theminji/v2ray · Datasets at Fast360

input stringlengths 5 4.07k ⌀	output stringlengths 5 4.07k
horsecode wait you can't fine-tune GPT3.5 API models, right?	v2ray nope you cant only base models
v2ray nope you cant only base models	horsecode sigh <Re: v2ray> well, I do have my own datasets... just wondering how much money will that drain
horsecode sigh <Re: v2ray> well, I do have my own datasets... just wondering how much money will that drain	v2ray <Re: horsecode> fine tuning through openai api can easily go through 1k+
v2ray <Re: horsecode> fine tuning through openai api can easily go through 1k+	horsecode <Re: v2ray> 💀
6706full If I have a Rtx 4090 and i9-13900K,can I game while allocating 16 threads to a VM (my server)? yes	v2ray why would you do that
v2ray why would you do that	horsecode <Re: v2ray> and the other part is that i.e. in my native language, even GPT-4 struggles at times.
horsecode <Re: v2ray> and the other part is that i.e. in my native language, even GPT-4 struggles at times.	v2ray probably it doesn't have enough data in your language during base model training
v2ray probably it doesn't have enough data in your language during base model training	horsecode <Re: 6706full> you can do it especially with Hyper-V enabled (are you using VMware or what?) <Re: v2ray> my model oddly enough has a better syntax at times
6706full <Re: horsecode> yes.I know	v2ray <Re: horsecode> which dataset are you using are you training base model or fine tuning
v2ray <Re: horsecode> which dataset are you using are you training base model or fine tuning	horsecode <Re: v2ray> my own 4.6GB dataset that I've scraped from everywhere.
horsecode <Re: v2ray> my own 4.6GB dataset that I've scraped from everywhere.	v2ray ok😢
horsecode built it from scratch with a custom BPE tokenization	v2ray <Re: 6706full> xeon
6706full up to 1tb instead of 128gb	v2ray use multiple xeon cpus with high thread count is always better
horsecode it's not just cores or threads but also the other technologies	v2ray and with 8 channel ram the ram latency is low
6706full <Re: horsecode> New Intel AI technology	v2ray unlike the trash dual channel i9
horsecode what kind of abstraction layers does it have, what kind of thread/core load handling etc.	6706full only in Xeon <Re: v2ray> Cpu inference=Possible
horsecode there's threads and then there's threads	v2ray <Re: 6706full> no >:(
v2ray <Re: 6706full> no >:(	horsecode <Re: v2ray> also this, the bandwidth in many lanes can be much higher.
horsecode <Re: v2ray> also this, the bandwidth in many lanes can be much higher.	v2ray no cpu inference i hate cpu inference fuck
v2ray no cpu inference i hate cpu inference fuck	6706full <Re: v2ray> better than the i9 ngl,but still worsr than a gpu worse
6706full Frontier:exists note:Imagine doing that for only AI inference 37K accelerators 5PB RAM 4k CPUs	v2ray plz 5090 48gb vram :( i hate nvidia
v2ray plz 5090 48gb vram :( i hate nvidia	6706full <Re: v2ray> 5090 will have 36
6706full <Re: v2ray> 5090 will have 36	v2ray low vram on consumer gpu
6706full h100 has 48	v2ray <Re: 6706full> 80
6706full <Re: horsecode> "6gb" cant even run 13b cough	v2ray i think nvidia is trying to lock resolutions with the vram
horsecode 6GB is less than what I have on my laptop's GPU	v2ray and lock consumer gpus out of ai inference its evil
6706full rtx 4090x4 96gb	horsecode <Re: v2ray> idk about "locking out" though, you mean like on a hardware level?
horsecode <Re: v2ray> idk about "locking out" though, you mean like on a hardware level?	v2ray <Re: 6706full> just buy an a100 at that point <Re: horsecode> yeah, low vram=impossible to inference large models and offloading kills the speed
6706full <Re: 6706full> only $150k	v2ray gaming on xeon💀 you'll need high singlecore freq for gaming
v2ray gaming on xeon💀 you'll need high singlecore freq for gaming	6706full <Re: v2ray> best thing that should be 2x i9 or sth
6706full <Re: v2ray> best thing that should be 2x i9 or sth	v2ray so i9 better <Re: 6706full> i9 doesn't support dual cpu
purplesci Which girl looks the most believable?	v2ray cpu inference all ai generated
purplesci So there's no way to make them look more real?	v2ray idk they lack emotions look like zombies actually they trigger some uncanny valley on me especially the first 2
6706full Like,a secondary machine w i9 and 2xrtx 4090 should be able to game	v2ray just dont game why would you want to game
6706full the first has 2xthreadripper,10xH100,2TB RAM	v2ray :thonk:
v2ray :thonk:	6706full <Re: v2ray> didnt wanted to gaming time:<2% on avg <Re: 6706full> ONLY $500K!FULL POWER BILL!FAST INFERENCE(faster than any demo you EVER SEEN!1!1!2!2!)!ENJOY THE FINEST GPUS!
6706full <Re: v2ray> didnt wanted to gaming time:<2% on avg <Re: 6706full> ONLY $500K!FULL POWER BILL!FAST INFERENCE(faster than any demo you EVER SEEN!1!1!2!2!)!ENJOY THE FINEST GPUS!	v2ray :5938_think_gmagik:
6706full "full power bill" sus	v2ray ok
6706full <Re: dawn.dusk> or momey sus	v2ray <Re: dawn.dusk> openai
6706full *money <Re: dawn.dusk> 1 person vs 10k mods	v2ray openai have both money and api
dawn.dusk /s	v2ray 🤬
v2ray 🤬	purplesci <Re: v2ray> So you're saying it's better to use an anime or art model?
purplesci <Re: v2ray> So you're saying it's better to use an anime or art model?	v2ray <Re: purplesci> probably rip hands
insaneness <Re: purplesci> Second pic looks interesting	v2ray <Re: purplesci> all looks ai
insaneness Redditors should keep this for a month But No 2 days Reddit gives a fuck then	v2ray its like reddit will care about 2 days lol
v2ray its like reddit will care about 2 days lol	insaneness <Re: v2ray> One month and they will care. But 2 days. They have a deadline now Just waiting Redditors are dumb
insaneness <Re: v2ray> One month and they will care. But 2 days. They have a deadline now Just waiting Redditors are dumb	v2ray but you are a redditor 😊
insaneness I just 3 accs there Actually 10	v2ray you are multiple redditors
purplesci <Re: purplesci> Hot or not?	v2ray 😩 <Re: purplesci> misplaced foot weird hand but ok
purplesci I noticed. Is there a way to avoid this?	v2ray inpaint it until it get it right
purplesci I did it in Catbird.	v2ray or controlnet it and fix it in photoshop then i2i it back
insaneness Yea	v2ray what is catbird
insaneness Leonardi Can do inpainting	v2ray not local😩 just rent a gpu
insaneness Also automatic1111	v2ray on lambdalabs cheap
purplesci Not free though.	v2ray just run sd locally
insaneness Yup	v2ray you only need to pay electricity bills
insaneness Lmao	v2ray then rent gpu 😩
insaneness Then use playground or leonardi ai	v2ray why not its only 0.5$ per hour for rtx6000
insaneness #No #No	v2ray 😩
insaneness Use kaggle Only phone number required	v2ray or colab ig colab performance is shit but ok vram
v2ray or colab ig colab performance is shit but ok vram	insaneness <Re: v2ray> Banned
insaneness U can't use automatic1111 on colab	v2ray <Re: insaneness> what? wdym? noooooooooooooooooooo but why 😢
v2ray <Re: insaneness> what? wdym? noooooooooooooooooooo but why 😢	insaneness <Re: v2ray> Colab bans automatic1111 users
insaneness <Re: v2ray> Colab bans automatic1111 users	v2ray google is fucking evil
v2ray google is fucking evil	insaneness <Re: v2ray> Guess who reported it?
insaneness <Re: v2ray> Guess who reported it?	v2ray fuck
trent_k <Re: purplesci> They look too perfect. They need like a pimple or a fucked up tooth or a weird shaped eye. AI girls always look like models in a way that makes it obvious they're fake	insaneness <Re: v2ray> SD mod reported it to Google
insaneness <Re: v2ray> SD mod reported it to Google	v2ray why ban webui what did webui do
insaneness Cuz they want u to buy their one SD wants money	v2ray but idc ig, i dont even use colab😊
insaneness Same	v2ray <Re: insaneness> but its free and opensource
v2ray <Re: insaneness> but its free and opensource	insaneness I have automatic1111 at home <Re: v2ray> They have cloud
insaneness Lol	v2ray claude
insaneness Nope	v2ray claude
v2ray claude	goni0755 <Re: v2ray> is that better than 3.5 turbo?
goni0755 <Re: v2ray> is that better than 3.5 turbo?	v2ray yes
insaneness Lmfao	v2ray ofc
thundersnow_ claude instant, on the other hand, might not be	v2ray but only if you can pass the waitlist lol
goni0755 is it more reliable than gpt4?	v2ray i heard claude added internal moderation its more closed than #ClosedAI now
insaneness Which llm surpasses gpt4? I think none atm	v2ray none human brain
thundersnow_ Would the brain really surpass it? Well, it can do a lot of other tasks, but 4 is good at what it is.	v2ray ai generated
insaneness <Re: thundersnow_> Ye cuz 1 billion or more data	v2ray because the cloths dont have any texture
insaneness Parameters	v2ray this is scary wtf
v2ray this is scary wtf	insaneness <Re: v2ray> It always looks soft.
insaneness <Re: v2ray> It always looks soft.	v2ray uncanny valley moment
thundersnow_ It shrunk her shoulders. Made her look soft.	v2ray sd is bad at photorealistic
insaneness <Re: purplesci> Civitai has good models embeds	v2ray <Re: insaneness> only a few are good
v2ray <Re: insaneness> only a few are good	insaneness and loras <Re: v2ray> I only choose the top 1
insaneness and loras <Re: v2ray> I only choose the top 1	v2ray anything v3&5 are good
insaneness But hands Meh	v2ray use controlnet
insaneness Ye ik	v2ray then steal some images from pixiv ez godtier gen
insaneness Ye But hands	v2ray pose model is not really effective compared to canny
purplesci Can I use CivitAI with Catbird?	v2ray what even is catbird
insaneness <Re: purplesci> It's a archive of models lol https://civitai.com Always use safe tensor Never ckpt	v2ray nah civitai is pretty safe i dont think there're viruses
purplesci Does Leonardo have Dark Sushi Mix?	insaneness <Re: v2ray> #safe tensor
insaneness <Re: v2ray> #safe tensor	v2ray no
insaneness Dunno	v2ray no i hate safe
insaneness Use search button	v2ray tensor for no reason at all
insaneness Lmao	v2ray 🤔 i hate to hate safe tensor
insaneness #keep yourself safe	v2ray #kiss me softly
insaneness #No	v2ray #keep yourself safe #Damn Samsung keyboard

horsecode wait you can't fine-tune GPT3.5 API models, right?

v2ray nope you cant only base models

horsecode *sigh* <Re: v2ray> well, I do have my own datasets... just wondering how much money will that drain

v2ray <Re: horsecode> fine tuning through openai api can easily go through 1k+

horsecode <Re: v2ray> 💀

6706full If I have a Rtx 4090 and i9-13900K,can I game while allocating 16 threads to a VM (my server)? yes

v2ray why would you do that

horsecode <Re: v2ray> and the other part is that i.e. in my native language, even GPT-4 struggles at times.

v2ray probably it doesn't have enough data in your language during base model training

horsecode <Re: 6706full> you can do it especially with Hyper-V enabled (are you using VMware or what?) <Re: v2ray> my model oddly enough has a better syntax at times

6706full <Re: horsecode> yes.I know

v2ray <Re: horsecode> which dataset are you using are you training base model or fine tuning

horsecode <Re: v2ray> my own 4.6GB dataset that I've scraped from everywhere.

v2ray ok😢

horsecode built it from scratch with a custom BPE tokenization

v2ray <Re: 6706full> xeon

6706full up to 1tb instead of 128gb

v2ray use multiple xeon cpus with high thread count is always better

horsecode it's not just cores or threads but also the other technologies

v2ray and with 8 channel ram the ram latency is low

6706full <Re: horsecode> New Intel AI technology

v2ray unlike the trash dual channel i9

horsecode what kind of abstraction layers does it have, what kind of thread/core load handling etc.

6706full only in Xeon <Re: v2ray> Cpu inference=Possible

horsecode there's threads and then there's threads

v2ray <Re: 6706full> no >:(

horsecode <Re: v2ray> also this, the bandwidth in many lanes can be much higher.

v2ray no cpu inference i hate cpu inference fuck

6706full <Re: v2ray> better than the i9 ngl,but still worsr than a gpu worse

6706full Frontier:exists note:Imagine doing that for only AI inference 37K accelerators 5PB RAM 4k CPUs

v2ray plz 5090 48gb vram :( i hate nvidia

6706full <Re: v2ray> 5090 will have 36

v2ray low vram on consumer gpu

6706full h100 has 48

v2ray <Re: 6706full> 80

6706full <Re: horsecode> "6gb" cant even run 13b *cough*

v2ray i think nvidia is trying to lock resolutions with the vram

horsecode 6GB is less than what I have on my laptop's GPU

v2ray and lock consumer gpus out of ai inference its evil

6706full rtx 4090x4 96gb

horsecode <Re: v2ray> idk about "locking out" though, you mean like on a hardware level?

v2ray <Re: 6706full> just buy an a100 at that point <Re: horsecode> yeah, low vram=impossible to inference large models and offloading kills the speed

6706full <Re: 6706full> only $150k

v2ray gaming on xeon💀 you'll need high singlecore freq for gaming

6706full <Re: v2ray> best thing that should be 2x i9 or sth

v2ray so i9 better <Re: 6706full> i9 doesn't support dual cpu

purplesci Which girl looks the most believable?

v2ray cpu inference all ai generated

purplesci So there's no way to make them look more real?

v2ray idk they lack emotions look like zombies actually they trigger some uncanny valley on me especially the first 2

6706full Like,a secondary machine w i9 and 2xrtx 4090 should be able to game

v2ray just dont game why would you want to game

6706full the first has 2xthreadripper,10xH100,2TB RAM

v2ray *:thonk:*

6706full <Re: v2ray> didnt wanted to gaming time:<2% on avg <Re: 6706full> ONLY $500K!FULL POWER BILL!FAST INFERENCE(faster than any demo you EVER SEEN!1!1!2!2!)!ENJOY THE FINEST GPUS!

v2ray :5938_think_gmagik:

6706full "full power bill" sus

v2ray ok

6706full <Re: dawn.dusk> or momey sus

v2ray <Re: dawn.dusk> openai

6706full *money <Re: dawn.dusk> 1 person vs 10k mods

v2ray openai have both money and api

dawn.dusk /s

v2ray 🤬

purplesci <Re: v2ray> So you're saying it's better to use an anime or art model?

v2ray <Re: purplesci> probably rip hands

insaneness <Re: purplesci> Second pic looks interesting

v2ray <Re: purplesci> all looks ai

insaneness Redditors should keep this for a month But No 2 days Reddit gives a fuck then

v2ray its like reddit will care about 2 days lol

insaneness <Re: v2ray> One month and they will care. But 2 days. They have a deadline now Just waiting Redditors are dumb

v2ray but you are a redditor *😊*

insaneness I just 3 accs there Actually 10

v2ray you are multiple redditors

purplesci <Re: purplesci> Hot or not?

v2ray 😩 <Re: purplesci> misplaced foot weird hand but ok

purplesci I noticed. Is there a way to avoid this?

v2ray inpaint it until it get it right

purplesci I did it in Catbird.

v2ray or controlnet it and fix it in photoshop then i2i it back

insaneness Yea

v2ray what is catbird

insaneness Leonardi Can do inpainting

v2ray not local😩 just rent a gpu

insaneness Also automatic1111

v2ray on lambdalabs cheap

purplesci Not free though.

v2ray just run sd locally

insaneness Yup

v2ray you only need to pay electricity bills

insaneness Lmao

v2ray then rent gpu 😩

insaneness Then use playground or leonardi ai

v2ray why not its only 0.5$ per hour for rtx6000

insaneness #No #No

v2ray *😩*

insaneness Use kaggle Only phone number required

v2ray or colab ig colab performance is shit but ok vram

insaneness <Re: v2ray> Banned

insaneness U can't use automatic1111 on colab

v2ray <Re: insaneness> what? wdym? noooooooooooooooooooo but why 😢

insaneness <Re: v2ray> Colab bans automatic1111 users

v2ray google is fucking evil

insaneness <Re: v2ray> Guess who reported it?

v2ray fuck

trent_k <Re: purplesci> They look too perfect. They need like a pimple or a fucked up tooth or a weird shaped eye. AI girls always look like models in a way that makes it obvious they're fake

insaneness <Re: v2ray> SD mod reported it to Google

v2ray why ban webui what did webui do

insaneness Cuz they want u to buy their one SD wants money

v2ray but idc ig, i dont even use colab😊

insaneness Same

v2ray <Re: insaneness> but its free and opensource

insaneness I have automatic1111 at home <Re: v2ray> They have cloud

insaneness Lol

v2ray claude

insaneness Nope

v2ray claude

goni0755 <Re: v2ray> is that better than 3.5 turbo?

v2ray yes

insaneness Lmfao

v2ray ofc

thundersnow_ claude *instant*, on the other hand, might not be

v2ray but only if you can pass the waitlist lol

goni0755 is it more reliable than gpt4?

v2ray i heard claude added internal moderation its more closed than #ClosedAI now

insaneness Which llm surpasses gpt4? I think none atm

v2ray none human brain

thundersnow_ Would the brain really surpass it? Well, it can do a lot of other tasks, but 4 is good at what it is.

v2ray ai generated

insaneness <Re: thundersnow_> Ye cuz 1 billion or more data

v2ray because the cloths dont have any texture

insaneness Parameters

v2ray this is scary wtf

insaneness <Re: v2ray> It always looks soft.

v2ray uncanny valley moment

thundersnow_ It shrunk her shoulders. Made her look soft.

v2ray sd is bad at photorealistic

insaneness <Re: purplesci> Civitai has good models embeds

v2ray <Re: insaneness> only a few are good

insaneness and loras <Re: v2ray> I only choose the top 1

v2ray anything v3&5 are good

insaneness But hands Meh

v2ray use controlnet

insaneness Ye ik

v2ray then steal some images from pixiv ez godtier gen

insaneness Ye But hands

v2ray pose model is not really effective compared to canny

purplesci Can I use CivitAI with Catbird?

v2ray what even is catbird

insaneness <Re: purplesci> It's a archive of models lol https://civitai.com Always use safe tensor Never ckpt

v2ray nah civitai is pretty safe i dont think there're viruses

purplesci Does Leonardo have Dark Sushi Mix?

insaneness <Re: v2ray> #safe tensor

v2ray no

insaneness Dunno

v2ray no i hate safe

insaneness Use search button

v2ray tensor for no reason at all

insaneness Lmao

v2ray 🤔 i hate to hate safe tensor

insaneness #keep yourself safe

v2ray #kiss me softly

insaneness #No

v2ray #keep yourself safe #Damn Samsung keyboard