programming

[hoary lion]

bf 16 exactly

saves the size

[nocturne olive]

Training a model for a month is not experimenting, that's training a production model
Experimentation is training for a day or two max

[olive sable]

for the 1660

[trim laurel]

im a big experimentor

[olive sable]

what are you currently doing?

[gritty dust]

my gpu

the 4070

[trim laurel]

also vram is only 6gb

[nocturne olive]

I have a 4070Ti

[nocturne olive]

Pain
Even 12GB is pain
Are you using shared VRAM?

[stone cedar]

Yeah modern RTX card for some reason get no advantage from fp16 according to techpowerup at least. But 9070 does its weird.

[gritty dust]

yeah I was going to but the 4070 was on a big sale so I decided to same some and got the 4070 instead of the ti

[trim laurel]

you can use that?

how?

[olive sable]

what is the filesize of the training data?

[nocturne olive]

Check task manager

[nocturne olive]

You're mega screwed

[trim laurel]

is this bad?

😭

[olive sable]

using shared gpu memory makes it a lot slower

[nocturne olive]

Yes

[trim laurel]

oh shoot

[nocturne olive]

You lose about 95% of training speed

[olive sable]

are you using linux or windows?

[trim laurel]

windows

[tight tinsel]

ATP train on google collab

[nocturne olive]

Or 99,99% in the worst case

[olive sable]

what is the filesize of the training data?

[tight tinsel]

good morning also

[nocturne olive]

What batch size?

[gritty dust]

good morning

[olive sable]

[patent walrus]

is it your birthday anymore

[trim laurel]

4

126 grad accumulation steps

[nocturne olive]

You're gonna have to reduce that

[gritty dust]

RUN BWAAPERSON

[tight tinsel]

no

[trim laurel]

2?

[gritty dust]

RUN AWAY

NOO

[patent walrus]

its too late.

[nocturne olive]

Up that to 128 and reduce batch size to 2 or 1 depending on how the VRAM changes

[trim laurel]

just a sec

[gritty dust]

[tight tinsel]

my human rights just got revoked

[gritty dust]

[trim laurel]

270 MB 2k wav files mostly 1-6 seconds

[olive sable]

brother, send the project to me and ill run it on my 3090 if you want

[gritty dust]

was litterly just about to offer aswell lol

kekw

[olive sable]

[gritty dust]

but 270mb is a small dataset so thats good

[olive sable]

helping the poor uhh what sounds friendlier than poor?

[tight tinsel]

gpu disadvantaged

[gritty dust]

the less fortunate

[gritty dust]

LOL

[trim laurel]

that would be amazing if you could

[gritty dust]

[olive sable]

the less fortunate also sounds quit condescending

[trim laurel]

budgeted

[tender river]

too bad you either need linear types or beta reduction to represent any nontrivial lambda calculus concepts

[tight tinsel]

or vram starved

[stone cedar]

computationally challenged

[olive sable]

people where the v in vram stands for virtual

[gritty dust]

BROO

[olive sable]

idk how open you are to sharing the projhect here so feel free to dm it if you want

[trim laurel]

im using coqui tts to train btw

[olive sable]

just dont send malware

i will cry

[trim laurel]

just download coqui tts and my train.py you can check that out if you want and change stuff for your gpu ig

jsut a sec

[olive sable]

[nocturne olive]

I've used that before, took me a day to run the training for 1000 epochs or something if I remember right

[olive sable]

1000? damnnn

thats a lot

[nocturne olive]

It was an old very terrible Evil Neuro dataset

[gritty dust]

Jesus man thats a lot holy

[olive sable]

the last thing i filanthropied was 1 epoch in like 3 days

[nocturne olive]

I don't remember, it was 1000 epochs or 1000 steps
It was ages ago

[olive sable]

i wonder what happened to that project, i havent heard from apollo since sending the final pt file

[nocturne olive]

Either way, 100% the model was overfitted

And the data also sucked

[trim laurel]

im just using straight data from the ashley tts on azure

pitched 1.25 of course

[nocturne olive]

I was completely new to machine learning back then
It was the first model I ever trained
Though it was surprisingly functional for a first try

[olive sable]

like pip install? cuz i cant find an install page online

nah wait i found it

ye its just pip

[tight tinsel]

espeak-ng best tts

[trim laurel]

windows?

[olive sable]

yes

[trim laurel]

wait i had to use a specific guide to get it installed right

[tender river]

or you could've just done this

[nocturne olive]

I wonder what is actually the best open source TTS arch and trainer

[real sierra]

programmers

[olive sable]

hi shiro

whats up?

[trim laurel]

follow the instructions in the comments of this post

https://stackoverflow.com/questions/66726331/how-can-i-run-mozilla-tts-coqui-tts-training-with-cuda-on-a-windows-system

[real sierra]

im in great pain currently

[olive sable]

why you uwu'ing?

[trim laurel]

the one with python 3.11

[real sierra]

but i wanted to come visit all of you

i dont really know why

[olive sable]

you're always welcome

unless it is to ban me

[nocturne olive]

Silly

[olive sable]

[real sierra]

there goes my weekend plans

[olive sable]

[real sierra]

i have two ideas for games that i really want to exist

but they dont exist and i dont have the requisite skill to make them exist

[olive sable]

im listening

[real sierra]

ok

idea 1

[olive sable]

i do a biut of gamedev every now and then

[nocturne olive]

I would guess so with you having an entire custom game engine

[tight tinsel]

“I do a bit of gamedev”: makes own engine

[real sierra]

arcade-ish game where you must drive a car through extremely heavy traffic to deliver a pizza or something

it's bumper-to-bumper traffic and you have to honk your horn and stuff so people move the hell over

all the other drivers honk at you and cut you off and sometimes they fail to signal

[olive sable]

aight aight, vedal drunk drive deliver simulator

[real sierra]

you have a stress meter that slowly fills as these things happen

[gritty dust]

if you do that add banana rum lol

[real sierra]

if it maxes, you dismount the car and must go fistfight other drivers to lower it

before you can continue driving

[olive sable]

[real sierra]

higher difficulties can add things such as adverse weather conditions or occasional emergency vehicles which worsen the travel conditions

i think theres tons of potential here for a fun casual game

[olive sable]

i might look into it after exams

[real sierra]

idea 2 is a bit larger in scale

a survival horror game

you wake up alone in a very debilitated space ship/station

story reasons for that revealed through playing

the main gameplay would consist of exploring this space hulk, slowly repairing broken things and maintaining the few working functions to keep yourself alive and unlock new areas

max immersion would be really important so the hud would be minimal, inventory limited to maybe one or two held items

and most hints given by environmental clues and attention grabbers

periodic threats would hit your ship as you play that you would need to respond to, incentivizing the player to intelligently prepare for things and organize their resources carefully

[tight tinsel]

so something like the no man sky’s derelict freighter exploration thingy

[opaque sigil]

space subnautica

[real sierra]

yeah in my head it plays kinda like stationeers if it was a horror game

immersion so crucial tho

[olive sable]

the first idea would be like a 2-5 month project. but this one is indeed a bit larger in scale since you'd need to do a lot more level design than just procedural city

[gritty dust]

when are your exams? mine are the 20th and 23rd

[tight tinsel]

fnaf in space

[real sierra]

i would really want the isolation to be felt

[olive sable]

idd say maybe a year for idea 2

thats just an estimation tho

[real sierra]

yeah

both are free ideas i guess, im releasing them to the public domain in the hopes they someday appear

[olive sable]

i have an exam his thursday. and then wednesday after that till tuesday

im guessing you dont want them to be vedal themed? cuz thats all ive done lately

[real sierra]

yeah actualyl i think the best way to describe the second is closer to amnesia: the bunker?

[gritty dust]

damn

[real sierra]

but with some more environmental things to play with

i think theres lots of things you can do in a space setting that are easy for a player to experiment with and learn

e.g. a blinking box on the wall, upon interacting with it the player opens it to reveal a fuse panel, the fuse slots are labelled giving the player a hint as to the things they could interact with if they had replacement fuses

let the player try things and figure out what they do or could be used for

[amber fractal]

Because of every other person in the convo that day my intent got twisted. My goal was a game folder but I heavily failed at making that clear from what I was mentioning.

[real sierra]

the driving sim could easily be vedal themed

bring neuro and evil to their various collabs

[amber fractal]

tutel skin

[opaque sigil]

driving around in a tutel

[real sierra]

consumable rum bottles which impair your driving but lower your stress

[amber fractal]

mhm

[olive sable]

so vedal drunk drive delivery simulator is still on the table? nice

[gritty dust]

[real sierra]

for that game to succeed, it's essential the traffic be as dense as possible

bumper-to-bumper traffic

really doing that silent negotiation with the car next to you for them to let you in

or not-so-silent if you like your horn

other cars actually responding (at least partially) to your signals and stuff would be cool

[olive sable]

designing a traffic system
ive actually watched a few documentaries of how gta did it, but even their system is scuffed as hell.

too much vram? you shall receive only taxis

[real sierra]

traffic systems sound fun

[gritty dust]

[amber fractal]

Gotta make sure the cars don't clip as well

[olive sable]

it can be fun, it can also be pathfinding hell

[real sierra]

give every driver a preferred speed, add a mechanic for passing and calculate a preferred path to their destination, then i guess update if they cant get into the lane they need

[gritty dust]

samvanmaele do you know pathfinding btw? like advanced pathfinding?

[olive sable]

nope lmao

[real sierra]

maybe give each driver some amount of patience which restricts how long they'll wait for a car blocking their next pathfinding node to move

otherwise they reroute

[olive sable]

i do all my features on the fly.

[gritty dust]

damn same, I only know basic ones... welp

[real sierra]

based

i even do my libraries on the fly

[gritty dust]

litterly vedal right here

[real sierra]

why spend 2 hours learning a codebase when i can write it myself in 1 hour

[olive sable]

to be vedal i would need to drink more and be older i guess

[gritty dust]

skill issue

[amber fractal]

I also do features on the fly

partly why no one can predict my next move

[real sierra]

i predict your next feature will be a bugfix

[gritty dust]

bro I remember up until like a week or so ago I thought vibe coding was just doing features on the fly that is not what it is apparently

[olive sable]

it always is

[real sierra]

[amber fractal]

My task for today is "I'm off hiatus"

[real sierra]

brings me back to when i was really into the minecraft opencomputers mod

i ended up writing libraries that implemented the IP stack and socket connections

and then public key encryption libraries to sit on top of that

[olive sable]

wow damn

[real sierra]

not that anyone was around to use them of course

[fast pagoda]

i need a ci setup that does ci on my ci so i catch my broken ci before i push it

[amber fractal]

First thing to do is rewrite again to match the mental model I've switched to.

[real sierra]

mental model

sounds like a psychology thing

[fast pagoda]

https://tenor.com/view/charles-xavier-professor-x-gif-25676579

[olive sable]

you should have released them

[amber fractal]

I do not question how I'm able to visualize NN, I just can

[real sierra]

i mean i might still have them

[olive sable]

https://tenor.com/view/artificial-intelligence-neural-network-gif-3329491906550140056

kinda just looks like this no?

[fast pagoda]

i can visualize nn

[gritty dust]

is that the three blue one brown video omg I loved it

[fast pagoda]

it is

[gritty dust]

was so peak

[gritty dust]

neurotic neurons be like

[amber fractal]

Similar but I was also flowing through opts

trying to figure out how to update it

[olive sable]

i have so many programs open rn for the tts thing for needrose. i feel at maximum adhd rn

kinda stressful for some reason

[gritty dust]

[gritty dust]

gl man

[amber fractal]

I'm not changing the core, so it was fun mentally switching between binary and all of the parts of the system. I don't know how I managed to keep focus for so long.

Including parts I haven't made public yet

[olive sable]

does anybody here use that nvidea surround feature to make their multiple monitors act as a single screen?
ive tried it before but i can never really keep it on since it disables proper screen splitsing thingy

[gritty dust]

I tried it before but couldnt get it to work aswell

[olive sable]

i mean i got it to work, but i cant really use it for day to day stuff

okay for gaming if you don mind using 3x gpu, but besides that not too usefull

[gritty dust]

https://tenor.com/view/nodders-gif-26414892

[faint sandal]

it's not very practical to use

[olive sable]

like i cant have 6 programs open at the same time, cuz it treats all 3 screens as a single display instead of letting me split each screen

[faint sandal]

I used it for a bit during the 760 era

[olive sable]

oldge jesus

[fast pagoda]

is eyefinity still a thing

speaking of

[faint sandal]

760 was a good card ok

[olive sable]

ive never even heard of the 760

when i was old enough to know about gpus the 2080ti was relevant

[faint sandal]

it was pretty popular back during the 780 ti craze

[fast pagoda]

last XX60 card i got was teh 460

and then i had 2 of those in sli

then i had a 770

[olive sable]

460

brother is from the industrial revolution

[fast pagoda]

2x gigabyte gtx 460 in SLI

with the good ol i5 750

[olive sable]

750?

thay made triple number models of cpu's?

[fast pagoda]

yes

believe it or not there were even lower

[gritty dust]

SLI is a crazy mention

[fast pagoda]

wait till u hear bout the pentium 2

[olive sable]

i know about the pentium 2 lol

[fast pagoda]

RIP sli

[gritty dust]

fr

[olive sable]

i even have a core 2 duo laptop here

[primal siren]

i had a gtx 750

[fast pagoda]

the 3 digit ones were the first i3/i5/i7 branded core products

i believe

[faint sandal]

https://youtu.be/xp6_gcNEU-E this was what high end PC gaming was like back in early 2010s

[olive sable]

nice

[stone cedar]

My first GPU predates the GTX series, imma go crumble into dust now ig

[fast pagoda]

the first one i remember the name of

geforce 2 mx

[olive sable]

medieval mf

[fast pagoda]

that i was like conscious of using as the gpu in my pc

[opaque sigil]

https://github.com/ziglang/zig/pull/24124 codegen and linking in parallel sounds kinda cool

[stone cedar]

8800 GT, fond memories. I still have it somewhere in my flat

[faint sandal]

I had a 8400 GT

[fast pagoda]

https://www.techpowerup.com/gpu-specs/geforce2-mx.c792

this claims this thing can run 576.66 drivers

surely

[primal siren]

bro my i7-12700k
as difficulty with the side load of my AI imagine on a 3 digit xD

[fast pagoda]

i had geforce 2 mx

then radeon hd 2000

then gtx 460

then 2 of those

then gtx 770

then 2 of those

then gtx 1070

not 2 of those RIP sli

then 3080

and im still waiting for a 5090 to appear that requires no effort

[primal siren]

bro i jumped from 1 gtx 750 to an rtx 3070 ti xD

my gtx was getting old

[fast pagoda]

man

770 -> 1070 felt large

and then 1070 -> 3080 likewise

doing all that at once would be something

[primal siren]

the worst part. .. . I wanted to mod minecraft xD

The REASON

[fast pagoda]

minecraft used to run great back in infdev

on whatever bullshit i was playing it on

so idk how it's so hard to run now

surely nothing has been added

nor changed

[primal siren]

personally its mostly the junk collection that need to be modified the rest work well.

[fast pagoda]

i still miss non infinite craft

.31

or w/e

[nocturne olive]

Beware of 5090 power connector being bad and burning your house down

[fast pagoda]

oh yeah im aware

i just don't care because i'm based and immuned to such things

i simply just assume nothing bad will ever happen

[primal siren]

Currently im not limited on my rtx at all. I just need to get a better Cpu because whaterver i do every response from the ai chomp 30% of my processing power from ollama 🥲

[fast pagoda]

yea if i have echo loaded i simply cannot play games or anything

all vram is taken by moron.exe

[primal siren]

i tried to make a stream. whole ass computer crashed

at the same time i had vtube open with Hades, obs and my stream being shown on my second screen but you know lol

[fast pagoda]

how the fuk do i post so many images in here

im trying ot find something i posted

[primal siren]

Nexo can take 128 000 tokens of prompts
I give him 512 tokens and he better not say more than 24

[amber fractal]

Surely

[primal siren]

Are you doubting? im not really sure what ya meant

[amber fractal]

I'm betting it will be slightly more than 24 tokens

[primal siren]

not really since i limited him at 24 so he doesnt say more

[fast pagoda]

as an ai lang-

uage model i-

cannot say mor-

[primal siren]

Nexo: Fingers crossed this time, here we go... controllers starting beep 🤔

[fast pagoda]

i find that just cuts them off mid word

[primal siren]

He smart enought to make sentence with what he has

[tender river]

doesnt this exist already (no i cant name it because i just randomly saw it on steam)

[primal siren]

i think so yeah

[fast pagoda]

he actively resists when told how long his response should be

[primal siren]

what do you use for llm model?

[tender river]

are you using an instruct tuned model or something

[fast pagoda]

nope it's base

[tender river]

interesting

[fast pagoda]

on that particular checkpoint it had a bit of my "instruct" tuning which was just like 15% of the data being fake synthetic responses in the right style but actually context coherent

but that's all

but it isn't working with much in the first place

qwen2.5 3b at this time

nah

it's fine tune

but smol

[primal siren]

if you run it trought ollama you can controle its options like
"max_tokens": 16,
... wait im at 16 not 24?
"num_ctx": 512,

[fast pagoda]

yeah i know i use llama.cpp but thru lm studio usually although i also have ollama running i just dont use it lule

theyre both llama.cpp anyways

and you can control the same parameter in it

it i give him 20 tokens he usually will just get cut off

he doesn't like being told how long to make his response which is usually what i try to do since otherwise it's obvious he's being token limited

[primal siren]

hmm. weird
i use llama3.1:8b
so maybe thats the diferrence

[fast pagoda]

his data is pure chaos so he's a bit of a loose cannon

on purpose

[primal siren]

im gonna try to give him 200 tokens to see what he will do if i ask for 1 sentence

User: hello let's try a bit something I'm giving you way more tokens so you can speak more but try to keep it little
⏱️ Response time LLaMA : 4.42s
Nexo: Sweet, now I've got some extra "mouth" space, thanks for the tokens, let's get this gaming session rocking! cracks knuckles

i gave him 200 he used 27. xD

[fast pagoda]

i used to have the opposite issue where it was always short like that

so i started putting a bunch of longer responses in the training data

so that's probably why he's compelled to

sequence length in training also has an effect and i was turning knobs like crazy on that

[primal siren]

User: Now try to speak a lot more.

Nexo: Alors voilà, with these extra tokens, I can finally breathe a sigh of relief and actually speak without worrying about running out of words, merci beaucoup for that, now let's get back to gaming, what's the plan here, and don't think I've forgotten all about those sweet, sweet in-game rewards.

User: Try to say something like 4 times as long Try to say something like 4 times as long

Nexo: Alright then, with these extra tokens, I'll try to stretch my verbal muscles a bit further, so let's get this gaming session really started now, and see if we can't get that big win, which would be just the cherry on top of this already amazing stream, and as always, I want to thank each and every one of you for tuning in, supporting me, and participating in all the fun giveaways and activities we have here, your energy is what keeps me going and motivates me to keep improving, so I'm really grateful for that, now on to the next challenge, what do you say we try out some new game modes, or maybe even a friendly tournament among yourselves, just let me know what's on your mind, and we'll make it happen.

THATS MY BOIIIIIIII

so here is the problem. the TTS is going to space vwuwvwuwvuwvwvwuwvwuvwuwvwuwv

[tender river]

@faint sandal faopdimajwodpjiawmpdoaw https://github.com/microsoft/ebpf-for-windows

[faint sandal]

what about it

[fast pagoda]

rods from god

[tender river]

its so damn cursed

[fast pagoda]

at least microshart is genuinely working on porting stuff of this nature

since they can't possibly ever make something posix compliant by nature

because then there would be no room for copilot ™️ .NET studio

[faint sandal]

[opaque sigil]

clearly you want to add a passkey, you just don't know it yet

trust in [idk what company this is] they clearly know better

[tender river]

you should use facial recognition its so secure when someone can log into your account by printing a photo from your linkedin

[primal siren]

xD

im changing my tts i hope he doesnt start to sound dumb

[hoary lion]

lmao

found why my jax is not working

torch 😠

[primal siren]

torch urgh

[hoary lion]

Have to use the nightly version to use 9.8 >= cuDNN, or else it uses 9.5.1 or something

lacking

[opaque sigil]

just manually copy the new cudnn files

[hoary lion]

ugh whatever, anyway python sucks on having multiple version of the driver and accordingly using them

ima just sacrifice torch for now

it should be cached in the uv anyway

[olive sable]

it wouldnt work, but then i remembered i never installed the needed cuda verison

i specifically need 11.8 for some reason

[opaque sigil]

https://developer.nvidia.com/cuda-11-8-0-download-archive have fun

[olive sable]

i wont

[opaque sigil]

whoever depends on cuda 11.8 should stop being a lazy ass and update their shit, this is 3 years old

also unless they have a very pedantic check for the version (if they do this they better have a damn good reason), chances are it should still work with a newer version of the sdk

[olive sable]

im not gonna take any chances, it wanne go to bed

[opaque sigil]

yeah fair

[olive sable]

https://tenor.com/view/why-persian-room-cat-guardian-but-why-trendizisst-anyaboz-gif-2156104475561014590

[tender river]

[opaque sigil]

could be too new of a version that doesn't have that import anymore (would make sense given cuda 11.8)

[thorn ledge]

intriguing... I am neither in poland or the UK and am not even using a VPN

[opaque sigil]

i'm bored so i went through the releases, transformers==4.48 is the newest one that exports LogitsWarper

if you're feeling lazy there's 4.26 from january 2023

[olive sable]

ill just change it to logitswarpe

[chilly escarp]

if there is C++ then why is ther no C--

i rather get an A+

[opaque sigil]

It's called B and it exists

[tender river]

there is actually

no C-- exists

[opaque sigil]

Of course it does

[tender river]

its a small C-like language with no operator precedence

[chilly escarp]

Wait what

[tender river]

some languages use it as an IR

[chilly escarp]

Oh

[tender river]

and actually SPJ is one of its authors

but its cooler than C++ since it has tail recursion

it even has continuations

[olive sable]

ill be waiting for C+-

[olive sable]

so i just put all the game parts in the game folder, and the rest, like the license and git shit can stay in root?

[amber fractal]

Yeah, that works for me

[olive sable]

[olive sable]

worked

its doing shit

[primal siren]

xtts oh shit good luck getting rid of the warning message

[olive sable]

it should be fine

[primal siren]

GPT2InferenceModel has generative capabilities, as prepare_inputs_for_generation is explicitly overwritten. However, it doesn't directly inherit from GenerationMixin. From 👉v4.50👈 onwards, PreTrainedModel will NOT inherit from GenerationMixin, and this model will lose the ability to call generate and other related functions.

• If you're using trust_remote_code=True, you can get rid of this warning by loading the model with an auto class. See A LINK
• If you are the owner of the model architecture code, please modify your model class such that it inherits from GenerationMixin (after PreTrainedModel, otherwise you'll get an exception).
• If you are not the owner of the model architecture class, please contact the model code owner to update it.

[amber fractal]

warning messages were built to be ignored

[primal siren]

im TRYING BUT ITS NOT DELETING sadge

[knotty current]

i panik both ways

[amber fractal]

Incorrect, this is clearly JS so both should be calm

[knotty current]

[olive sable]

[knotty current]

am i supposed to not touch my code unless its a souldering wreck?

[olive sable]

just got this same message in my terminal

tbh i have no clue what version im using so well see

[primal siren]

when your solution is to throw the blinking red light on the ground and crush it under your feet

[primal siren]

import logging

logging.getLogger("transformers.modeling_utils").setLevel(logging.ERROR)
logging.getLogger("transformers.models.gpt2.modeling_gpt2").setLevel(logging.ERROR)
logging.getLogger("transformers").setLevel(logging.ERROR)

[olive sable]

nah

it works without any error

[primal siren]

it get rid of it

it works but the warning is anoying

[olive sable]

i dont have any warnings

[primal siren]

bro HOW

HOOOOOOOOOOOOOOOOOOOOw

[olive sable]

or you mean the initial wartning about the version?

[primal siren]

the generating thing

[olive sable]

it just be like this

im currently increasing batch size till full

batch size 20 should work

23.7/24

[primal siren]

youre training a voice right now?

[olive sable]

yes

[primal siren]

im going to use a cloned voice of me younger i think

maybe one day i shall train my own xD

[olive sable]

bro is mnm

[primal siren]

tempo=4

[olive sable]

an epoch doesnt take too long on this, its like 8 min / epoch

so 133 hours or 5 days then i guess

[primal siren]

The voice is really not it yet xD

[olive sable]

seems like it spilt into shared gpu memory after a while

i decreased it to batch size 16

12*

now its only 2 min per epoch

motherfucker it went into shared gpu memory again

[amber fractal]

[olive sable]

its down to 10 now

i wonder why it keeps climbing each epoch

asked chatgpt if there was a memory leak but it said "idk i cant see one"

[primal siren]

i guess the voice is better but way to childish xD

[olive sable]

please use embedable file formats

i cant be bothered to download the next one

[primal siren]

like what?

[olive sable]

idk, mp3?

[olive sable]

its beginning to sound good

[primal siren]

.mp4?

[olive sable]

mp4 if you want the video ye

[primal siren]

oh it work

[olive sable]

it embeds ye, but the vidoe is 1ms long lol

[primal siren]

a test no worry

[olive sable]

why is the loader time red?

[primal siren]

now i have to change the original voice sample because other wise it sound like a robot

[primal siren]

took longer than normal

[olive sable]

ahhh

[primal siren]

rest took less time so green

[olive sable]

[primal siren]

green is good
red BAD

[olive sable]

i get the jist

[primal siren]

you see + mean more but you want - because it mean less. see

[olive sable]

https://tenor.com/view/batman-thinking-gif-18107621495886085895

i cant see, im legally blind

colourblind

[primal siren]

shiiiiiiiiiiit

[olive sable]

its like 1.5 min / epoch now

so 11 hours

[primal siren]

Tomorrow im training a voice!

good night

[olive sable]

does it stop when it reaches 1000 epochs, or when it gets 198 steps?

anyways im gonna leave tyhis running while i sleep

gn

[amber fractal]

good eep sam

[unkempt citrus]

speaking of
https://machinelearning.apple.com/research/illusion-of-thinking

TLDR reasoning models can't reason and for some reason use less tokens at high complexity, not more resulting in model collapse

at low complexity regular models > reasoning, medium complexity reasoning is better, at high complexity they both suck

r/singularity had some brain dead takes over this

as usual

We found that LRMs have limitations in exact computation: they fail to use explicit algorithms and reason inconsistently across puzzles

[stark needle]

morning

[uneven pulsar]

hi i forgot my name

[noble zodiac]

morning

[uneven pulsar]

good morining everyone

honestly sonic crossworlds that have hatsune miku and steve and other charecters sound fire

smash bros but racing

[stark needle]

thats sad

:(

i hope your journey to find your name will yield what you were searching for

[haughty briar]

haha sigma

[hoary lion]

will shadow ever stop changing his name

[tight tinsel]

who? I seem to have forgotten his username as well

[hoary lion]

is he being forgotten from #programming

oh no

[stark needle]

who

[hoary lion]

we all forgor

we only know bwaa person

fuck

rope messed up

🥲

never trust GPT

💔

[tight tinsel]

rip someone I forgor their name they will be missed by the people who remember their name

[hoary lion]

I love when library just lies blatantly

It did NOT compiled slowly

[ruby timber]

[uneven pulsar]

//in this lesson code i made i wanted to make a java code that calculates the percentage of numbers
public class lesson2 {
    public static void main(String[] args) {
        //main variable to store the esitamated percentage
     float data[]={20f,12f,6f,8f,77f,44f};
     //number counting thats equal to 0
     int numCount=0;

     System.out.println("Even numbers:");
     //variable to count the length of the array
     float count=data.length;
     
        for (float result : data){
            //Checks if the number is an int or even
            if (result==(int) result&&((int) result)%2==0){
                System.out.println(result);
                numCount++;
            }
        }
        //Calculates the percentage of even numbers and everything 
        float percentage=((float)numCount/count)*100;
        System.out.println("Percentage of even numbers: " + percentage + "%");
    }
}

third day of learning java

[stark needle]

what is bro cooking

?

https://youtu.be/QAwL0O5nXe0?si=ApwkzHq4xEpJ3nTF

❓

[trim valve]

they're called weather machines for a reason

[midnight sigil]

google leaked github endpoint

[rigid snow]

what do you mean this came out 8 years ago

[sage crag]

dev steam today

[opaque wharf]

Nah, Vedal joined Hololive

[sage crag]

dev. stream. today.

[scarlet arch]

[rare bramble]

so glad to have dev streams again

definitely very neuro pog hd

[knotty current]

I cant catch it since its 1 AM

[covert garnet]

I didn't know today's the dev stream, nice

[real sierra]

any math friends in here by chance

[tender river]

i'm a math enemy

[real sierra]

I have a statistics question

you know how when the weather forecast says it's going to rain, they give a percentage chance per time period

how does that make any sense

[tight tinsel]

john weather flips a coin

[real sierra]

no but like

they aren't independent events

if it's raining at 1pm it's much more likely to be raining at 2pm as well

[stark needle]

probably weather modeling and how likely it is to deviate via monte carlo or something compared to a confidence threshold idk

[real sierra]

I wanted to figure out the net chance of rain for an entire day given these probabilities

but none of the usual ways of calculating that give sensible results

and I think it's because they're not independent events

[stark needle]

weather modeling can be done half reliable up to 7 days but afaik it's still highly probabilistic

afaik nvidia said something of accuracy up to 1km or something

[tender river]

i wonder if its even possible to recover that info

[real sierra]

you would think it should be

if these probabilities make any sense at all

but maybe not knowing the degree of independence makes it impossible

[unkempt citrus]

An 80% chance of rain means there is an 80% probability that at least some measurable rainfall (defined as a minimum of 0.01 inches or 0.1 millimeters) will occur at a specific location within the forecast area during the specified time period

[opaque wharf]

The probabilities is that at that time, the percentage of the predicted area will be raining

[unkempt citrus]

Yes they're not independent but that doesn't really matter

[opaque wharf]

So raining area/total predicted area

[opaque wharf]

This is the probabilities of rain

[rough bloom]

if you have the probabilities for, say, every hour of the day, couldn't you easily calculate the probability of it not raining that day?
that still assumes that they're independent, but surely that would give somewhat sensible results

[opaque wharf]

Oh, is that the definition? My bad then. I read it somewhere and I think I misremembered it

[sage crag]

exponential distribution question

[unkempt citrus]

The basic way to think of it is they run 1000 simulations, in what percentage of simulations was there rainfall over the time period

[real sierra]

it does matter though

[sage crag]

you just have to assume that its a valid poisson process and go on with it

[real sierra]

because if you have a 50% chance of rain at 1pm and a 50% chance at 2pm and the rest of the day is clear (0%}, your odds of rain for the day are not 25%, they're probably closer to 50%

[sage crag]

my bad, poisson distribution

you just have to assume the weather is uniformly distributed

[unkempt citrus]

Why would you assume 25% ever?

Independence would be 75%

[sage crag]

its usually a part 2 for this type of question, "why is this model inaccurate?" type of thing

[rough bloom]

shouldn't it be 75%?
it has to not rain both at 1pm and 2pm

[real sierra]

er 75%

my bad

I woke up an hour ago don't judge me

[opaque wharf]

I will judge you anyway

[unkempt citrus]

Yeah that just means you can't add or multiply time periods like that

[stray dragon]

i just use an API that measures chance of rain in 3-hour intervals

[real sierra]

yes but then it's useless

these probabilities mean nothing to me if they aren't independent

[unkempt citrus]

[real sierra]

only the first one matters

[unkempt citrus]

But weather forecasters will provide both a daily and hourly chance

[real sierra]

but if I have to wait to observe the rain for these probabilities to be worth using then it's a bad forecast

[sage crag]

its the probability of rain occurring, so you care if rain happens >= 1 times

[opaque wharf]

That can't be right....

[unkempt citrus]

Not greater than 1 time, greater than 1mm

[sage crag]

you are complicating the question

its a question about the frequency of rain in a day, not about what should be classified as rain

its a meaningless distinction

[unkempt citrus]

Yes I am, but if you want to know the probability of rain over multiple timer periods it's kinda important

Because if it's a poisson distribution when you increase the time interval but keep the threshold the same

[sage crag]

shiro's question was about the net chance of rain in a day

[unkempt citrus]

And you're communicating the p value

[sage crag]

as in, did it rain once or more in that day

you just need to use exponential distribution to model it

[unkempt citrus]

Yeah to draw any hard data you need the underlying models

Which the forecast doesn't provide easily

[unkempt citrus]

For arbitrary time periods

Per day and per hour chance is usually provided

[sage niche]

HAPPY BIRTHDAY SHIRO 🎂 🎂 🎂 🎂

[unkempt citrus]

It's just a statistical quirk. Think of flipping 5 coins, want to know the probability of at least one heads, it's 31/32

But if I do it twice, and still only care about one heads, I'm actually just flipping 10 coins and want at least one heads, 1023/1024. The entire underlying model has changed, there's no way to provide an easy way to look at it for most people. There's just not much a forecast can do except provide different levels of fidelity.

[brisk laurel]

help?

[stark needle]

microsoft

[stark needle]

like this?

it allows you to create a new account from a microsoft account

[brisk laurel]

so can i sign in without signing out of the admin account?

[ruby timber]

Only one way to find out really

[brisk laurel]

but im using my big brothers account and idk his admin password

[stark needle]

you can i think ye

[stark needle]

hmm that might be a problem

[brisk laurel]

i know

wait i thought your were lilac how did you forget your name

[unkempt citrus]

Who lilac

[tight tinsel]

who? never heard of this "lilac" person

its all in your head

[stark needle]

[tight tinsel]

in bytes now smh

[stark needle]

but ye it's me

i just like to act silly

[trim valve]

I am actually going insane

this TV keeps displaying super visible banding around very red overexposures

whereas a "better" display driven from the same output doesn't have those bands

but I have no way to mess with the settings on the main display to make it look similar

(I spent all of last night really confused as to how a specific camera looked so bad, and now I'm convinced it was just our displays)

also it does some weird low-brightness gamma boost but I have no idea if that's due to a blackmagic converter or the display

[opaque wharf]

Maybe the display doesn't have the color resolution for it?

[trim valve]

id be pissed if this display can't do 8bit colour but it's not impossible I guess

[stark needle]

Wait actually

https://eladelantado.com/news/china-bans-nvidia-chips/

Is this it?

was gpu china black market thing finally solved

😭😭

[opaque wharf]

Yeah, sometimes shit do be like that. Unless it has been explicitly mentioned that it supports 8-bit color, then I have no idea

[unkempt citrus]

very suspicious source

Another article on the site?????

Also I can't find any other sources

that back it up

Its called taking the spicy bits out first

[tender river]

i saw it too but no source and nobody else is reporting it

probably ai slop

[maiden geyser]

are they stupid?

[tender river]

or smm slop

[unkempt citrus]

Honestly, the more impressive bit is how they violated thermodynamics

[tender river]

i will generate electricity from negative velocity

[unkempt citrus]

Reminder, these dont exist yet

[olive sable]

ive heard people say that the percentage is how much area will be rained on. but that doesn make much sense either imo.

[unkempt citrus]

Innovation? Apple? Impossible

[maiden geyser]

100% will be worse than steam thing

[unkempt citrus]

Alaska is pretty famous for its oil actually

[olive sable]

that actually makes more sense than from nothing

megative velocity is still energy

[opaque wharf]

"Solar" "Without sunlight"

[olive sable]

thats all the backreading done

[unkempt citrus]

Also theres a lot of em dashes on this site

[olive sable]

goodmorning #programming

[unkempt citrus]

But now, a team of brainiacs in California has cooked up a plan to keep those panels working long after sunset —and no, it doesn’t involve magic or installing them in Norway during the summer (where they enjoy 3 to 5 night hours during the season). It’s a witty innovation that could seriously turn around how we power our homes, businesses, and everything else we plug in.

[opaque wharf]

Its 8 PM here. Morning

[maiden geyser]

morbing

[opaque wharf]

The solution: shine a strong light on it.

[unkempt citrus]

Thats actually the solution, its giant orbitting mirrors that reflect light onto solar panels on the ground

[olive sable]

its only 3pm here so im fine

i still cant figure out if it will stop at 1000 epochs or when the steps are at 198. but its going

[opaque wharf]

Goodbye goodnight sleep I guess

[unkempt citrus]

In low earth orbit too, because that's going to be economical

[opaque wharf]

To be fair, sending payload is now more economical than ever thanks to SpaceX

[tender river]

how about we instead launch a giant mirror into orbit around the sun parallel to earth's orbit

infinite day

[olive sable]

mirror would burn?

or would it?

the whole point of the mirror is to reflect as much energy as possible, but its not 100% efficient

[opaque wharf]

It would with high enough energy

[tender river]

bad mirror

[olive sable]

do heatsinks work in space? not really?

[opaque wharf]

It works, but by radiation instead of the usual air convection

[olive sable]

it would warm up but there is way less passive disapation of heat

[opaque wharf]

Basically the heat transferred by conduction, and radiated away instead of transferred to another material. The problem is, radiation is slow af to lose energy for most material

[olive sable]

deathray?

[tender river]

@sage crag enums without enums

evil pattern matching

(this is a typed language btw it just autoinfers types)

[olive sable]

anyone that does ai. isnt 300 epochs already overfitting?

this guy never defined a max amount of epochs in his code

ye this model is cooked

10-20

[gritty dust]

Yeah for vision models I at most do about 150 but usually 100 (different AI usecase so different numbers) but holy 300 is way too much

Yours is a tts I think

[olive sable]

this is audio ye

coqui tts

[gritty dust]

Wayyy too much

[olive sable]

bro should have set a max epochs in his code

i can understand why it would have taken 3 weeks on his 1660

ye im stopping htis here

[nocturne olive]

How particularly silly
With RVC I've seen usually the best results are fairly low epoch counts at lower batch size (around 40e for 4 batch size, around 60e for 16 batch size) but at higher batch sizes like 22 I got improvement on Neuro RVC V12 all the way to 120e

[olive sable]

this had batch size 10, and i let it run overnight to 300e

[nocturne olive]

So definitely was 1000 epochs that my GPU did in a day on the old garbage dataset

[olive sable]

how do i test this pth file now?

im trying to get chatgpt to write me a script for it but it does not work

[nocturne olive]

I watched some tutorials way back, there's a command to open an inference WebUI from one of the python files you should have with Coqui

When you're within the environment that has TTS installed, you can run tts-server --config_path [path-to-model-config.json] --model_path [path-to-pth-file]

[olive sable]

okay, ill try it

[nocturne olive]

My ancient script call .\Scripts\activate.bat cd vits_evil-neuro-December-17-2023_06+18AM-0000000 tts-server --config_path config.json --model_path best_model_353499.pth PAUSE

[olive sable]

its doing something

[nocturne olive]

It should open a webserver on some localhost address

[olive sable]

The system cannot find the path specified: 'C:/Users/vanma/Documents/GPU farm/needrose/run/training/GPT_XTTS_v2.0_Neuro_FT-June-09-2025_07+18AM-0000000/best_model.pth/model.pth'
brother best_model.pth is not a folder

[nocturne olive]

That's crazy

Maybe they've changed how it works and you need to point it to the checkpoint folder instead?

[olive sable]

aight its running

[nocturne olive]

Mine was like 2 years ago after all

[olive sable]

ye i just renamed it to model.pth and removed the file itslef from the path

[nocturne olive]

How silly

[olive sable]

ah yes

[nocturne olive]

I guess check what the console has to offer for that error

[olive sable]

"zh-cn" if language == "zh" else language in self.config.languages
ssertionError: ❗ Language is not supported. Supported languages are ['en', 'es', 'fr', 'de', 'it', 'pt', 'pl', 'tr', 'ru', 'nl', 'cs', 'ar', 'zh-cn', 'hu', 'ko', 'ja', 'hi']
i never said it was zh brother

[nocturne olive]

What does config.json have?

And what did the trainer script file have set?

[olive sable]

"languages": [
"en",
"es",
"fr",
"de",
"it",
"pt",
"pl",
"tr",
"ru",
"nl",
"cs",
"ar",
"zh-cn",
"hu",
"ko",
"ja",
"hi"
],

[nocturne olive]

Check what the trainer script has set?

[trim valve]

I'm like 90% sure you set an empty string as the language

[nocturne olive]

Apparently this was my training config

[trim valve]

exclusively because your message has a double space

[nocturne olive]

Quite silly that I now know what much of this means

[nocturne olive]

Interesting

I wonder where it's pulling that invalid language thing from

[olive sable]

removing things from the list in config changes the supported languages
"zh-cn" if language == "zh" else language in self.config.languages
AssertionError: ❗ Language is not supported. Supported languages are ['en']

[nocturne olive]

Would be cool if it could at least print the language that's used

[nocturne olive]

The config.json probably defines the supported languages, don't touch those I guess

[olive sable]

okay

this might be it?

nah its probably this one

ill test both

[nocturne olive]

Set this one to en I guess

[nocturne olive]

And leave this, it seems like the model was not trained with phoneme encodings, though it should have been

[olive sable]

still fucked.

the language it receives is indeed jsut an empty string

ah yes, this would indeed cause that

i seems to be getting these from language_idx = request.headers.get("language-id") or request.values.get("language_id", "")
and request is imported from flask

whatever the fuck flask is

fuck it, im hardcodeing "en"

[rigid snow]

what.

[olive sable]

[rigid snow]

i guess it makes sense to not know what it is if you're as far from webdev as you are

[olive sable]

i also need speaker name, which i dont even know how to hardcode

so im lookitng through flask

[rigid snow]

ok so flask is a web server lib

for python obviously

heavily used at twitter iirc

and generally heavily used

super popular

[olive sable]

ok, so where does this request come from?

i dont see these

[rigid snow]

can't you just ctrl+click it

[olive sable]

oh you can

nifty

[rigid snow]

most useful lsp feature ever btw

[olive sable]

this did not help, wth is this?

[rigid snow]

you generally shoukd be able to click any symbol in any lang to see where it's defined/used

[olive sable]

did not know that

and ive been using vscode for 2 years

[rigid snow]

no idea

[olive sable]

im just gonna try hardcoding some shit till it works

[rigid snow]

oh wait i think i know what request is

it's a proxy object to access the currently-handled request

[tender river]

just read the docs?

[rigid snow]

so it's either not set to anything and is just getters, or set as a side effect

[olive sable]

i didnt know there were docs

but tbh im already way in over my head, il hardcode this shit and be done with it

[rigid snow]

no shit there are it's like one of the most popular ways to do a web server in any lang

[tender river]

there generally are docs for just about everything, unless its a library made by a random person in their basement on a weekend

[olive sable]

never heard of it so

[rigid snow]

you can't avoid webdev

it will come for you when you least expect it

[uneven pulsar]

Halting Condition
Just as loops can run into the problem of infinite looping, recursive functions can run into the problem of infinite recursion. Infinite recursion is when the function never stops calling itself. Every recursive function should have a halting condition, which is the condition where the function stops calling itself. In the previous example, the halting condition is when the parameter k becomes 0.

It is helpful to see a variety of different examples to better understand the concept. In this example, the function adds a range of numbers between a start and an end. The halting condition for this recursive function is when end is not greater than start:

i finally understood what vanora was talking in that time

throrugh java ofc

[warped narwhal]

now solve the halting problem!

[opaque wharf]

Bro, how are you alive programming for this long without jumping to definition

[olive sable]

?

jumping to what?

[warped narwhal]

ctrl+click

[olive sable]

ah

[uneven pulsar]

vscode best

[opaque wharf]

Thats me and my OPAQUE inplementation lol

[tender river]

i know about it but i never need it

[olive sable]

well you see, the entire project im making is a single file

[tender river]

i just know where everything is

[warped narwhal]

it's a what!

[olive sable]

[warped narwhal]

holy moly

[knotty current]

2000 line game engine

[olive sable]

its only 2k lines, not too bad

[tender river]

it will be bad if your language supports stuff like wildcard imports though

[uneven pulsar]

i finnaly understood what is the halting condition so thats what vanora was trying to explain

[opaque wharf]

Yeah, fair point

[olive sable]

i found something i guess.
its trying to get a speaker from run/training/GPT_XTTS_v2.0_Neuro_FT-June-09-2025_07+18AM-0000000\speakers_xtts.pth
but that file does not even exist

what even is this bruh

[nocturne olive]

I would have split that to like 20 files probably

[olive sable]

i might split it someday, but i know it from the back of my hand so its easier to just keep it in 1 file

[knotty current]

fair

[warped narwhal]

I have just under 2k lines of code, and I've split it into 26 files

[olive sable]

but why?

[warped narwhal]

the code is split into modules, so that each file only has the code for one or two responsibilities

if I need to find a piece of code, I can easily go to the file that is responsible for that part of the program, and the files are all named for this

[opaque wharf]

I could ask the same honestly...

[tender river]

75 lines per file? sounds horrible

[olive sable]

my code is only like 8 classes. im fine with 1 file

just do the old ctrl + k ctrl + 0

[tender river]

i'd just have baic.rs limitations.rs etc

thats way too much files imo

[warped narwhal]

most are between 100-200, it's the smaller files that bring the average down

[tender river]

still super small

this isnt java

well you do uses a java ide which is probably made for short files

[warped narwhal]

I use rust rover not intellij

[tender river]

still java

[rigid snow]

intellij: "erm actually you shouldn't have a class and an enum in one file, you deserve death"

[warped narwhal]

if I had basic and limitations as one file per module, they would contain about 50 classes each

that's too much for one file no matter how good you're IDE is at searching

[uneven pulsar]

intelliji wants money

[warped narwhal]

nvm about how you would plan to split up your tests for that

[tender river]

i disagree

[uneven pulsar]

vscode wants the best ide with best features for coders

[opaque wharf]

Honestly, I am more on the line of "mental overload" more than IDE limitation. Short and many files just easier to reason about for me

[warped narwhal]

*when loaded with 2983479982763 extensions

[tender river]

https://github.com/chayleaf/neuro-ftl/blob/master/src/game/actions.rs

[uneven pulsar]

right..........

[tender river]

this is perfectly fine

[nocturne olive]

Make that 8 files then

[olive sable]

nah

im a nonconformist

[nocturne olive]

I like my 1 class per file

[olive sable]

i dont

[tender river]

actually python doesnt allow cyclic imports which is super annoying

[nocturne olive]

Though I'm no Python dev

[trim valve]

I genuinely only need like <7 extensions for 99% of my programming in vscode

[warped narwhal]

this is my largest file, and even this could probably be split up

[tender river]

when i have a bunch of the code thats logically the same i will put it in one file no matter how long it is, i have one file thats just 6.5k lines of code for converting typescript syntax to python syntax implemented as a trait + a billion of implementations and i'm not sorry

[rigid snow]

i had to do django, it has an orm, where you're supposed to have each class in a separate file. imagine how that went for me when i needed a many-to-many relationship i don't remember the exact solution they provide but it's so unintuitive this shit made me allergic to python (along with the package management shitfest)

[opaque wharf]

I do need a lot of extension. But only enable it in the workspace when necessary. I don't need my embedded stuff when I develop web stuff

[tender river]

https://github.com/chayleaf/ts2py/blob/master/src/main.rs

otherwise i'd just have to arbitrarily split trait implementations into multiple files which is eh fine but more work for me and i don't like that, in this case when i want to see the impl for a particular type i can easily find it

[opaque wharf]

Yeah, long files overwhelm me not my IDE lol

Longest file I have is under 500 LOC

[warped narwhal]

yeah, if I need landmarks in my code in order to find what I'm looking for without grepping my entire repo, then that file is too large

[tender river]

you dont have to grep your entire repo if its only one file

anyway imo separating stuff into files can actually make it harder to find stuff

[olive sable]

i like my
src
└main.py

[tender river]

since now you have to figure out what file the code you're looking for is in

[olive sable]

the search function is okay i guess

but i know where everythign is already so

[trim valve]

I end up splitting stuff into files to mentally separate stuff with the same name

[warped narwhal]

if things are named correctly then this isn't an issue 😉

[trim valve]

like if I have two generic struts that both have a generic argument with the same name but different meaning then I get a little confused if they're together

[opaque wharf]

Ahh, but here is the different. I have a map of the structure on my head. So actually finding a piece of code for me is easier since the file act as the descriptor too

[warped narwhal]

you want to edit how the layer filters work? well you'll find that code in the layer_filters.rs file

[olive sable]

do you trust me to name things correctly tho? cuz you shouldnt

[uneven pulsar]

// the code here is an example of the halt condition
public class lesson3 {
    public static void main(String[] args) {
        int result = sum(4, 8); // in this line we declare the int var and we gace it the method of the entere args 4 and 8
        System.out.println("Sum: " + result); // it will display the results out put would be 30
    }
    // the used method to make the halt condition
    public static int sum(int start, int end) {
        if (end > start) { // it will occure if the end greater than start also the halt condition would trigger if the start > end
            return end + sum(start, end - 1);
        } else {
            return end; // base case: when end == start
        }
    }
}

the halt condition in java

[opaque wharf]

Code line position could change, filename seldomly change

[uneven pulsar]

its really a cool concept

[nocturne olive]

I like this

[olive sable]

i dont

makes my skin hurt

[tender river]

i never even look at the lines, i just / for whatever i need

[nocturne olive]

That's crazy

[trim valve]

https://github.com/vanorsigma/neuro-chat-elo/blob/live-elo/lbo%2Fsrc%2Flib.rs

[olive sable]

short

[uneven pulsar]

i hope vanora see it

i really want to show it but i dont want to bother her

[trim valve]

imo I like the extra separation that having stuff I'm extra files gives here

admittedly the other files are very small

[olive sable]

if i want sepration i just ad a #####################################################################

[knotty current]

ow

[trim valve]

and I would rather not use the mod keyword here

[dry charm]

all IDEs are shit

[rigid snow]

i'm so not sold on kotlin as a java replacement but i'm also very unfamiliar with it, can you convince me it's worth it

also why are the packages named like that (where com.package.name thing )

[olive sable]

make your own and increase the problem

[gritty dust]

lmafo

[dry charm]

Java has domain named packages to avoid clashing

[nocturne olive]

Much less useless code, type inference is good stuff, null safety exists, fully compatible with anything that works with Java
And it looks much cleaner than Java code usually

[rigid snow]

that's why i use a code editor (vscode) instead of an ide

[dry charm]

I use Word 2009 cause it has fully customizable syntax highlighting

[olive sable]

i use my brain

[rigid snow]

i know why it has that but kt is a jvm language, surely you should still do that for java interop?

[dry charm]

Also has a high functional approach, so most of the classes have lambda, iteration or otherwise very helpful functions

[dry charm]

Technically it does not matter, it is not a fixed thing, your package can just be a simple program., but there could be very high change of clashing, meanwhile com.sqbika.myproject has 0% of clashing

But the best example I can give is minecraft modding.
Every mod depends on a Lib or another mod.
One mod decides to ship wrongfully one of the library mods and now you have 2 versions of the same lib mod causing clashing code and crashing

Especially if the two libs are different versions, then it cannot even start

[rigid snow]

yeah but that's more in the dependency realm

[dry charm]

That is why packages are like this

[rigid snow]

you can't necessarily do anything in that scenario

[uneven pulsar]

i use the cmd

[dry charm]

You do

[uneven pulsar]

best code editor

[dry charm]

Gradle shadowing slaps the dependent package under my.minecraft.mod package (obviously the real name, not this) so it becomes my.minecraft.mod.com.minecraft.library

[opaque wharf]

You joke, but an old lecturer at my old uni use MS word to write code

[uneven pulsar]

https://tenor.com/view/arceus-x-google-keyboard-microsoft-swift-roblox-roblox-hacker-gif-12698266540257236113

[rigid snow]

oh cool

[olive sable]

dont care

[dry charm]

One thing to note, is that there is no import aliasing in Java

so this is why all of these are required

[rigid snow]

not com.minecraft though, that's part of the game

right?

also it's net.minecraft

[dry charm]

You can create any packages in com.minecraft

or net.minecraft

[rigid snow]

should you

does anyone

[dry charm]

Absolutely not

Yes

[uneven pulsar]

its funny how people became java pros juz beacuse of minecraf

[dry charm]

It is just another goal to achieve and learnt coding

I learnt JS cause I wanted a discord bot

[uneven pulsar]

nice

i am now learning java

[rigid snow]

because it's an enterprise-circlejerk lang you learn to land a job otherwise

[uneven pulsar]

// the code here is an example of the halt condition
public class lesson3 {
    public static void main(String[] args) {
        int result = sum(4, 8); // in this line we declare the int var and we gace it the method of the entere args 4 and 8
        System.out.println("Sum: " + result); // it will display the results out put would be 30
    }
    // the used method to make the halt condition
    public static int sum(int start, int end) {
        if (end > start) { // it will occure if the end greater than start also the halt condition would trigger if the start > end
            return end + sum(start, end - 1);
        } else {
            return end; // base case: when end == start
        }
    }
}

[dry charm]

Not really tbh

[uneven pulsar]

i also learned this from java third day

[dry charm]

There are plenty C#, C Cpp jobs

[rigid snow]

i know there are

what i mean is practically nobody learns java for funsies outside of minecraft modding

[uneven pulsar]

i finnaly understood halt condition from java

[dry charm]

I enjoyed learning java and I still think it is a fun language

[rigid snow]

[dry charm]

just a bummer that it has so many nuances

but you cannot deny that Java has the best enum system

[rigid snow]

TRUE

[dry charm]

even though it is a fake enum

[uneven pulsar]

minecraf is every gen z childhood game

probably

[dry charm]

There are so many times during C# that I'm like "oh yeah, in C# enum is a glorified number not a glorified class"

If you throw away the enterprise garbage like the factory/builder paradigms, the "make everything an interface+SomethingImpl" java becomes a much better language

less robust but still fun

also java kills the convo

[sage crag]

as it should

[sage crag]

its not as fun

[nocturne olive]

I just do Discord bots in Kotlin

[dry charm]

Yeah, the JS one was years ago, now my bots are in Kotlin as well

Mostly everything is in Kotlin that is in active development

[nocturne olive]

[gritty dust]

Yeah I got it in 2013 on my dad's laptop was fire (I was 4)

[rigid snow]

i write all my discord bots in C guys

[uneven pulsar]

C???????

[rigid snow]

it’s a joke btw would’ve been a crazy flex otherwise

[tender river]

fair! i'm currently trying to understand how real dependent types work so i can see how much of it i can use in lsp (probably none at all since its just comptime javascript)

[opaque wharf]

Why would it be a flex? C is not that complex either. Just very verbose to do

[rigid snow]

because there’s zero reason to not use a higher-level, more abstracted and more modern lang for that, that’s why

[tender river]

i find it amusing how hblang development is all "yay inlining register allocation x86" and then i come and ask for first class functions and equirecursive types

[rigid snow]

also - c isn’t complex, writing in c is

[opaque wharf]

Embedded device development begs to differ. You can make discord bot run on MCU

And do interesting thing with it to interact with IRL stuff

[rigid snow]

nostd rust

also why are you embedding a discord bot

[tender river]

realistically not even nostd, you need so much stuff for a discord bot i doubt it would run without allocators, even time may be useful for something like ssl

[opaque wharf]

Hmmm, yeah you could do that. Or use C and its myriad of ecosystem. For rust, the 2 big framework that I've played with is embassy and one other for STM32 that I forgot

[opaque wharf]

As I said, doing interesting stuff with it to interact with IRL stuff

[tender river]

imagine a buffer overflow leading to ACE in a discord bot