foilfoilfoil/FireCheese · Datasets at Hugging Face

text stringlengths 0 2.01k
NotYourMom#5178:unity
cell_destroyer#8513:What data rate is this?
Jiray#3048:UUGGGGGGHHHH
NotYourMom#5178:"theres like this Very successful game using unity, looks good too (disregarding texture/model popins) :Clueless:"
NotYourMom#5178:"oh also
https://www.youtube.com/watch?v=x65PPDn9ndM"
NotYourMom#5178:nvm its a reupload
SaskAudioGuy#2444::Clueless:
Olrak#9007:Genshit :DLSSoyjak:
NotYourMom#5178::LeoKek:
nadd#3217:"Memory standards are hilariously slow to come out, when AMD and Intel start going ham on 2000mm2+ silicon monstrosities in a few years, how the hell is it gonna get fed?"
Jiray#3048:Hearthstone?
Jiray#3048::kek:
NotYourMom#5178:wait heartstone uses unity?
NotYourMom#5178:TIL
Jiray#3048:Wanna know some more games that use Unity?
MCH#5588:Based
JamesCJ#2023:"Here’s some more games :KEKW:
The Forest
Sons of The Forest
Hollow Knight
Cuphead
Escape from Tarkov
Ori and the Blind Forest
Fall Guys
Untitled goose game
Rust
Outer Wilds"
Olrak#9007:More caches
Olrak#9007::PepePoint:
Olrak#9007:"Wait, fall guys?"
Olrak#9007:That genuinely looks like a UE game
NotYourMom#5178:https://tenor.com/view/reaction-game-gaming-fire-reactions-gif-25416214
NotYourMom#5178:huh the 3050m clocks the mem at like 1.6Gbps under low power
Jiray#3048:"Beat Sabre
Cities: Skyline
Subnautica
Pillars of Eternity"
Jiray#3048:"Yes, Pillars of Eternity uses Unity"
nadd#3217:"Ah yes the cache processing unit, also called CPU"
Olrak#9007:The cerebras WSE is also a giant SRAM pile
MCH#5588:Can we talk
MCH#5588:About how shit google translate is?
Jiray#3048::kek:
MCH#5588:an easy language it’s good enough
NotYourMom#5178:"its all sram?!
always has been 🔫"
MCH#5588:It butchers hard languages so bad though
JamesCJ#2023:"Cult of the Lamb
Totally accurate battle simulator
Alto’s adventure
My friend pedro
I’m not listing anymore :kek:"
Jiray#3048:It's so shit at German.
Olrak#9007:My problem with google translate is that it fucks the original formatting of the shit it translates
MCH#5588:Yea
MCH#5588:GPT-4 is better
Olrak#9007:Especially tieba
MCH#5588:Like literally
nadd#3217:True. I used babblefish for years and GTL never crossed my mind
MCH#5588:I can ask bing chat to translate the webpage
NotYourMom#5178:for short sentences its fine/usable
MCH#5588:And it produces a significantly better TL
NotYourMom#5178:but for lots of words :/
NotYourMom#5178:or i2t yeaaa...
Jiray#3048:SteamVR is also Unity. :HyperKek:
Jiray#3048:https://steamdb.info/tech/Engine/Unity/
Olrak#9007:Wallpaper engine best game
MCH#5588:"whisper model, a video transcription model, is better at translating audio Japanese => English text than Google tl is at text to text"
MCH#5588:So sad
MCH#5588:google has really given us their C effort
MCH#5588:because no competition
Gifted Gummy Bee::SufferingInside: inference costs
NotYourMom#5178:TPU :Clueless:
Olrak#9007:Just buy H100s bro
Gifted Gummy Bee::Clueless: :Clueless:
Gifted Gummy Bee:"Also, google’s AI division is also kinda shit rn for some reason"
Gifted Gummy Bee:People do better work on colab over a nice weekend than the entire fucking AI division
MCH#5588:Yea they were nickel dimming us
Gifted Gummy Bee:I heard google has a cult or something
Gifted Gummy Bee:I do wonder
Gifted Gummy Bee:Why didn’t google just throw 3t tokens at the problem
MCH#5588:even bard
MCH#5588:Is optimized for inference performance
Gifted Gummy Bee::SufferingInside: llama literally proved that you just need to train on more tokens to get performance
JamesCJ#2023:https://twitter.com/tomwarren/status/1654140648512016385
MCH#5588:Time
Gifted Gummy Bee:Not really
MCH#5588:GPT-4 took 45 days
Gifted Gummy Bee:Bungiee cord a bunch of TPU
MCH#5588:On 10K A100s
Gifted Gummy Bee:You think google has less compute than fucking OAI?
Jiray#3048:"TBF, at Googles scale, inference perf is very important."
Gifted Gummy Bee:Custom silicon + custom software = actually some decent optimisations
MCH#5588:"Nope, but they aren’t training a chatGPT competitor in 7 days"
Gifted Gummy Bee:This is literally the BOM vs perf argument
Gifted Gummy Bee:This is LITERALLY
MCH#5588:"Yes, but value matters too"
Gifted Gummy Bee:7900xtx vs 4090
MCH#5588:Google wanted to keep their sweet high margin SEO optimized garbage search

NotYourMom#5178:unity

cell_destroyer#8513:What data rate is this?

Jiray#3048:UUGGGGGGHHHH

NotYourMom#5178:"theres like this *Very* successful game using unity, looks good too (disregarding texture/model popins) :Clueless:"

NotYourMom#5178:"oh also

https://www.youtube.com/watch?v=x65PPDn9ndM"

NotYourMom#5178:nvm its a reupload

SaskAudioGuy#2444::Clueless:

Olrak#9007:Genshit :DLSSoyjak:

NotYourMom#5178::LeoKek:

nadd#3217:"Memory standards are hilariously slow to come out, when AMD and Intel start going ham on 2000mm2+ silicon monstrosities in a few years, how the hell is it gonna get fed?"

Jiray#3048:Hearthstone?

Jiray#3048::kek:

NotYourMom#5178:wait heartstone uses unity?

NotYourMom#5178:TIL

Jiray#3048:Wanna know some more games that use Unity?

MCH#5588:Based

JamesCJ#2023:"Here’s some more games :KEKW:

The Forest

Sons of The Forest

Hollow Knight

Cuphead

Escape from Tarkov

Ori and the Blind Forest

Fall Guys

Untitled goose game

Rust

Outer Wilds"

Olrak#9007:More caches

Olrak#9007::PepePoint:

Olrak#9007:"Wait, fall guys?"

Olrak#9007:That genuinely looks like a UE game

NotYourMom#5178:https://tenor.com/view/reaction-game-gaming-fire-reactions-gif-25416214

NotYourMom#5178:huh the 3050m clocks the mem at like 1.6Gbps under low power

Jiray#3048:"Beat Sabre

Cities: Skyline

Subnautica

Pillars of Eternity"

Jiray#3048:"Yes, Pillars of Eternity uses Unity"

nadd#3217:"Ah yes the cache processing unit, also called CPU"

Olrak#9007:The cerebras WSE is also a giant SRAM pile

MCH#5588:Can we talk

MCH#5588:About how shit google translate is?

Jiray#3048::kek:

MCH#5588:an easy language it’s good enough

NotYourMom#5178:"its all sram?!

always has been 🔫"

MCH#5588:It butchers hard languages so bad though

JamesCJ#2023:"Cult of the Lamb

Totally accurate battle simulator

Alto’s adventure

My friend pedro

I’m not listing anymore :kek:"

Jiray#3048:It's so shit at German.

Olrak#9007:My problem with google translate is that it fucks the original formatting of the shit it translates

MCH#5588:Yea

MCH#5588:GPT-4 is better

Olrak#9007:Especially tieba

MCH#5588:Like literally

nadd#3217:True. I used babblefish for years and GTL never crossed my mind

MCH#5588:I can ask bing chat to translate the webpage

NotYourMom#5178:for short sentences its fine/usable

MCH#5588:And it produces a significantly better TL

NotYourMom#5178:but for lots of words :/

NotYourMom#5178:or i2t yeaaa...

Jiray#3048:SteamVR is also Unity. :HyperKek:

Jiray#3048:https://steamdb.info/tech/Engine/Unity/

Olrak#9007:Wallpaper engine best game

MCH#5588:"whisper model, a video transcription model, is better at translating audio Japanese => English text than Google tl is at text to text"

MCH#5588:So sad

MCH#5588:google has really given us their C effort

MCH#5588:because no competition

Gifted Gummy Bee::SufferingInside: inference costs

NotYourMom#5178:TPU :Clueless:

Olrak#9007:Just buy H100s bro

Gifted Gummy Bee::Clueless: :Clueless:

Gifted Gummy Bee:"Also, google’s AI division is also kinda shit rn for some reason"

Gifted Gummy Bee:People do better work on colab over a nice weekend than the entire fucking AI division

MCH#5588:Yea they were nickel dimming us

Gifted Gummy Bee:I heard google has a cult or something

Gifted Gummy Bee:I do wonder

Gifted Gummy Bee:Why didn’t google just throw 3t tokens at the problem

MCH#5588:even bard

MCH#5588:Is optimized for inference performance

Gifted Gummy Bee::SufferingInside: llama literally proved that you just need to train on more tokens to get performance

JamesCJ#2023:https://twitter.com/tomwarren/status/1654140648512016385

MCH#5588:Time

Gifted Gummy Bee:Not really

MCH#5588:GPT-4 took 45 days

Gifted Gummy Bee:Bungiee cord a bunch of TPU

MCH#5588:On 10K A100s

Gifted Gummy Bee:You think google has less compute than fucking OAI?

Jiray#3048:"TBF, at Googles scale, inference perf is very important."

Gifted Gummy Bee:Custom silicon + custom software = actually some decent optimisations

MCH#5588:"Nope, but they aren’t training a chatGPT competitor in 7 days"

Gifted Gummy Bee:This is literally the BOM vs perf argument

Gifted Gummy Bee:This is LITERALLY

MCH#5588:"Yes, but value matters too"

Gifted Gummy Bee:7900xtx vs 4090

MCH#5588:Google wanted to keep their sweet high margin SEO optimized garbage search