A general dedicated to the discussion and development of local language models.

AIcg niggers need to go back.

Previous Threads: >>97923309 & >>97918783

►News
>(12/15) 3x prompt processing speedup in llama.cpp for Mixtral
>(12/14) Oobabooga adds Mixtral support for Nvidia and CPU-only
>(12/13) Theoretical possibility of 1Bit quantization of MoE layers https://github.com/ggerganov/llama.cpp/issues/4445#issuecomment-1854380383
>(12/13) Koboldcpp merged Mixtral support
>(12/13) Mixtral support added for llama.cpp
>(12/12) Mistral 7B v0.2 release
>(12/08) Mistral released 8x7B MoE model & Together.ai released StripedHyena-7B SSM
>(12/06) Oobabooga adds QuIP# support
>(11/29) DeepSeek-67B released
>(11/28) Yuan 2 released (2B,51B,102B) & Qwen-72B released
>(11/09) Open LLM Leaderboard adds 3 new benchmarks
>(11/05) EXL2 adds 8-bit cache for decreased VRAM usage
>(11/04) koboldcpp-1.48.1 adds Context Shifting for less/no reprocessing (not Smart Context)
>(11/02) Yi-34B model released (200k context also available): https://huggingface.co/01-ai/Yi-34B

►FAQ: https://rentry.org/er2qd
►Links: https://rentry.org/LocalModelsLinks

►Getting Started
https://rentry.org/MixtralForRetards
https://rentry.org/llama-mini-guide
https://rentry.org/8-step-llm-guide
https://rentry.org/llama_v2_sillytavern
https://rentry.org/lmg-spoonfeed-guide

►Further Learning
https://rentry.org/machine-learning-roadmap
https://rentry.org/llm-training
https://rentry.org/LocalModelsPapers

►Benchmarks
General Purpose: https://hf.co/spaces/HuggingFaceH4/open_llm_leaderboard
Programming: https://hf.co/spaces/bigcode/bigcode-models-leaderboard
Censorbench: https://codeberg.org/jts2323/censorbench

►ERP/RP Datasets
https://rentry.org/qib8f
https://rentry.org/ashh2

►Alpha Calculator
https://desmos.com/calculator/ffngla98yc

►Text Gen. UI
https://github.com/oobabooga/text-generation-webui
https://github.com/LostRuins/koboldcpp
https://github.com/lmg-anon/mikupad
https://github.com/turboderp/exui

And despite of all that, men still will never be women.

Why are Tier 1 ISPs doing internet (((content moderation))) in 2023?

/hsg/ - Home Server General
RTFM/JFGI edition.

READ THE WIKI! & help by contributing:
https://wiki.installgentoo.com/wiki/Home_server

>NAS Case Guide. Feel free to add to it:
https://wiki.installgentoo.com/wiki/Home_server/Case_guide

/hsg/ is about learning and expanding your horizons. Know all about NAS? Learn virtualization. Spun up some VMs? Learn about networking by standing up a OPNsense/pfsense box and configuring some VLANs. There's always more to learn and chances to grow. Think you’re god-tier already? Setup OpenStack and report back.

>What software should I run?
Install Gentoo. Or whatever flavor of *nix is best for the job or most comfy for you. Jellyfin to replace Netflix, nextcloud to replace Googlel, ampache to replace spotify, the list goes on and on. Look at the awesome self-hosted list and ask.

>Why should I have a home server?
/hsg/ is about learning and expanding your horizons. De-botnet your life. Learn something new. Serving applications to yourself, your family, and your frens feels good. Put your /g/ skills to good use for yourself and those close to you. Store their data with proper availability redundancy and backups and serve it back to them with a /comfy/ easy to use interface.

>Links & resources
Server tips: https://anonbin.io/?1759c178f98f6135#CzLuPx4s2P7zuExQBVv5XeDkzQSDeVkZMWVhuecemeN6
https://github.com/Kickball/awesome-selfhosted
https://old.reddit.com/r/datahoarder
https://www.labgopher.com
https://www.reddit.com/r/homelab/wiki/index
https://wiki.debian.org/FreedomBox/Features
List of ARM-based SBCs: https://docs.google.com/spreadsheets/d/1PGaVu0sPBEy5GgLM8N-CvHB2FESdlfBOdQKqLziJLhQ
Low-power x86 systems: https://docs.google.com/spreadsheets/d/1yl414kIy9MhaM0-VrpCqjcsnfofo95M1smRTuKN6e-E


Previous: >>85835302

>!!ATTENTION!!
/pcbg/ is currently being spammed. To differentiate your post from the spam, use the timeless 4chinz method of getting attention: include an image with your post.
>UPGRADE & BUILD ADVICE
Post build list or current specs including MONITOR: https://pcpartpicker.com/
Provide specific use cases (e.g. 4K editing, high FPS gaming)
State budget and COUNTRY or you will not be helped
>NEWS
RTX 40 Super refresh expected to be announced in early January at CES
AMD is expected to launch new AM4 X3D CPUs (R7 5700X3D and R5 5500X3D)
>12VHPWR
DO NOT USE CABLEMOD 12VHPWR ADAPTERS, PARTICULARLY THE ANGLED ADAPTER
Fully seat a 12VHPWR connector in its socket, otherwise the connection can melt.
https://youtu.be/ig2px7ofKhQ?t=1345 [Embed]
>CPUS
HTPC(4K60)/Web Browsing: i3 12100
Gaming: R5 7600, 12600K; budget: R5 5600
eXtreme gaming: R7 7800X3D; budget: i5 14600K
Multicore: R9 7950X; budget: i5 13500
>Avoid iGPU-less CPUs
>Existing AM4 board with an old chip? Consider a Zen 3 CPU
>GRAPHICS CARDS
1080p: RTX 4060; budget: RX 6600; ultra budget: used RTX 2060
1440p: RTX 4070; RT: RTX 4070 Ti
2160p: RTX 4080; budget: RTX 4070 Ti
Amateur production: RTX 3060 12GB, RTX 4060 Ti 16GB, used RTX 3090, RTX 4090
>RAM
DDR4: Zen3/AM4 - 2x 16GB 3600 MT/s CL18; for locked "non-K" 12th/13th/LGA1700 - 2x 16GB 3200 MT/s CL16
DDR5: Zen4/AM5 - 2x 16GB 5600 MT/s CL28; for unlocked "K" 13th/14th/LGA1700 - 2x 16GB 6400MT/s CL32
>COOLERS
Standout: Peerless Assassin 120 (or variant)
https://linustechtips.com/topic/891730-cpu-cooler-performance-tier-list/
>PSUS
Not worth buying a new PSU unless it's ATX 3.0 compliant
Aim for 50-75% PSU utilization at full system load
https://cultists.network/140/psu-tier-list/
>SSDS
Standout: WD SN770 2TB
Avoid: outdated Samsung 970 Evo Plus, cheapest NVMe SSDs
https://ssd.borecraft.com/
>MOTHERBOARDS
Good power delivery is required for Intel CPUs like the i5 13500 and above
>GAMING MONITORS
https://pcpartpicker.com/user/pcbg/saved/dP3v4D
Previous: >>97800155

A general dedicated to discussion & development of AI Chatbots

flumfy tail Edition

▶News
Google released Google Gemini Pro - Out on Bard now / Gemini Ultra to release in January 2024
https://blog.google/technology/ai/google-gemini-ai
Claude 2.1 is out, switch to the Claude-2 model in SillyTavern to use it
https://twitter.com/AnthropicAI/status/172700177388865975
Slaude is unusable unless you use/have a corporate account


▶Bots
https://chub.ai
https://chatbots.neocities.org/new_chatbots
https://booru.plus/+pygmalion
https://rentry.org/cai-list
https://rentry.org/meta_bot_list
https://chub-archive.evulid.cc

▶Botmaking
https://github.com/notfiz/de3u
https://agnai.chat/editor
https://avakson.github.io/character-editor
https://rentry.org/meta_botmaking_list

▶Agnai
https://agnai.chat
https://rentry.org/agnai_guides
▶Risu
https://risuai.xyz
▶Silly
https://sillytavern.app
https://rentry.org/aicg4retards

▶Claude
https://rentry.org/meta_claude_list

▶Local models
>>>/g/lmg/
https://rentry.org/meta_golocal_list

▶Other
Character.ai: https://c.ai
Proxy list: https://chub-archive.evulid.cc/proxies
OAI key checker: https://github.com/CncAnon1/kkc
Tech rankings: https://rentry.org/aicg_meta
OP templates: https://rentry.org/aicgOP
Logs: https://chatlogs.neocities.org
JBs: https://rentry.org/Jail-Breaks-for-Different-Models

▶Additional details, guides, links, screenshots: https://rentry.org/aicg_extra_information

murdered in cold blood: >>>/g97823978

>onii-san, will you teach me how to use Linux?
How do you respond?

Spammer is a corrupt biased shill who spams /g/ with shitty threads that accuse our favorite jannies of crnsorship, and forbid discussion about AI and tranime