Free from Drama edition>LOCAL UIreForge: https://github.com/Panchovix/stable-diffusion-webui-reForge | Stable commit: 6964cebComfy: https://github.com/comfyanonymous/ComfyUI | https://comfyanonymous.github.io/ComfyUI_examples>NOOBAI-XLhttps://civitai.com/models/833294/noobai-xl-nai-xlhttps://huggingface.co/Laxhar>NOOBAI SHITMIXES102d custom: https://civitai.com/models/1201815?modelVersionId=1491533291h: https://civitai.com/models/1301670/291h>RESOURCESComment too long. Click here to view the full text.
>>107655588>>107655617I am a close friend of Julien, and during a private conversation, he shared with me that he feels shy and insecure about his UI, and might not want to include it.
>>107655520You forgot the link to Novel AI, you should include it in the next OP. Thanks!https://novelai.net/image
ara ara! sucha new general!
Oekakis incoming!
what's the use case for this?
>>107655526>>107655554this is about as good as those coke holiday commercials.
>>107650844It's a pretty funny benchmark for video generation.
>>107647764loli hentai
>>107655526>>107655554this is real.
>>107647781Kill it with fire
why? whats going on?
>>107655720but the archive wouldnt remove it, its not cp
>>107655639Idk but it's pretty based, hope it continues, in real time too.Check this out:https://raw.githubusercontent.com/DandelionSprout/adfilt/refs/heads/master/AntiPepeList.txt
>>107655720It looks like it's not working if that's the case https://desuarchive.org/g/search/filename/frog/
>>107655738had a good laugh thanks
>>107655738>"its the chudcels!!!">mentally ill mantainerkek
/lmg/ - a general dedicated to the discussion and development of local language models.Previous threads: >>107643997 & >>107636165►News>(12/22) GLM-4.7: Advancing the Coding Capability: https://z.ai/blog/glm-4.7>(12/17) Introducing Meta Segment Anything Model Audio: https://ai.meta.com/samaudio>(12/16) MiMo-V2-Flash 309B-A15B released: https://mimo.xiaomi.com/blog/mimo-v2-flash>(12/16) GLM4V vision encoder support merged: https://github.com/ggml-org/llama.cpp/pull/18042>(12/15) llama.cpp automation for memory allocation: https://github.com/ggml-org/llama.cpp/discussions/18049►News Archive: https://rentry.org/lmg-news-archive►Glossary: https://rentry.org/lmg-glossary►Links: https://rentry.org/LocalModelsLinks►Official /lmg/ card: https://files.catbox.moe/cbclyf.pngComment too long. Click here to view the full text.
Tired of pretending dense models beyond 4b have any future in this space. Some people enjoy them but that's the problem; they enjoy them, but the expense of tk/s is a tradeoff no one can actually afford, considering that applications beyond the saturated chatbot market demand high throughput almost real time responses. The closer you move to domestic hardware, the more problematic the density of the model is, because the hardware Nvidia let's us have at that level sucks.Anyway. Even if a 14b beat at some edge case a massive model I wouldn't care. To do hard engineering with AI you need low latency response time. This is why Chinese models, specifically qwen3 next are the rage, even when trained in garbage Huawei hardware.
>>107654309One where the merchants didn't turn it into a secular consumerist sham, I guess.
>>107655084bait used to be believable
>>107655145That world isn't real. It's also long been a secular holiday as well even aside from the consumerism. Also many cultures and subcultures have incorporated it in their own different ways since it was thrust onto them everywhere or had preexisting traditions that overlap with christmas.
>>107654882Maybe I've been placebo-ing myself. Feel thinking would help most in lengthy/lore detailed scenes or when the model starts doing impossible anatomy/inconsistent clothing etc. It gets autistic about details and when/how to incorporate things in the <think> and intuitively here 90% of the tokens are planning the response - 10x computation surely extracts more of the model potential than yoloing into an answer?Most of my use is assistant desu or rerolling the same handful of opening RP chats to test model behaviour.The speed is really nice tho, bigger factor than I had considered ig if 90% of TG is inside <think>Will give nothink a chance for RP tho, thx anon you pushed me to see for myselft. 16K context is all you need anon
How hard is learning VHDL/verilog gonna be if I'm an okay programmer but know nothing about electronics? I want to design obscure CPUs, lisp machines or something
>>107655518>I want to design obscure CPUsyou read yesterdays' /dpt/, didnt you?the cpuanon's story was pretty inspiring, wasnt it?i too wanna do vhdl+verilogbut i dont care how hard its gonna be, im just gonna fkn do it and thats itim too retarded to realize things are too hard and thats how i end up succeeding
>>107655564>you read yesterdays' /dpt/, didnt you?Not really, what's the story in question? Idk, it seems interesting, but I absolutely don't want to learn physics because I was bad at it when I was in middle school. I started thinking about it because I wonder if creating a modern lisp machine is actually possible.
>>107655518All you need is NAND to Tetris and Digital Design and Computer Architecture (RISC-V Edition) by Sarah Harris and David Harris..NAND to Tetris teach you digital design not VHDL or System Verilog they use a custom HDL language.
>>107655628it not much of a story to be honest>be anon>write a risc cpu in an fpga using verilog>use said cpu to control a led panel>the endi still found it to be based as fucki already had vhdl/verilog on my bucket list but now i really really wanna do this shit, for reals now
>>107655644>NAND to Tetristhank you for the recommendation, I'll check it out>>107655678yeah, sounds cool. what ISA did it use?
https://www.thurrott.com/dev/330980/microsoft-to-replace-all-c-c-code-with-rust-by-2030>Our North Star is ‘1 engineer, 1 month, 1 million lines of code.’ To accomplish this previously unimaginable task, we’ve built a powerful code processing infrastructure. Our algorithmic infrastructure creates a scalable graph over source code at scale. Our AI processing infrastructure then enables us to apply AI agents, guided by algorithms, to make code modifications at scale. The core of this infrastructure is already operating at scale on problems such as code understanding.”LOL
>>107654925Some student was working on a prototype bringing Rust to .NET. Probably only a few years away from Microsoft officially adopting it and calling it R#.
>>107655120A formal proof would require explaining to you what Rust is and how it works.
>>107655668You can skip that. I have been programming in Rust and many other languages for years.Simply show in what way is unwrap special that does not apply to assert.
>>107655685So you already know in what way unwrap is special and I don't need to explain.
>>107655734The burden of proof lies with the one making the claim.
Why doesn't MS get enough credit for DirectX?Every GPU innovation ever only exist because Microsoft kickstarted it through DirectX. Remember when GPU went from the separated Vertex and Pixel shaders paradigm to the unified shaders paradigm around 2006 ? That was because Microsoft rewrote their graphical API stack for WIndows Vista and the Xbox 360. NVidia and ATI only created GPUs that would work with the new API.>DirectX 8 wasn't an explosive update, but it completely altered the way modern rendering works. Every single device we use right now, from phones to consoles to computers, they wouldn't work the way they do without employing DX8's fundamental principle of control>DirectX 9 was first released in 2002, games are still made for it and in some cases only for DX9>Xbox 360 went significantly beyond Direct3D 9 / DX9 / Shader Model >DirectX12 brought us a ton of new stuff, such as low level access to hardware, better CPU parallelization, ray-tracing, async computing, mesh shaders, autohdr>DX11 and DX12/Vulkan are designed to solve very different problems>DX12 unifies Windows PC and Xbox Console development with one API, making it easier for game devs >Your GPU is built for DirectX
>>107652850Your hypothetical better solution requires more computing hence even stupider.
>>107652893this
>>107652470posted it again award>>107652893this
>>107652470DirectX was Microsoft's way of segmenting their operating systems, whereas OpenGL applications could run on anything from Windows 98 to the present day.
>>107652470>look mom I posted it again!https://desuarchive.org/g/search/text/innovation%20directx/type/op/
What kind of physical media do you collect, anon?
>>107653896That works with DVD, not Blu-ray
>>107654567>vinyl and cd'sI want to do this but I keep telling myself it's simply more convenient to have a big HDD with my music.
>>107650128still collect? none. bit i still have vhs tapes, audio tapes, cds and dvds. only have the means to read the cds and dvds though.
I stopped coomlecting physical media a number of years ago.
I don't collect anything. My grandparents had a bunch of shit when they died and it was a pain to deal with. Never again.
Is Ublock losing the Youtube war?
>>107655000Cute image anon(male), but this changes nothing.
Not on Brave.
>>107645568PAY OR ELSE
>>107645568Werks on my machine.
>>107646536>Look at /fit/but I ain't gay
Sam is based for this.Post yours.I love ChatGPT.I hope it wins the AI war.
>>107653389I only use AMD products.
>>107655064and? not him but i've had no issue running llm's and txt2img models on my amd cards
>>107640770it wants memory turned on, unfortunately.a bunch of these companies lure people into giving away data with the prospect of pointless bumhole gazing, see 23andme's 'welcome to you' in exchange for giving Mrs. google your dna
>>107653004
>>107655316openai has the data and every summaries it wants about whether or not you've turned memories on. this feature only affect YOUR access to the data.
Previous /sdg/ thread : >>107636815>Beginner UIEasyDiffusion: https://easydiffusion.github.ioSwarmUI: https://github.com/mcmonkeyprojects/SwarmUI>Advanced UIComfyUI: https://github.com/comfyanonymous/ComfyUIForge Classic: https://github.com/Haoming02/sd-webui-forge-classicStability Matrix: https://github.com/LykosAI/StabilityMatrix>Z-Image Turbohttps://comfyanonymous.github.io/ComfyUI_examples/z_imagehttps://huggingface.co/Tongyi-MAI/Z-Image-Turbohttps://huggingface.co/jayn7/Z-Image-Turbo-GGUFComment too long. Click here to view the full text.
gm
>gm
Everything is already from China, but here we discuss the cheap chink shit you see on various sites.1st rule of /csg/: If it looks too good to be true, it probably is.Useful links>TWS IEM reviews: www.scarbir.com/>Guide: csg-guide.neocities.org/>Wiki: https://igwiki.lyci.de/wiki/Chink_Shit_General>What headphones/earbuds should I buy?>>>/g/iemg>I want a cheap smartphone what should I buy?>>>/g/spg/>I want to buy some sort of emulation device>>>/vr/hhg/Comment too long. Click here to view the full text.
>>107654716XIAOMI MIJIA
>>107654716iirc Dyson cyclone patent has expired so now everyone can make actual good copies, but many chink vacuums only copy external appearance but not the internal engineering.
>>107654716buy a dyson with a broken battery and buy a dyson to makita adapter and use some reputable fake makita batteriesor if you have tools use that tool battery adapter
update on >>107633696opened a dispute and ali sided with me no questions asked (pleasant but that's how it should be, tracking clearly said Delivery unsuccessful)My grievance is that this is the only listing on the site selling this product, will Chang be angry if I order again after the dispute? will the courier fail to deliver again? No other packages of mine have gone lostaliexpress.com/item/1005007274193336.htmlit went up in price too...
>>107655710i doubt itbut at worst just pay family or a friend to buy another
HDB Edition >Not sure what private trackers are all about?A private tracker is an invite-only torrent website. Each member shares common goals: collecting, preserving and discussing media.>Have a question?- FAQ https://archive.is/UVQkn- WIKI https://igwiki.lyci.de/wiki/Private_trackers- NEWFAG PYRAMID https://inviteroute.github.io/graph or https://inviteroute.github.io/sheet/- STUDY https://www.cambridge.org/core/services/aop-cambridge-core/content/view/2F379FE0CB50DF502F0075119FD3E060- SPREADSHEET https://hdvinnie.github.io/Private-Trackers-Spreadsheet/- TEN CURRY COMMANDMENTS https://pastebin.com/raw/dBbdE73M (embed)- TEN NEON COMMANDMENTS https://pastebin.com/raw/Ud2pGYaE (embed)- RED SPAMMER'S BIBLE https://rentry.org/69zbxh4h- #ptg is on irc.sageru.org but it's pretty deadComment too long. Click here to view the full text.
>>107655470that's stelks. this dude should be banned for discussing recruitment in the public. crazy that known edge lords run aither.>>107655501yup any other recruitment is 5x easier. grohl is expecting high volume when there are 21 titans lmao. most uploader and trustee ranks don't even meet these reqs. does grohl even have titan stats?
>>107655470>>107655638What an asshole
>>107655638It doesn't matter. They aren't interested in taking new members, but to say "no, but see? look, we are recruiting now. You should try join us there are no excuses now"
>>107655684stop crying and start contributing neon
can you not buy upload on ATH with bonus points or something?
Yearphone of the ear editionHow to request advice:>Budget>Intended use (media, source, environment)>Frequency response preference and music examples>Past gear and your thoughts on themFAQ:>Where do I buy IEMs?Amazon, Aliexpress, Linsoul, Hifigo, Shenzhenaudio>Shopping Guide (IEMs, PMPs, Cables, Ear Tips, etc.):https://rentry.org/consoomer_guideComment too long. Click here to view the full text.
> you can't bake HRTF's into a BRIR. ~/.config sox --combine mix -v 1 BRIR_impulcifer_hrtf.wav -v -1 BRIR_personal_hrtf.wav diff.wav ~/.config sox diff.wav -n statSamples read: 470386Length (seconds): 0.699979Scaled by: 2147483647.0Maximum amplitude: 0.026776Minimum amplitude: -0.034804Midline amplitude: -0.004014Mean norm: 0.000052Mean amplitude: 0.000000RMS amplitude: 0.000338Maximum delta: 0.042889Minimum delta: 0.000000Mean delta: 0.000082Comment too long. Click here to view the full text.
>>107651614it's sad he'll never post collection (he doesn't have one) because i'd like to see the pair that caused his hearing loss. he threw it out in a fit of rage and asshurt kek. along with his previous dongles
>>107655538It's sad that you'd rather think there's a conspiracy among all IEM manufacturers to fake sensitivity numbers than to measure your listening volume.
>>107654945>i can bake someone else's BRIR (their HRTF + room) that was measured with binaural mics with a DTF (HRTF minus DF) from EAC, it's not wrong, okayy??>i can neither measure my pos in situ nor make a compensation filter for it because it's a pos, but it's not wrong, okayy??>i can't use head-tracking, so whenever i move my head i will have wrong directional cues, and that's amazing okayy??? ok retard, no one cares, enjoy the mess
If you want spaciousness for music just add some light reverb, from your room made with a measurement mic, not in-ear mic, with the IR EQd to be reasonably flat to decouple reverb from tonality. Speakers without reverb sound in your head too so HRTF is not the important part, HRTF only matters when playing native surround content.
/aicg/ - A general dedicated to the discussion and development of AI chatbotsIt's Over Edition>NewsZ.ai releases GLM 4.7 https://z.ai/blog/glm-4.7Google releases Gemini 3 Flash https://blog.google/products/gemini/gemini-3-flashOpenAI releases GPT-5.2 https://openai.com/index/introducing-gpt-5-2Deepseek releases V3.2 https://huggingface.co/deepseek-ai/DeepSeek-V3.2Anthropic releases Opus 4.5 https://www.anthropic.com/news/claude-opus-4-5Google releases Gemini 3 Pro https://blog.google/products/gemini/gemini-3Additional info: https://aicg.neocities.org/info.html>FrontendsComment too long. Click here to view the full text.
Is AI really going to replace all jobs in 18 months? Bunch of people on Twitter are posting exactly that.
>>107655478Yes, only whores like your mom will stay safe
>>107655473>If you're using OR, yesoh, that explains it. Uh, so is .99 the highest creativity temp it can go?
>>1076554891.0-0.9 is fine. haven't tested OR in ages. quality depends on whatever provider OR picks per request.
>>107655536yeah, it's working now. one more question, should I be <think>ing with 3.2?
