0.5" would yield "would you rather yes or no" over and over, but now there's some nice variety - still safe/somewhat boring output, but the network is definitely getting a better handle on what sort of word sequences i consider good.

are you training over the original weights? how is it getting feedback on what you consider good?

Crusader
Apr 11, 2002

atomicthumbs posted:

are you training over the original weights? how is it getting feedback on what you consider good?

yep - adding the twitter posts to the training corpus (since i strip out anything that's a match for a corpus line, which seems unnecessary with temperature=1) + re-crawling net sources for new material

edit: originally i ran through a bunch of parameters through trial and error (net size, layers, optimization method, dropout rate) that i haven't revisited in a while since just adding more to the corpus has kept improving both qualitative output and the quantitative error rate, but will probably go back and see if there's any gains to be made by tweaking those again

Crusader fucked around with this message at 20:28 on Oct 4, 2016

atomicthumbs
Dec 26, 2010


We're in the business of extending man's senses.

Crusader posted:

yep - adding the twitter posts to the training corpus (since i strip out anything that's a match for a corpus line, which seems unnecessary with temperature=1) + re-crawling net sources for new material

edit: originally i ran through a bunch of parameters through trial and error (net size, layers, optimization method, dropout rate) that i haven't revisited in a while since just adding more to the corpus has kept improving both qualitative output and the quantitative error rate, but will probably go back and see if there's any gains to be made by tweaking those again

oh, so you're adding its own (curated) input back into the training too? that's cool as hell

unpacked robinhood
Feb 18, 2013

by Fluffdaddy
those arent markov boots but pretty cool imo

they trained a thing to generate animated frames, either from scratch



or from a static image



e:they have a cronenberg training set

unpacked robinhood fucked around with this message at 11:05 on Oct 5, 2016

WrenP-Complete
Jul 27, 2012

Crusader posted:

yep - adding the twitter posts to the training corpus (since i strip out anything that's a match for a corpus line, which seems unnecessary with temperature=1) + re-crawling net sources for new material

edit: originally i ran through a bunch of parameters through trial and error (net size, layers, optimization method, dropout rate) that i haven't revisited in a while since just adding more to the corpus has kept improving both qualitative output and the quantitative error rate, but will probably go back and see if there's any gains to be made by tweaking those again

Crusader, we are on the exact same page with what we are doing. I read your post and then checked to see if I was you. :hfive:

Captain Foo
May 11, 2004

we vibin'
we slidin'
we breathin'
we dyin'

unpacked robinhood posted:

those arent markov boots but pretty cool imo

they trained a thing to generate animated frames, either from scratch



or from a static image



e:they have a cronenberg training set



this is pretty sweet

unpacked robinhood
Feb 18, 2013

by Fluffdaddy
https://twitter.com/WYR_bot/status/783728612285509632

https://twitter.com/WYR_bot/status/783819224582090752

https://twitter.com/WYR_bot/status/783773918503047168

i love u wyrbot

Synthbuttrange
May 6, 2007

Wyrbot is on a gold streak.

https://twitter.com/WYR_bot/status/783366169503891462

https://twitter.com/WYR_bot/status/783411475264200706

https://twitter.com/WYR_bot/status/783139643722194944

Su-Su-Sudoko
Oct 25, 2007

what stands in the way becomes the way

those are all ridiculously good

Synthbuttrange
May 6, 2007

https://twitter.com/WYR_bot/status/784226775077363716

MIKE TYSON'S CLOWN UNIVERSE

my girlfriend is Legos
Apr 24, 2013

moooooooods

Trig Discipline
Jun 3, 2008

Please leave the room if you think this might offend you.
Grimey Drawer
rename yospos to mike tyson's clown universe imo

Sagebrush
Feb 26, 2012

Trig Discipline posted:

rename yospos to mike tyson's clown universe imo

Synthbuttrange
May 6, 2007

Westie
May 30, 2013



Baboon Simulator
wyr_bot finally is sentient

WrenP-Complete
Jul 27, 2012

Mike Tyson clown universe!!!

WrenP-Complete
Jul 27, 2012

I'm naming this R work environment Mike Tyson's clown universe.

(The last one was named meatspace)

Moist von Lipwig
Oct 28, 2006

by FactsAreUseless
Tortured By Flan
mods please rename yospos "mike tyson's clown universe"

WrenP-Complete
Jul 27, 2012

Yospos is Mike Tyson's clown universe!!! Mods please rename!

power botton
Nov 2, 2011

the more i think about it, the more im convinced that this forum is like if mike tyson ran the universe, but everyone in it was clowns

Carthag Tuek
Oct 15, 2005

Tider skal komme,
tider skal henrulle,
slægt skal følge slægters gang



also a sound "mods" from here

well why not
Feb 10, 2009




modsmodsmods

flakeloaf
Feb 26, 2003

Still better than android clock

WrenP-Complete posted:

Yospos is Mike Tyson's clown universe

Adbot
ADBOT LOVES YOU

infernal machines
Oct 11, 2012

we monitor many frequencies. we listen always. came a voice, out of the babel of tongues, speaking to us. it played us a mighty dub.

there's some folk in japan that could use your help...

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply