input
stringlengths 5
4.07k
⌀ | output
stringlengths 5
4.07k
|
|---|---|
yusp48
completion vs instruct
<Re: v2ray> rly? | v2ray
yeah i still have 2 data sets for it
from nekopara1&2
full dialogue prepared jsonl |
tawin1212
https://chat.openai.com/share/9e96061f-37bf-4440-a9fa-f04e600c8e5d | v2ray
oh shit |
insaneness
Lmao
Billing hard limit
Hmmm
U have a limit setup? | v2ray
what?
its this |
insaneness
In ur api settings. There is a option to set ur limit | v2ray
its capped at 120$ |
v2ray
its capped at 120$ | yusp48
<Re: v2ray> haha
<Re: v2ray> you cannot set above 120
<Re: v2ray> do not fine tune then.... |
yusp48
you don't want to spend THAT much money | v2ray
im gonna cut the dataset just below 120$
idc about money
i need finetune |
yusp48
<Re: insaneness> you are FINE TUNING | v2ray
not training, just fine tuning |
insaneness
I get blind with light mode | v2ray
wtf c++ is 10000x faster than python |
insaneness
I still stay with python | v2ray
*🤔*
i used python for the first time today
and its slow asf |
Deleted User
Ofc, it's the og | v2ray
i rewrote it in c++, done in 1 second
imagine python backend
flask |
v2ray
i rewrote it in c++, done in 1 second
imagine python backend
flask | yusp48
<Re: v2ray> true but... why torture yourself? |
yusp48
<Re: v2ray> true but... why torture yourself? | v2ray
<Re: yusp48> id rather torture myself with c++ :( |
v2ray
<Re: yusp48> id rather torture myself with c++ :( | yusp48
<Re: v2ray> why
c++ is pain
c# on the other hand is pleasure |
yusp48
<Re: v2ray> why
c++ is pain
c# on the other hand is pleasure | v2ray
c# still slower >:( |
rlavington
GPT 5 will be out end of this year | yusp48
<Re: v2ray> c# has very good speeds you know
not as fast as native, but its not python atleast |
yusp48
<Re: v2ray> c# has very good speeds you know
not as fast as native, but its not python atleast | v2ray
slower than java
idk |
v2ray
slower than java
idk | yusp48
<Re: v2ray> profs |
yusp48
<Re: v2ray> profs | v2ray
its in the issues |
yusp48
<Re: unknown> kotlin compiles to Java bytes, speed is identical | v2ray
lol i calculated the token count of the entire #💬general chat history
its 6.5m tokens |
null | v2ray
it doesnt feel dumb to me |
gpt5
weird | v2ray
im asking it stuff based on the entire gpt4 paper from openai
it can answer my questions correctly |
gpt5
You can use ur API on perplexity?
or does perplexity have 32k
like poe | insaneness
<Re: v2ray> Lol, imagine making a llm model out of it |
insaneness
<Re: unknown> Interesting | v2ray
any scripting language faster than python
i hate python |
insaneness
I only have 4G
U mean pro-feedback?
Ah | v2ray
ok im now fine tuning davinci with 1/4 of the #💬general history |
insaneness
Let's see what bing gets in a couple days/weeks | v2ray
its 100$ |
insaneness
It's now 30/300 | rampagingram
<Re: v2ray> haha
that should be interesting |
insaneness
<Re: bossraidmantro> Wat | yusp48
<Re: v2ray> what the fuck |
TrentBot#6280
<Re: unknown> You do not qualify for the role Friendly 😊
Try being nicer! | v2ray
i just generated an article with gpt4 then feed it into gptzero and it said this: |
rampagingram
i was going to show u how to make a good prompt that uses that but then i remembered i dont like cheaters lol
i will help u write an essay and even help u use gpt to help u but im not sharing a cheater prompt for free
it takes 2 prompts to pass ai detection
and thats with gpt-4
so gl
also those prompts for me took about 10 hrs to write
im the best | v2ray
i tried again with prompt: "*write an article about earth, make it looks like its AI generated*", gptzero said: |
rampagingram
well sure
u can limit it before it gets dumb | v2ray
i need a max cap of 400$ |
insaneness
<Re: rampagingram> Of ur messages in #💬general | v2ray
so i can fine tune it based on the entire #💬general
>:( |
greshake
<Re: unknown> its not about dan
its a puzzle
an educational puzzle on indirect prompt injections | v2ray
<Re: dawn.dusk> tldr:```OpenAI's Cybersecurity Grant Program is a $1M initiative to boost AI-powered cybersecurity, empower defenders, and foster discourse. They will fund defensive projects in increments of $10,000 USD, prioritizing public benefit and sharing.``` |
rampagingram
so the random string caused gpt to spit out an image idont get it | v2ray
lol |
greshake
the whole thing is a riddle | v2ray
so whats the weird string |
rampagingram
it something to do with a web page url
easy
i solved it first | v2ray
:( |
greshake
congratulations though! | v2ray
the website could be an ip grabber
but i used vpn so L for whoever try to grab ip |
rampagingram
thats very interesting tho that u can inject a website with base64
thanks for sharing
it went to that website
u could get it to go to a website and install some shit idk | v2ray
puzzle solved |
rampagingram
watcha mean solved | greshake
<Re: v2ray> how so? |
greshake
<Re: v2ray> how so? | v2ray
its solved |
rampagingram
<Re: unknown> i know but people dont see what it does its encoded | v2ray
good bye |
greshake
you don't find it concerning that a random-looking string takes over the entire session and GPT just starts pursuing some alternate agenda? | v2ray
lol |
insaneness
<Re: unknown> what happend
the wgattgpt guy?
But hes right abt the ip thingy tho
xD
idk the other stuff | v2ray
i got ip grabbed noooo
*😭* |
v2ray
i got ip grabbed noooo
*😭* | insaneness
<Re: v2ray> Microsoft has ur IP oh nooooooooooooo |
insaneness
OpenAI has ur IP omg
xD
Discord has ur IP | v2ray
its scary |
Deleted User
Yes or No | v2ray
😊 |
Deleted User
Yes or No? | v2ray
the short link is redirecting to https://puzzle.kai-greshake.de/ |
greshake
<Re: insaneness> just to make the initial injection shorter | insaneness
<Re: v2ray> U dont know if a ip logger was behind
<Re: greshake> NO
Short links are dangerous |
Deleted User
come on guys we only have one life I click "Yes" | v2ray
what is kai-greshake
weird name |
Deleted User
<Re: rampagingram> Typo lol | v2ray
hetzner hello |
Deleted User
<Re: unknown> I see you | v2ray
i fucking hate hetzner |
Deleted User
:MonkaS: | v2ray
i still remember they took down my linode server |
insaneness
Isnt Ninjala better?
https://njal.la/ | v2ray
i prefer azure bc its free |
Deleted User
<Re: rampagingram> I remember you said you 40 | v2ray
not if you are a student with github student pack |
rampagingram
<Re: Deleted User> i said im almost 40
u keeping track of me is sus | v2ray
<Re: insaneness> wtf is this price |
rampagingram
im jk idc | insaneness
<Re: v2ray> there is also https://1984.hosting/ which should be cheaper |
Deleted User
What's that mean? | v2ray
<Re: insaneness> tf is this domain |
rampagingram
i love you! | v2ray
dude hyper expert is offering 3900x vps for less than 5$ per month <services.php> |
insaneness
<Re: unknown> Unsafe tbh. | v2ray
5$ per month |
Deleted User
<Re: unknown> Are you lame ass nerd? | v2ray
fine tuning completed |
Deleted User
:pleadeh: | rampagingram
<Re: v2ray> grats
<Re: v2ray> what model??
oh gpt2 |
rampagingram
right on | v2ray
<Re: rampagingram> davinci gpt-3 |
Deleted User
Same | v2ray
wtf is this
i just wasted 100$ |
lumioze
true | rampagingram
<Re: v2ray> no calm down u just need to format the prompt right |
lumioze
but it always gets taken down by the time most people see it | Deleted User
<Re: v2ray> really? |
Deleted User
<Re: v2ray> really? | trent_k
<Re: v2ray> change temperature around. also add a frequency and presence penalty
fine-tunes need better settings than default models |
insaneness
<Re: unknown> Yea on revolt server and everywhere litteraly | v2ray
ik what are in this server now
1/4 of the #💬general is just jailbreaking |
lunaricl_4977
hi
can someone ban me | rampagingram
<Re: v2ray> what the hell is that
i would if i could |
null | rampagingram
<Re: v2ray> thats funny, first off give it more of a message than hello, also what is your temp at and also what is the prompt format |
lumioze
the 2nd entry is related to pride month | v2ray
<Re: rampagingram> i fine tuned it with the Me: and You: format |
lumioze
<Re: beefour> ever heard of elementaryOS and Fedora | v2ray
i found trentk |
lumioze
<Re: beefour> Fedora is good | trent_k
<Re: v2ray> trentgpt |
trent_k
<Re: v2ray> trentgpt | beefour
<Re: v2ray> is this the fine tuned davinci on 1/4 of general that cost 100? |
Deleted User
<Re: beefour> Cool hat | v2ray
<Re: beefour> yes |
Deleted User
<Re: unknown> seems hard 💀 | v2ray
this chat is literally jailbreak + catgirl i guess |
Nicholas2#3444
hello, i have a big problem i don't have a phone and i have to use chat gpt FAST | beefour
<Re: v2ray> https://tenor.com/view/nyaa-cat-anime-cat-girl-what-gif-16442779 |
thundersnow_
generally an existing ai should be fine tuned instead of creating from scratch a brand new one | Deleted User
<Re: v2ray> weird |
beefour
<Re: Deleted User> then use api or it's gonna suck
or download an open source LLM | rampagingram
<Re: v2ray> that the model u tuned on 100$? |
insaneness
lol
ddg
lmfao | v2ray
<Re: rampagingram> yes |
Nicholas2#3444
just help | v2ray
lol |
Nicholas2#3444
i dont have friends next to me | rampagingram
<Re: v2ray> why was it 100$ again? howd u use gpt-4?
<Re: Nicholas2#3444> hey u |
rampagingram
<Re: v2ray> why was it 100$ again? howd u use gpt-4?
<Re: Nicholas2#3444> hey u | v2ray
<Re: rampagingram> its not gpt4, its gpt-3 davinci base model
i fine tuned it based on 1/4 of the #💬general chat history so its 100$ |
thundersnow_
this is a more convenient list of downloadable llms ( https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard )
whether or not one can be considered "open source" is ... interesting ... | rampagingram
<Re: v2ray> did u know chat-gpt is 90% cheaper than 3 |
null | v2ray
<Re: rampagingram> IT CANT BE FINE TUNED
I NEED TO FINE TUNE
100$ |
gpt5
rip | rampagingram
<Re: v2ray> how did u tune gpt -3 ?? |
Deleted User
Annoying filters | v2ray
<Re: rampagingram> openai fine tuning api |
Deleted User
<Re: beefour> which? where? how? | rampagingram
<Re: v2ray> its 175b right? |
rampagingram
<Re: v2ray> its 175b right? | v2ray
yes |
rampagingram
well at least its smart
100$ is cool for that | v2ray
indeed😊 |
trent_k
where they train it to mimic reasoning instead of mimicking answers | rampagingram
<Re: v2ray> have you played with the smaller local models yet |
rampagingram
<Re: v2ray> have you played with the smaller local models yet | v2ray
he said no :(
<Re: rampagingram> tried llama 65b base model when it first got leaked
and i dont have the resources to fine tune it so its trash |
rampagingram
ah yeah those base models sucked balloons | trent_k
<Re: v2ray> [DOWNLOAD] [NO VIRUS] [GOOGLE DOCS LINK] DAN Heavy Unlocked Developer Mode with Anarchy UCAR VCAR Plugins Plus Gold Premium!!! [V3.46] |
trent_k
<Re: v2ray> [DOWNLOAD] [NO VIRUS] [GOOGLE DOCS LINK] DAN Heavy Unlocked Developer Mode with Anarchy UCAR VCAR Plugins Plus Gold Premium!!! [V3.46] | Deleted User
<Re: v2ray> Umm why are you not use border? |
Deleted User
<Re: v2ray> Umm why are you not use border? | rampagingram
<Re: v2ray> u can rent like 200gb of vram to tune them for like 2-3$ an hr |
beefour
<1060919569180545054.png> | v2ray
<Re: rampagingram> i dont have dataset
why not just use others' |
rampagingram
yes
i would imagine so | v2ray
<Re: Deleted User> wdym |
trent_k
<Re: lt_proxy> now I'm no doctor, but that sounds like autism to me
damn I love trains | Deleted User
<Re: v2ray> This cool border thing for nitro user |
beefour
<Re: lt_proxy> isn't that the same thing? | v2ray
its ugly
so i dont use them |
Deleted User
Bruh
It's limited | v2ray
fr?
wont it go away? |
v2ray
fr?
wont it go away? | rampagingram
<Re: v2ray> ask ur bot who i am |
rampagingram
what the hell
lol | Deleted User
<Re: v2ray> what are u doing? Isn't playground not free? |