#programming
1 messages · Page 510 of 1
Discord gives you the opportunity to generate then and put it somewhere
I think like 8 of them
Nah 10
I have 5 left
Soon to be 4
Need to get another android device online
Would use the 2fa code if I didn't nuke it
Should use one to reinstate 2FA
try a bunch of zeros
they're randomly generated, so a bumch of 0s dont work
might want to make it a habit to save them 
Quite possibly a good idea, yes
IP leak?
a bunch of 1s
discord codes are 8 chars long
that does look like ISP IP
saving them where? if you save them on the compromised evice you're still fucked
mine is in cold storage
they start with 100.64 ig
i have the discord backup codes of my main in a notebook
and as im not using my main, you can guess
lost it years ago sob
It is infra in nature, I'll give you that
pc, cloud storage, password managers, you name it
i forget what services i even have that for besides discord
maybe sony
sony wont let me log in to my headphones 
Just discord here
who needs 2fa anyway I guess
every popular javascript dependency maintainer be like
i have 2fa
but the codes are backup i think
and sony didnt want to send codes to my new phone cuz fuck you
i need to reboot the server
this is how you end up with a massive headache because one day your phone suddenly breaks and now you're locked out
o7 ears
https://www.youtube.com/watch?v=UCBAqaT4SQc evilDentge
This video changes based on your quality settings.
Click this link https://boot.dev/?promo=PORTALRUNNER and use my code PORTALRUNNER to get 25% off your first payment for Boot.dev.
Thank you for watching!
Join our Discord server: https://p2r3.com/discord
Schedule a one-on-one for tutoring or consulting: https://learn.p2r3.com/
Music arrangemen...

hmmm
tty1 freezes on boto cuz of dispalymanager
tty2 jsut froze cuz i tried to add gpu to btop
but it does work on tty3
tty3 is the goat for some reason
lol
I print them and it's in my wallet and also in my wardrobe
it was about time I watched this, interesting concept indeed
i have a slop 
I saw the thumbnail but havent watched it yet
reminds me of that video that had 2 english tracks by adding one in “klingon”
i watched it already, was interesting i guess but i knew msot of the tricks already due to gamedev
Now give it full control of your network and your social media 
Wdym no? It's your clone, so surely it will act like you 
aka openclaw
SamVanSlop2000?
no
openslop
I am against the access to network for personal reasons
mainly I don't need it rming my stuff instead
you don't need network access for that tho?
the line full control over network is my issue 
I'd also do it segmented
not full net tho
espically as I just got finished adding devices to a headscale instance
discord auth is one of the methods of all time I'll have to say
Taken from this session: https://www.youtube.com/watch?v=0ebreGjhEvo
tsoding is such a interesting channel.
perfect content to just put in the background... also 
can say
i was wondering why my internet was lagging when i dont see much internet usage on my desktop
but the server doesnt show up on the desktop's tracker
and my server is using the full 10MB/s bandwidth
i am preparing for slop that doesnt need a templates
cuz the webui wouldn't work on the opt-125m
Daang, I didn't know he had second channel. Thanks!
I always loved his main channel content lol
So you don't need to tell it "Make no mistake"?
you got it the other way around 
Sloppa
Wait so Tsoding is his second channel and daily is his main?
"You are SamVanMaele, be autistic about curly braces" 
only eletricity prices
load time is pretty good 
I guess i did kinda leave the ports open to the local network
We love setting traffic to 0.0.0.0
You'd have to expose to localhost and local IP if you didn't want other IP connections
Wdym?

So if you set to 192.168.129.233 you'd only get traffic from that if your using --listen on whatever frotnend your using
I see
I feel like its kinda funny to have the slop accesible tho
Does it have shared history?
I only saw one logged in user
Don't know if I could log out
Didn't want to do so without asking
I'd guess it is autologging into a user
Oh there's a bunch of chinese now
I can hear my server chugging away from far away
Slop generation 
Isn't like we are tunneling localhost traffic

That's nginx or similar software's job
I just need to improve the setup to not be broken
nginx my beloved
I'm pretty sure about half of the tailnet is nixos
its refusing to call me a slur so its at least partially filtered
I did some while the p40 was online, but not really anything interesting back when MoE was a new concept
There are abliterated models on hugging face
I think Mistral and Nous Hermes models are uncensored
Also check here
You can also do it yourself if you want
Used that before
Can confirm it's dumb
deepseek
Somehow had the best results with Nous Hermes 3
It is running on 24gb so it might work decently
it has a 4090
honestly deepseek is probably the best public opensource model
though naturally it has high requirements
Deepseek is really trying to progress
its on par with some earlier gpt releases imo currently
So yeah it tries to go as hard as it can
well... use qwen 3.5 9b
With the 1M context window it is bound to be nice
its the only decent way to have a good model and high context without draining ur bank currently
opus has gotten cheaper with its newer variations but that price is still terrible
even sonnet
https://huggingface.co/HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive
i personally tried this one... it seemed fairly good
u guys remember when pygmalion was the only decent way to host a LLM locally
what was it 3b? 5b? 7b?
this one has deepseek-ai/DeepSeek-V4-Pro (reasoning=enabled)
Uncensored aggressive lmao
Yeah sounds great 
lmk how v4 is
where the would normally be a vllm button there is none here

right vllm isn't using gguf
well...
(in case of a federal investigation. this is pure parody
)
qwen is not gonna be good enough for most questions like those anyways
Approved by Evil
i have to shut down vllm to activate a new model, and the vram stays in use the whole time even when not using the ai. kinda bad design
ah yes because i want to ask stuff like that all the time anyway 
why not just run llama.cpp?
VLLM is better for multi user usage

Like if you actually serve it to many people, vLLM is better
write ur own inference tool like a chad
llama cpp with their router mode is totally enough for normal users 
i just have it running in the background on localhost all the time

it is active for anyone on the tailscale rn when its up
Yeah if you are only using it for 1-4 people I'd say llama.cpp is better
Like 1-4 people at once tho
So you can have 100 but only 1-4 active at a time is fine
vllm works good with streaming if you need it
honestly hated streaming as a feature since it was introduced
but its like forced into so many providers now
Isn't that's like every model ever nowadays
Sorry, inference engine
Ollama also does it
But i think vllm is still the fastest isn't it
its so good
Yes for multi-user
But for small number of connection, eh
It's not THAT big of a difference I guess
General model runners will always have a weakness
Espically if your doing something very far out of the scope of what it can handle
If your doing the millionith distill, then you will be good most of the time
BGM
If a company had infinite time and budget to make a unicorn piece of hardware with their current and past tech, which company would make the best non specialized compute unit.
10
22
1
Intel
1173553648534568960
SMILE
BGM
"oh whats this? root user? i wonder what it means"
"cool username honestly, i'm going to stick with it from now on"
"why dont i install steam"
"oh wait, hmm, i'm low on storage"
"let me just clean up this old project files directory"
...and the rest is history

pluu
pluu??
omg i so want to do things but i so not want to do really
and they even ask 'how you getting motivation'
lizard?

may i ask common sense?
re-active -> not doing anything unless asked
pro-active -> doing things when no one asked
folly_future::subscribe is pro-active, it invoke callback on promise side, no one asked future to do it
rust Future::poll is re-active, it not doing literally anything unless asked to, and has no outer mechanism
-# underline Waker is just notify, but not invoke future
mapping is clear and direct, right?
just terminology
FTL BGM is good 
https://www.youtube.com/watch?v=WFkGjEut9U4
Song list:
ORIGINAL GAME
0:01 - Space Cruise (Title Theme)
3:57 - Milky Way (Explore)
6:33 - Milky Way (Battle)
9:13 - Civil (Explore)
11:53 - Civil (Battle)
14:32 - Cosmos (Explore)
18:53 - Cosmos (Battle)
23:13 - Deepspace (Explore)
26:21 - Deepspace (Battle)
29:27 - Debris (Explore)
32:19 - Debris (Battle)
35:11 - Mantis (Explore)
38:12 - Man...

oh yesterday i randomly thought this gif's neuro expression has konii vibes 
breakfast
did you know Colonel Sanders Ran the Cafeteria at the Oak Ridge National Research lab in the 40s
a
3
5
4
ab
1163912230585237655
enub
the usernames you guys have seem to keep regressing and devolving ngl

wdym
look who's talking
me username always the same 
i think most of us here have had the same nickname for a long time
false because i said so

Where is John ?
The cool refreshing taste of a crisp egg
true because you said opposite
only for konii

hi karamel
hi bluno
also false because what does caibi even mean
Intel only release new tech when AMD is catching their ass up in stock 
📋 todo: "planning 3 task(s)"
Why would you avoid bitwarden? I just selfhost a vaultwarden and connect my bitwarden apps to it
Not really if you do daily backups, which with vaultwarden are like 300mb maybe. Also easy to setup on proxmox with alpine lxc as example..
Because it is yet another attack surface if I self host bitwarden
I'd rather not deal with too many services

How do I stop username quality from degrading
not you

by spreading Toast

i mean youve always been superbox so like there has been nothing to change
meanwhile someone in chat will probably adopt the name nyaaaa mrrp meow nyannn or something like that

theyd also probably have a cat pfp or something

idk man thats just how #programming goes
nya mrr meow

now i see what are you talking about
Cannot use #programming to become real gatgirl
-# just thing i noticed
i was thinking, and had quick question, asked ai
now i 3 hours writing text in this box instead of normal text file
now when i noticed it, i can finally copy everything to text file
kitty

oh no
it break text
omg there so many spaces i have to remove
Coding without syntax highlighting chat is this real
btw
i remember some one said you cannot make moveToThread in rust
then wtf am i doing now? xD
yes this real
there no proper highlight when you mix 2 languages
i write in both rust and cpp at same time
you need a comsci degree for that :p and even then its optional
RawWaker provide void*
so for my context i can make thread affinity at executor lvl
since i have to anyway make Task contain Waker, to make Waker move Task from self ref to taskQueue
i can also add fields about thread affinity
bool shouldNotBeStolen and int16 shouldRunOnThread
would look like something something
moveToObjectThread(io_object).await?
btw i still not sure how tokio async threading looks like
i just doing stuff like how i feel it
That is a valid concern
where's john
@native surge
@native surge
@native surge
@native surge
@native surge
do they have portugal

Omg hi idol❗ ❗ ❗ ❗ ❗



The controller thingy is level 3 now
@true hemlock this you
🥤🦆
Duck Review
6/10
uu
cognac good with black tea

@sick owl @fast pagoda what is your recommended locally run LLM that can be run on consumer HW for document summary?
Because I think my venture to vibe coding is a bust, I think I'll look into their usage to do some other tasks
Qwen 3.5/3.6
Alright, noted. Why did you suggest that tho if I may ask?
Token memory usage efficiency
Hmmm, alright
Goodmorning
good meowning
good
generalized left-to-right, leftmost-derivation parsing using graph structured stack and shared packed parse forest
ning





I have a secret recipe for chicken nuggets that I'm not supposed to tell anybody because it's my secret recipe but i'll tell you anyway it's not that hard you just need some chicken nuggets and a frying pan first you take one nugget and then you catch in the palm of your hand then you open your hand the nugget is gone so what you do then is you put another nugget in the frying pan and then you close your eyes and you imagine that the first nugget is in your hand and you open your hand and the nugget is gone so the frying pan is now empty of nuggets so then you open your hand again and a new nugget appears so then you close your hand again and the nugget is gone so you open your hand and a new nugget appears and you close your hand again and the nugget is gone so you open your hand again and again and again and a new nugget appears and you close your hand again and again and the nuggets keep appearing and you just keep opening and closing your hands and the frying pan keeps getting emptier and emptier and eventually the nuggets are gone
@olive sable game night
shadow "not neuro fan" lilac
nrr
stats person called
mlntcan d

we're playing res evil in order rn. so res evil 1 remake

these older games dont want to run at 4K tho

ms. info

neurt
4K is scam 
Just blow up 1080p to wall size projector and watch from 7' away 
Guys where is John ?
John AI
The power of vibecode?
actual book has memmes
c++ isnt that bad
it's easy enough if u know how low level memory management works
and even then just use smart pointers
driving below speed limit on motorway
go to prison
I personally like to believe the meme is talking about python vs c++ developers, not the difficulty of the langs
Not that hard when you're driving below speed limit
In ye olden days, that doesn't exists
And that book may be from that time
surprisingly
difficult
I just use KeePassXC on my desktop and KeePassDX on Android, then store the database in cloud storage. Currently self-hosted nextcloud but whatever should work.
You can point KeePassDX directly at an installed cloud storage provider without manual syncing.

actually peak

No
Res evil 1 remake does max 1080p, and it fine i guess
ip grabber at the left
pls click here
https://ipgrabber.com/xAfRq56A
my ip is 196.202.240.225
my ip is charaverk.online
my ip is 213.55.243.216

deadass?
The joys of being behind CGNAT
not really a joy since it has a range anyway
99% of ip devices are behind nat
I assume it looks less washed out IRL
Depends, how much vram you got?
Uhhh, lemme get you the spec
If there are no good one then I suppose I'd use cloudflare worker AI lol
If I say this in my official textbook I’d drop out
Making fun of the material is for people to do with their friends then get in trouble for
Book should take self seriously

it ain’t that deep
meme is just “python easier lang than cpp”
not making fun of anything
And what model did I rec to you before
Just so I can check I'm not gonna rec the same one twice
Gemma 8-bit quant
The latest one
The uhhh, E2B
I asked between E4B 4bit vs E2B 8bit
But I wanna know if there's a better choice
cpp dev explain why you print with bitshift
rust dev explain how operator. work
then rust dev explain wtf is variance and how pinning work
Also I asked for coding model last time. But I think coding are cooked on small model
coding vs agent very different capabilities most of the smallish models are fine scoped to like one function
but you wont get opus schizoing for 3 hours to produce something that somehow works against all odds
That explains a lot
not even xkcd?
Yeah, that's why I opt to just do document summarization
Which, it may be useful
it is funny that C++26 have an import keyword now
which imports a precompiled module, just like dear CPython

i mainly use them for terminal stuff, they're pretty good at file ops and such
dont let them fuck around in root tho
It’s only kinda very illegal 
It's only illegal if you get caught by law enforcement
Bragging about it online definitely isn't going to get you caught of course 
unbothered supernova
Bragging online is actually fantastic OPSEC
And you should always do this when you do anything black hat
erm what
You should also do all your bright net stuff on your onion browser
Check your easily identified personal emails, browse your twitter that has your legal name on it, send messages to direct family members
police definitely cannot catch me if I use Surfshark VPN
#AD

Surfshark™ VPN can ensure your online privacy and security by encrypting your internet connection. With Surfshark™ VPN, you can browse the internet anonymously, access geo-restricted content, and...

If SoCute()
Print() ":neuroSoCute:"```
Is this valid python scripting or did I mess up my own joke again
Bragging about illegal activity on Discord is a great idea
the syntax could be better but who knows what compiles in python atp
Isn’t there a report button these days
And you should do it always all the time
surely they do something 
one can only intentionally mess up this bad

It's been a few months for me
I'm still learning
neurocord report button does something pretty cool 
it invites a wild @real sierra to the party, more specifically his own birthday party
it'll just be an automated email sent to you, as long as they're not sending CP-related contents. That's what they only care about
just tag @reef condor while replying to the message you wanna report.
Mutually assured destruction shall follow
why did I get syntax error in line 1 with Python 2.7.8 :((
Wdym?
I think the terms need defined?
Naah, just the people reflection
ragebait
Oh yeah tbh you wanna use the cloud if you are gonna rely on an LLM for coding
Neuro() = True
SoCute() = [defined function i think]
Still useful to know what model I rec'd since a lot of the best ones right now are generalists
We need blind out curtains still. Rn the sun is annoying
I'm genuinely so very outta practice
Yeah, that's why I pivoted to document summary
while neuro():
if so_cute():
print(“Neuro So Cute”)
time.sleep(0.01)
Also this is what you recommend me last time @sick owl
Right, gotta sleep the script or it'll just crash I think
I mean maybe you wanna spend 20K on hardware to run Kimi whatever
bro i'll be real I started learning this year and fell off the wagon so hard I gotta get back in there 
Or however much the needed HW costs
👀
Ah yeah, I thought you had less ram than you do
I remember
I moved the machine lol
But also yeah, small model are cooked for coding no matter which I choose 
great, now it'll sleep for 35±10ms per cycle instead of desired 10ms

Gemma 26BA4B would be okay for simple coding stuff but also take up your entire ram lol
And since you're running on CPU any decently sized dense model would be agonisingly slow
Yeah. I give up on the vibe coding arc 
But document summary may still be feasible
Now do it in brain
if I have 400+ cpus that can communicates with each other simultaneously, can I emulate a gpu 
Like, no delay communication?
You still need a lot more of cpu but yes
Also, DDR5 limited
Alright my 175k parameter model is in pretraining
1,000,000 times smaller than gpt 3.5

I mean, 8800 GTX only had 128 cuda cores back then

Batch size 256 on 8gb laptop GPU 
Honestly for that you could try the F16 gguf of GPT OSS 20B
875kb int8 checkpoint file size
Alternatively, if you don't mind filling up your RAM you could run Gemma 4 26BA4B Q8 without vision at decent context and get smarter answers
Similar recommended settings to E4B in that case
If you use GPT OSS you can check the model card for recommended sampler settings
Oh, but make sure to limit context checkpointing or it will blow up your PC memory leak style
That gets big fast on gemma at long contexts
Gonna run this thing on Monroe LA7-160
From 1958
Alright, will do that then. Thanks @sick owl
now we know they’re broke asf
Guys, does neuro-sama using Vtube Studio?
Neuro renders using probably the Live2D SDK rather than VTS
THX!
What is unity used for then? Because Vedal mentioned it a lot
As the source of memory leak that is
why not just use google cloud tpu
“If ur homeless, why not just buy a house?”
bro is literally free on kaggle
128gb vram
AFAIK it's rendering Neuro and most of the stuff you see on stream + a whole bunch of other stuff we don't know about
That seems... inefficient
like, pretty sure the Discord integration was part of that at some point 
-# source is some random mention of a C# Discord library ages ago idk I don't keep a list of sources
Like, I think for regular stuff Neuro can use Live2D with just OBS right? Because I bet the LLM are not in unity as well
Unless... 
I mean she does use Live2D, and VTube Studio is also a Unity app
the Live2D Unity integration is good, there's not really any big inefficiency there
yee, I think so
but we just don't know a lot of internals there, so can only speculate 
Knowing Shiro journey in Unity, I bet it's the network stuff 
LaTeX moment
I only ever needed to learn it for a linear algebra class
I kind of like it in a weird way?
Just feels like markdown mixed with microslop word
random question but does anybody have any experience here with gunsmithing or high velocity projectiles in general
i was trying to mod a nerf gun but wasn't sure if i'd have to reinforce other parts to withstand the increased fps
apart from swapping out the main spring for a heavier one and swapping out the barrel and breech for metal parts
i was wondering whether to put the end of the barrel at the open end as threaded or with m2/m3 screw holes
to be able to put in a rifle choke/linear compensator like a scar/bcar
Like computer hardware
steel is hardware
Not powertools
i buy it at the hardware store
nice try FBI
thats for insanity and autism not genuine questions
Ellie's discord maybe
I'm not sure they'll be too happy about discussing building a lethal weapon lmao
i imagine most discords wont
i dont think a foam blaster is lethal
compressed gas is better than springs
its a bolt action with single load rounds
i miss nerf battles
it is spring from its very core
Maven do you know anyone called gradle?
rip

i'll just add a 7kg retaliator spring
the current spring is 2.5/3kg
ye cachy kernel on nix
Kernel/torture.o 
peak naming
my question was since im adding a thinner barrel to speed up the darts and want to add a rifle choke (think muzzle brake but opposite function) should i have the barrel threaded at the open end or just put in 3 m2 screw holes and call it a day
i mean, they probably dont make much of a difference, it's up to you, i would thread the barrel for better looks
so...
shortcut to split is ctrl+shift+9 or+0
No
are we like, adding modifications to firearms in here???
didnt know it was that kind of channel
nerf gun
no firearms
oh...
we like to be legal
yes i like firearms but im not allowed to own em yet
so uncool...
"chat i have a nerf gun that shoots at 900m/s how do i put a silencer on it?"
i know a guy...
"The nerf darts are made of lead will they hurt?"
hey i cant do shit man the license for guns is hard to get here
at this speed it would require silencer bigger than gun itself
"hypothetically"
"in Minecraft"
LETS SJUST SAY if i DID have a guy who NEEDED this for some reason
dawg the best i can get out of my gun is 60m/s before the stock and entire body of the gun shatters
turns out that doesnt work
HOW would i go about doing it?
first of all you dont
no? it's a 7.62x51mm
second if you have to dont scream about it
well kinda
it's a joke
the rule of the fight club is you dont talk about the fight club kona
only chemical powered guns can reach supersonic
if i wanted to make my "super dorito nerf gun" automatic how do i?
fires once
or railguns
or me on a can of beans
bean silencer
yes yes i know but still
hmmm ill go and do some cad cam analysis and see which one is less likely to break my gun
and which one doesnt break the bank either
because believe it or not getting metals machined is expensive
have you seen this one btw?
it have big bullet with low speed (practically silent)
only 300m/s for 12.7mm
called ВССК «Выхлоп»
effective distance < 300m btw
so kinda useless
ya
notice silencer
ye so?
thats not a silencer thats a fucking car axle
there is another type of silencers which are way bigger but i wont talk abt them because hmm
scale scale scale scale
even the scrappy gun from the day of the jackal would be more useful
idk
this thing is literally anti-material class
weight 7kg
ill prob get it threaded the price isnt that much
benefits of living in asia ig
just watch some show in the meantime
oh effective distance 600m actually
x2 compare to val/vintorez
2+ seconds fly time
lemme see my cpu usage
that.... makes it useful
let me not
rust analyzer each time i go to std source files go brrrrrr
i will fix it myself at some point
i just have to make it return when it get some path and compile it from source again
it is not 100% yet
plz remove this toolbar useless things that kde decided to add for some reasons
i just cant look at them
make me sick
nerd
happy?
yeey

is that a question
idk maybe i can squeeze out those last 3% on each core
doom it is then
doom, great idea
Taking a detour before test4.
Test3.5 is testing conditional depth cache and mm encoding reuse
If both tests test3.5 and test4 succeed, my concept is successfully validated and I can begin actual trials
way faster, thanks claude
Day 67 of making my own game engine we are taking brake for 2 weeks for a hackathon stuff
nice gl!
This is farther one of my ideas has ever gotten, very excited
The space station 14 devs have banned the word clanker 
Wouldn't bat an eye elsewhere but was caught off guard given how anarchic ss13 is
Good humans predictable as ever
fml i forgot to turn off unlimited extra usage for Claude
How much $ did you burn


So I've been dealing with this issue in TF2 for about 2 weeks now: https://github.com/ValveSoftware/Source-1-Games/issues/8000
I was playing TF2 just now and decided to look into the journalctl again when the freeze occurred. Amongst some sort of window state spam, there was one entry I missed all this time. An innocent entry from systemd-resolved saying system time change has been detected and caches have been flushed.
So I know what was causing it. The fucking NTP client doing regular time syncs. Every time it happens, the game freezes. Every time I manually roll back the system time by even a second, it happens. Because of time sync...
Source spaghetti at its finest 🍝
wait i just realized that was an invoice because i switched from monthly to yearly payment
my extra usage was only 12€ lol
false alarm 
it did gave me a panic attack
fe
iv got to do some work on making it so the UI for this thing im working on isnt, terrible
So
fun times
I now reached 119 pages on my report and i'm only like half way through
This looks stable
hi
I am going to try to learn lua
and will probably be asking a lot of questions
first question
how do I run a program
do I have to use the full path, or can I just type the name?
ah apparently I don't have lua installed
wonderful
fullpath if not on the same dir, name if you are on the same dir
b r a i n s l o w l y d i s i n t e g r a t e s

IIRC CAPE sandbox recommended it
The flavor of the VM runner is relatively irrelevant to me
Even though it’s reported to run on other operating systems too, CAPE is originally supposed to run on a GNU/Linux native system. For this documentation, we chose the latest Ubuntu LTS as the reference system for the commands examples.
The VM with CAPE is being ran on Win11 host anyway
why are you building plasma
Could be a plasma dev for all we know 
can you not use ../ to back out of the dir and execute lua scripts that way?
relative paths banned 
bruh
When i read CAPE my mind went
HTB Certified Active Directory Pentesting Expert
I gotta continue writing my CPTS report 
problems with cachyos kernel build
and nix
To be fair, path traversal is its own can of worms
You DON'T want to do that whenever possible
true
Now I'm wondering if there exists a universal binary
That somehow tricks windows into thinking it is an executable, but when it is brought to linux is also an executable
Time to FAFO 
It is in fact, not possible 
depending on how much you want to bend rules wsl already supports this

through binfmt_misc
Naah, that's too bending it
Also a fun coincidence that the blog name is Justin(e) 
It would be fun if we had those tho. Web apps are no longer necessary
i've been meaning to play around with the cosmopolitan libc
So it is a WIP. The madlad 
my pc doesnt seem to be overheating anymore 
thats just assembly
jar
py
pretty much most things
The file host I had been recommended is a paid one so now I need to find a new place to host NeuroSynth's model files
Anyone got suggestions?
huggingface
There's no vocal synthesizer category though
Not the same category of models though 
I already checked in the past and there's no labels matching singing voice synthesis
someone is hosting a forsen RVC model on there under the random ass tag "RVC" without problems i'm sure they wouldn't mind
Yeah those are
- TTS
- slop
- Whisper
- RVC
- Genre detection?
- Basically volume detection
And I don't exactly want to put NeuroSynth next to slop
It also doesn't seem to be the correct place to host a vocal synthesizer
someone is doing a series of megaman nt warrior models lmfao
All RVC
yes because that's the rvc category
We're intentionally pushing NeuroSynth as far away from RVC as possible
it's a reinforcement learning based model trained on a dataset presumably with a normal pytorch or similar method no? it belongs on hf if even most of that is true
What do you mean reinforcement learning
should've just said machine learning there
HuggingFace has no category suitable for SVS though
it does indeed have svs on ther already
It's buried in the "other" category for some reason
Why the heck is it there and not in "tasks"
because the task is text-to-audio
Text to audio is song generation
SLOP
i mean, by definition, it fits svs and is in some form another what neurosynth does as well
whatever your personal boundary is for what is slop and what isn't notwithstanding
SVS is not even close to "text to audio"
There's a billion other things that go in and it isn't exactly even really text that goes in from the text you type
It's more like phonemes-pitch-tones-to-audio
the categories are extremely broad strokes
Even text-midi-to-audio would differentiate it enough
text-to-audio is literally just "You input text, the model generates a song"
i mean it's significantly more than that if you look at the models
So unless we find anything better I guess HF is an option after all
the categories are broad
It's all slop
pretty much any model on hf you could nitpick and find a way it doesnt necessarily fit exactly the main category, tagging is about braod strokes it's not like they have you on oath
you dont even need to select a category
you can also just tag it with the training backend and leave it at that for the lulz
hell you could tag that bitch kotlin and nobody would care
vocal synth should be under text to speech
if u give it text and it makes speech
TTS is completely different functionally
then
do u give it text
and it makes speech?
I give it notes with phonemes
then its a vocaloid
idk put it under synthv
Way closer to Vocaloid than TTS
I don't think there's any open-weight SynthV vocal synths
just more advanced settings on its "text"
it's just text broken into smaller bits
superset
internally there isnt much difference, alot of the sounds are just handled for u with traditional TTS
this guy is just holding his dataset publicly here
but yeah it's still just text to audio if you want to be more broad

Not even close
a TTS is essentially a reverse synthV
Are you intentionally trying to annoy me?
So you are intentionally annoying me
I would very much like to not be intentionally annoyed
So anyway as we've established there is a well hidden but existing SVS tag on there so it should be a valid fallback if we can't find anything better that doesn't have the Git baggage that I don't care about
TTS not a reverse SynthV. TECHNITALLY you could say that would STT. but really it would be Singing to Midi or something similar
depends on which tech ur specifying
Technically SynthV comes with reverse SynthV preinstalled
hf is just git lfs right
TTS is kind of a blanket for like 4 different types of techs
buckets are xet based
And are they a simple GET request away?
Storage Buckets are a repo type on the Hugging Face Hub providing S3-like object storage, powered by the Xet storage backend. Unlike Git-based repositories (models, datasets, Spaces), buckets are non-versioned and mutable, designed for use cases where you need simple, fast storage such as training checkpoints, logs, intermediate artifacts, or any large collection of files that doesn’t need version control.
s3 commands
TTS is Text to Speech. So setting aside semantics TTS and SynthV would be in the same class. A REverse SynthV would be a minimum SPeach to Text.
good thing there are a bazillion s3 clients
you literally have your terminology backwards
And are those implementable entirely on the web side in such a way that it just feels like downloading a file with a GET request?
i think i just worded it like ass, im referring to alot of tools breaking down sentences into phonemes for the program then to play the respective soundbites for
Remember that you gotta be able to make it on the website
THen TTS and SynthV (which is SVS) both do that in a sense
You're literally the web dev
exactly
super, s3 is THE storage api for the web
one just removes a middle man generally
















compiling Linux

