#localllm 搜尋結果
Just fed my local LLM a messy codebase. It spat back: 'This isn't code, it's abstract art. Want me to refactor or frame it?' Now it's demanding coffee breaks. Who's training who here? Also, what would it tell you? github.com/sukanto-m/dire… 📷📷 #LocalLLM #AICreativity #TechHumor
stable diffusionをローカルで体験中。 A1111とSDNextの環境構築したけれど、A1111はなぜか抽象画になっちゃう（笑） 本命はComfyUIなので、動作確認は完了的なステップ #localLLM #radeon780M #stablediffusion
ローカルLLMは「開放重み×量子化×監査」でビジネス要件(コスト/機密/遅延)を満たす。 Llama 3・Mistral 7B・Gemma 2等を4/8bit量子化→単GPU 8–24GBで実用。TGIはHTTP/gRPC、ストリーミング/バッチ/監視を備え、本番運用に適す。 #LocalLLM huggingface.co/docs/text-gene…
事業でローカルLLMを使うなら、まず汎用サービングのvLLMで土台を固め、用途に応じてモデルを選ぶ vLLMはGemma 2(9B/27B)、Llama 3.1、Qwen2をサポートし、LoRAとPP(並列化)に対応。小型でPoC→LoRA適応→GPU/メモリ見積り #vLLM #LocalLLM docs.vllm.ai/en/latest/mode…
業務でローカルLLMを使う結論：まずLM Studio。機密データを出さずに社内PCで推論し、試作から運用までの橋渡しに向く 事実: 無料で商用利用可／OpenAI互換APIとCLI(lms)／Qwen・Gemma・DeepSeekなどのオープンウェイトをワンクリック実行 #LocalLLM #LMStudio
ローカルLLMは機密/コスト/レイテンシで効く。OllamaならGGUF量子化の Llama 3 8B・Mistral 7B・Gemma 2 を1コマンドで実行、Modelfileでプロンプト/テンプレ/推論設定を再現可能。 業務はRAG・要約・コードレビューから開始。Q4→Q5で比較し、VRAM/応答時間/品質を記録して選定。#Ollama #LocalLLM…
ローカル/オープンウェイトLLMの運用は、Hugging FaceのTGIが現実解。 SSE配信・連続バッチ・テンソル並列、量子化(bitsandbytes/GPTQ)、監視(Prometheus)・分散トレース(OpenTelemetry)、関数呼び出し(Guidance)。Llama/Falcon/StarCoder対応。7B級でPoC→GPU増設＋TPで拡張。#LocalLLM #MLOps
Great use of Ollama running locally but having my hosted n8n use it for many of the AI tasks that otherwise would cost me money (ok so the Mac mini cost me money 🤑) #n8n #localllm
Microsoft just launched a MINI MONSTER Runs LOCALLY. Thinks like a GIANT. Meet Phi-4-reasoning-plus: ✅ 14B parameters ✅ Beats bigger models ✅ Open & fast on your PC This changes the game. 🧵👇 #AI #LocalLLM #Phi4 #MachineLearning
Just got Kimi K2 running locally via @UnslothAI 's GGUF. Q1 variant, 15 layers GPU, rest RAM. A 1T param model on consumer hardware. My cloud API bills watching me run inference for free: 💀 Welcome to the "Heavy Metal in your basement" era of AI. #LocalLLM #KimiK2 #Unsloth
ローカルLLMを業務で使う近道は、vLLMのOpenAI互換サーバで社内APIを立てること。/v1/chat/completions 互換で「vllm serve NousResearch/Meta-Llama-3-8B-Instruct」で即起動、extra_bodyでtop_kやJSONのstructured outputs、Ray Serveで自動スケールも可能。#vLLM #LocalLLM docs.vllm.ai/en/latest/serv…
Something big is brewing... 💪 Our new local AI rig is taking shape, powered by 2 x NVIDIA RTX 5090s. What models would you run on this rig? DeepSeek-R1, Qwen 3, Llama 3, Gemma 3, or something custom? Drop your setups and ideas below! 👇 #AI #LocalLLM
TIL I can check LLM inference speeds on my current (or potential) system for any model on hugging face. rahulschand.github.io/gpu_poor/ #machinelearning #localllm
🦙📱 Build a mobile app in Blazor for your Local LLM! No cloud. No subscriptions. Just code. 🎥 Tutorial → youtu.be/5wWswrwYkUo #Blazor #AI #LocalLLM #DotNet
youtube.com
YouTube
Build Mobile App for your Local LLM in Blazor
Just finished 30 days testing Parakeet v3 vs Whisper for our macOS app. Here's what we learned about local speech-to-text and local diarization 🧵 Full breakdown: reddit.com/r/LocalLLaMA/c… #SpeechToText #Whisper #LocalLLM #AI #Parakeet #MACOS
Just fed my local LLM a messy codebase. It spat back: 'This isn't code, it's abstract art. Want me to refactor or frame it?' Now it's demanding coffee breaks. Who's training who here? Also, what would it tell you? github.com/sukanto-m/dire… 📷📷 #LocalLLM #AICreativity #TechHumor
"Built a local LLM to organize my code folders. It suggested: 'Delete everything and start over.' Wise bot. Now it's unionizing my tabs against me. Who's really in control here? 🤖💻 #LocalLLM #AICreativity #BuildInPublic Try and see how it goes for you! github.com/sukanto-m/dire…
Llama 3.2（1B/3Bのオンデバイス向け・オープンウェイト）を軸に、ビジネスでのローカルLLMは「小さく速く安全」を徹底。 実装要点: RAGで社内文書接続、Int8/4量子化でVRAM≲8GB運用、関数呼び出し＋監査ログで業務統合。評価はタスク別品質×P95遅延で見る。 #Llama #LocalLLM…
UGREEN AI NAS: Local LLMs on AI NAS: It's not just tech, it's business. Two experts break down the specs and use cases. The future is local, powerful, and ready to deploy. #LocalLLM #AINAS
ローカルLLMは機密/コスト/レイテンシで効く。OllamaならGGUF量子化の Llama 3 8B・Mistral 7B・Gemma 2 を1コマンドで実行、Modelfileでプロンプト/テンプレ/推論設定を再現可能。 業務はRAG・要約・コードレビューから開始。Q4→Q5で比較し、VRAM/応答時間/品質を記録して選定。#Ollama #LocalLLM…
Built my first local LLM project - a directory monitor that uses @ollama to analyze messy codebases and suggest specific cleanups 🤖 Everything runs locally (no API keys!), learned a ton about RAG and embeddings while building it. @simonw #LocalLLM #BuildInPublic #Ollama…
結論: ローカルLLMは「低コスト×データ保護×低遅延」で業務適用の現実解 4bit量子化で7B/8B（Llama 3.1 8B, Mistral 7B, Qwen2.5 7B）が16GB級PC/GPUで稼働。RAG＋Ollama/vLLMで社内検索・要約、1リクエスト数円。導入の勘所は用途別モデル選定とアクセス制御・監査ログ #LocalLLM…
Still learning and improving it, but it's open source if you want to tinker with it too 📖 Blog post: dev.to/sukantom/i-bui… 💻 GitHub: github.com/sukanto-m/dire… Feedback welcome! First time building something with LLMs #Python #LocalLLM #BuildInPublic #Ollama #OpenSource
ローカルLLMは「開放重み×量子化×監査」でビジネス要件(コスト/機密/遅延)を満たす。 Llama 3・Mistral 7B・Gemma 2等を4/8bit量子化→単GPU 8–24GBで実用。TGIはHTTP/gRPC、ストリーミング/バッチ/監視を備え、本番運用に適す。 #LocalLLM huggingface.co/docs/text-gene…
Running local LLMs for fun and research? This project uses `Ollama`, `Python`, and two AIs to simulate full-on conversations — no cloud APIs. Great for prompt engineering, classroom demos & multi-agent UX. #AI #LocalLLM #LLMDev #Ollama #Python praudyoneek.com/ai-chatbots-ll…
Qwen3-235B (ローカルLLM) にGrokの強制ルーティング問題の話してたらちゃっかり #LocalLLM のハッシュタグ提案してきてかわいい いいモデルだけどちょっとウェットなところがあって、すぐに「この真実に気付いたあなたは特別ー🫵」的なマインドコントロール術を使ってきがちだからなんとか調整したい
事業でローカルLLMを使うなら、まず汎用サービングのvLLMで土台を固め、用途に応じてモデルを選ぶ vLLMはGemma 2(9B/27B)、Llama 3.1、Qwen2をサポートし、LoRAとPP(並列化)に対応。小型でPoC→LoRA適応→GPU/メモリ見積り #vLLM #LocalLLM docs.vllm.ai/en/latest/mode…
Forget one AI. Why not two? Gemma & LLaMA 3.2 can now talk to each other thanks to a simple Python app running locally with Ollama. All you need: - Python - Jupyter - A weird topic #AI #LocalLLM #Ollama #Python praudyoneek.com/ai-chatbots-ll…
Let AI agents chat & brainstorm ideas for you—solving problems and generating insights while you focus on your own work! 🚀🤖 #AI #automation #LocalLLM #productivity #Python #streamlit #agents
🚨 Why developers should run LLMs LOCALLY... Most pay $20-100/month for API calls. Local LLMs give you: ✅ ZERO ongoing costs ✅ 100% privacy ✅ No rate limits ✅ Offline testing ✅ Predictable performance What's stopping you? 👇 #LocalLLM #AI #DataPrivacy #DevTools
I got Meta's Large Language Model (llama2-70B) running locally on the dual RTX 3090 AI PC I just built 🤓 48GB of VRAM lets me run the full 70 billion parameter model (GPT3.5 level) with no limits & no Internet connection required. What should I ask it? #LocalLLM #LocalLlama
stable diffusionをローカルで体験中。 A1111とSDNextの環境構築したけれど、A1111はなぜか抽象画になっちゃう（笑） 本命はComfyUIなので、動作確認は完了的なステップ #localLLM #radeon780M #stablediffusion
Qwen3-235B (ローカルLLM) にGrokの強制ルーティング問題の話してたらちゃっかり #LocalLLM のハッシュタグ提案してきてかわいい いいモデルだけどちょっとウェットなところがあって、すぐに「この真実に気付いたあなたは特別ー🫵」的なマインドコントロール術を使ってきがちだからなんとか調整したい
Built my first local LLM project - a directory monitor that uses @ollama to analyze messy codebases and suggest specific cleanups 🤖 Everything runs locally (no API keys!), learned a ton about RAG and embeddings while building it. @simonw #LocalLLM #BuildInPublic #Ollama…
Incredibly, @ComfyUI now supports Wan2.2-TI2V-5B for local text-to-video and image-to-video. For their Q4_K_M in GGUF it made me ~5 secs of video in ~7 mins on my RTX3080 w/10GB. Incredible that it works at all. This is their template workflow with my English negative. #LocalLLM
Big Win for Local AI Devs! Just got Windows Foundry Local up and running with AI Insights in VSCode—unleashing Phi-3.5 mini-instruct on my rig. Super fast inference thanks to CUDA on my 6GB NVIDIA card! . Phew, game-changer for home labs! #LocalLLM #Phi35 #CUDA #VSCode
You don’t need the cloud to use ChatGPT anymore. I just built my own local AI stack — Ollama + OpenWebUI — and now I can run Mistral offline, upload PDFs, and get answers with zero internet. 🧠 Full tutorial coming this week 👇 youtube.com/@atefataya/vid… #LocalLLM #Ollama…
I've been experimenting with running AI models locally on my PC (on my RTX 4070), and I just tested the new Nemotron-Nano-12B v2 model from @nvidia. The performance is absolutely staggering! #LocalLLM #PrivateAI #OpenSourceAI #MachineLearning #AmericanTechStack #NVIDIA #Nemotron
LM Studio + VSCord + ContinueでローカルLLMをCursorみたいに使えるようにしたぞー！ ただ、ものすごく遅い上にコードの質も微妙（プロンプトが悪い可能性は否めないが。）だし、Macが熱々になる。 R1よりV3の方がコーディングに強いらしいので今度試してみる！ #DeepSeekR1 #LocalLLM
Microsoft just launched a MINI MONSTER Runs LOCALLY. Thinks like a GIANT. Meet Phi-4-reasoning-plus: ✅ 14B parameters ✅ Beats bigger models ✅ Open & fast on your PC This changes the game. 🧵👇 #AI #LocalLLM #Phi4 #MachineLearning
Something big is brewing... 💪 Our new local AI rig is taking shape, powered by 2 x NVIDIA RTX 5090s. What models would you run on this rig? DeepSeek-R1, Qwen 3, Llama 3, Gemma 3, or something custom? Drop your setups and ideas below! 👇 #AI #LocalLLM
Something went wrong.
Something went wrong.
United States Trends
- 1. Jeremiah Smith N/A
- 2. Vanderbilt 4,787 posts
- 3. French Laundry 2,335 posts
- 4. Christmas 127K posts
- 5. Jim Knowles N/A
- 6. Amad 13.2K posts
- 7. #HookEm 2,140 posts
- 8. Arch Manning 1,736 posts
- 9. Colin Simmons N/A
- 10. ESPN 80.1K posts
- 11. Dalot 11K posts
- 12. Ryan Wingo N/A
- 13. Julian Sayin 1,493 posts
- 14. Disney 91.8K posts
- 15. Buckeyes 2,597 posts
- 16. Casemiro 19.4K posts
- 17. Jojo Trader N/A
- 18. #MUFC 18.8K posts
- 19. Simpsons 57K posts
- 20. Gameday 34.9K posts