/lmg/ - a general dedicated to the discussion and development of local language models.New Years Eve EditionPrevious threads: >>107709248 & >>107700909►News>(12/29) HY-Motion 1.0 text-to-3D human motion generation models released: https://hf.co/tencent/HY-Motion-1.0>(12/29) WeDLM-8B-Instruct diffusion language model released: https://hf.co/tencent/WeDLM-8B-Instruct>(12/29) Llama-3.3-8B-Instruct weights leaked: https://hf.co/allura-forge/Llama-3.3-8B-Instruct>(12/26) MiniMax-M2.1 released: https://minimax.io/news/minimax-m21>(12/22) GLM-4.7: Advancing the Coding Capability: https://z.ai/blog/glm-4.7►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.png►Getting Startedhttps://rentry.org/lmg-lazy-getting-started-guidehttps://rentry.org/lmg-build-guideshttps://rentry.org/IsolatedLinuxWebServicehttps://rentry.org/recommended-modelshttps://rentry.org/samplershttps://rentry.org/MikupadIntroGuide►Further Learninghttps://rentry.org/machine-learning-roadmaphttps://rentry.org/llm-traininghttps://rentry.org/LocalModelsPapers►BenchmarksLiveBench: https://livebench.aiProgramming: https://livecodebench.github.io/gso.htmlContext Length: https://github.com/adobe-research/NoLiMaGPUs: https://github.com/XiongjieDai/GPU-Benchmarks-on-LLM-Inference►ToolsAlpha Calculator: https://desmos.com/calculator/ffngla98ycGGUF VRAM Calculator: https://hf.co/spaces/NyxKrage/LLM-Model-VRAM-CalculatorSampler Visualizer: https://artefact2.github.io/llm-sampling►Text Gen. UI, Inference Engineshttps://github.com/lmg-anon/mikupadhttps://github.com/oobabooga/text-generation-webuihttps://github.com/LostRuins/koboldcpphttps://github.com/ggerganov/llama.cpphttps://github.com/theroyallab/tabbyAPIhttps://github.com/vllm-project/vllm
►Recent Highlights from the Previous Thread: >>107709248--Text-to-animation model potential vs current limitations in NPC/game applications:>107713005 >107713022 >107713094 >107713093 >107713161 >107713166 >107713180 >107713199 >107713395 >107713478 >107714633 >107715953 >107715974 >107716259 >107716282 >107716362 >107716378 >107716412 >107716427 >107716455 >107716463 >107716491 >107716546 >107716556 >107716643 >107716752 >107716648--Choosing SillyTavern prompt modes based on model type and customization needs:>107715217 >107715245 >107715274 >107715300 >107715318 >107715651 >107715665 >107715758 >107716851 >107716897 >107716949 >107717038--TTS tool landscape: XTTSv2, Chatterbox, and lightweight alternatives:>107714613 >107714656 >107714858 >107714701 >107714733D--Kimi Linear architecture boosts efficiency and compatibility:>107712023 >107712128--Feedback on updated model recommendations list:>107715036 >107715051 >107715075 >107715081 >107715093 >107715088 >107715623 >107715752 >107715866--M2.1's jailbreak resistance and identity adherence challenges:>107714948--Llama.cpp API compatibility issues with enable_thinking flag:>107712820 >107712892--Optimizing inference with mixed GPU/CPU setups:>107714237 >107714606 >107714638 >107714341 >107714370--32GB VRAM roleplay model recommendations and narrative challenges:>107710932 >107710954 >107711017 >107711047 >107711311 >107711369 >107711409 >107711524 >107711716 >107711757 >107712384 >107714928 >107714953--Exynos 2600 NPU advancements vs Nvidia's market dominance:>107713883 >107713925--ikllama's declining Windows support and maintenance challenges:>107714744 >107714804--Logs: Migubench:>107715268--Logs:>107710745 >107712779 >107713077--Teto (free space):>107709736 >107709978 >107710051 >107711883 >107711900 >107712939 >107713166►Recent Highlight Posts from the Previous Thread: >>107709259Why?: >>102478518Enable Links: https://rentry.org/lmg-recap-script
GIWTWM
>>107717246I want to be the Teto
It's overReal life people talk like slop now
>>107717380that sounds absolutely awful.
What's the gooner recommendation for 96gb vram, assuming something like 4bit cope, I'm not able to offload to ram.
>>107717404StableLM 7B
>>107717404Use SaaS APIs.>t. someone with 96GB VRAM
Bald anime girls make better threads. Decline is already very visible here.
>>107717404Q4 or Q5 of air.
>>107717380You kind of have to wonder though, how much time a guy like Elon is actually spending each day talking with AI.
>>107717464LLMs are amazing for brainstorming and exploring new concepts.
>>107717246erotic colors
>>107717380If it is this piece of shit faggot proclaiming it is gonna happen then we are safe. On the other hand making a smartphone with some weird forked android OS that only lets you use twitter and grok should be easy enough that even with his incompetence it can happen.
>>107717472Brainstorming, eh? You want the details, the real deal?
>>107717404>96gb vramGet a second one and run 4.6/4.7
>>107717380>a screen with radios that pings Grok/xAI servers for everything in real timewho asked for this?
>>107717508it is being enforced on us like every other rich person who wants to let privacy become a thing of the past
>>107717308Bald is the new /lmg/ mascot.
>>107717575
Am retard. How do I stop sillytavern from showing shit in code blocks? They aren't hidden with some models and I don't know why. I still want to gem them, I just don't want to see them.
so what's a good story/rp telling model? been using mistral nemo 12b for a while, anything new to try?
>>107717380stop wasting time on gadgets that nobody's going to buy and finish up the full dive brainchips you rich cunt
>>107717669do you mean the "thinking" blocks?what exactly do you mean?
>>107717669regex
>>107717695hes an adhd rich idiot, investors had to give him a 1 Trillion carrot on a stick, just so he continues working on the electric cars
>>107717689Magidonia or Cydonia. Ur specs??
>>10771772916vram 32ram4060ti
>31 December 2pm in worst korea>bloody besterd mathrachods still haven't done needful release of the solar open 100b gooofsMay sacred cow curse their ancestors and descendants to toilet witch dimension
>>107717729Cydonia just shat me this lil nugget from just the title.https://rentry.org/opkobdrp
>>107717729>>107717758Die in a fire
>>107717758>chapter 2: finding your partnerPretty funny.
>>107717689Just keep using nemo desu.
8gb vram. 64gb ram. spoonfeed me model. i want for the BIG sex. i run utilize GLM-4.5-Air-Q4_K_S and NemoReRemix-12B-Q3_K_XL and sometime trashpanda-org_QwQ-32B-Snowdrop-v0-IQ4_XS there is it there anything better?
>>107717805not really, no. there has not really been much progress made in the past few months for us mortals.
>>107717805Why are there so many people asking this recently?>Hey guys what's the current best model????Bot spam again?
>>107718128That's a clever insight —
Hunyuan motion 1.0 is fun
