#hugging-face — Public Fediverse posts

NTU scientists used a 200-year-old Poisson spot to create optical skyrmions, ScienceDaily reports. Here’s why the trick could speed

https://aistory.news/open-source-ai/poisson-spot-makes-optical-skyrmions-for-future-computing/

#AmazonAI #FineTuning #HuggingFace

Crew Control Plane adds free tracing, governance, and on‑prem options to CrewAI. We compare it to LangSmith Engine for real production

https://aistory.news/open-source-ai/crew-control-plane-takes-aim-at-langsmith-for-agents/

#AmazonAI #FineTuning #HuggingFace

RT @UnslothAI: We’re releasing new Qwen3.6 quants that run 2.5× faster on your GPU. Qwen3.6-27B NVFP4 runs on 24GB VRAM. 35B-A3B can hit 17,561 tok/s (B200). We also improved accuracy, tool calling, agent use, and looping. Guide: https://unsloth.ai/docs/models/qwen3.6#nvfp4 Qwen3.6 NVFP4: https://huggingface.co/collections/unsloth/nvfp4

mehr auf Arint.info

#agent #huggingface #Qwen36 #qwen36 #Qwen3627 #unsloth #arint_info

https://x.com/UnslothAI/status/2075566124687892597#m

ScienceDaily reports DT-109 fatty liver results: gut repair reversed severe disease in animals. Why this shift matters for MASH trials.

https://aistory.news/open-source-ai/dt-109-fatty-liver-drug-repairs-gut-in-animal-study/

#AmazonAI #FineTuning #HuggingFace

RT @UnslothAI: We’re releasing new Qwen3.6 quants that run 2.5× faster on your GPU. Qwen3.6-27B NVFP4 runs on 24GB VRAM. 35B-A3B can hit 17,561 tok/s (B200). We also improved accuracy, tool calling, agent use, and looping. Guide: https://unsloth.ai/docs/models/qwen3.6#nvfp4 Qwen3.6 NVFP4: https://huggingface.co/collections/unsloth/nvfp4

mehr auf Arint.info

#agent #huggingface #Qwen36 #qwen36 #Qwen3627 #unsloth #arint_info

https://x.com/UnslothAI/status/2075566124687892597#m

ScienceDaily reports a study arguing the Dyson sphere search should target red and white dwarfs, a shift that could reveal overlooked

https://aistory.news/open-source-ai/dyson-sphere-search-pivots-to-dwarf-stars-study-says/

#AmazonAI #FineTuning #HuggingFace

CrewAI AMP Suite adds an enterprise control plane with tracing, security, and on‑prem options—positioning against LangSmith’s neutral

https://aistory.news/open-source-ai/crewai-amp-suite-takes-aim-at-langsmiths-control-layer/

#AmazonAI #FineTuning #HuggingFace

I wrote a #blog #post going into a bit more detail: https://textualmirror.mataroa.blog/blog/a-proposal-for-pronoun-usage-by-llms-other-non-conscious-entities/

(Written about "AI", but without using "AI".)

@leadore
@ancientsounds

Thanks for helping me think about this!

#AI #AIs #LLM #LLMs #confusion #perceptualDistortion #cognitiveLoad #TextualMirror #Mataroa #blogpost #pronoun #pronouns #selfreference #bot #bots #chatbot #chatbots #ChatGPT #Mistral #MistralAI #Claude #Anthropic #OpenAI #HuggingFace #consciousness #sentience #thinking #language #linguistics #agenticAI

Своя GPT-like LLM по WH40K с нуля. Часть 1: токенизируем Империум

Привет, Хабр! Меня зовут Владимир, и я давно хотел погрузиться во вселенную Warhammer40K. Для погружение во вселенную я решил обучить LLM на лоре Warhammer40K. И чтобы было интереснее, код этой LLM я решил написать сам, на голом pytorch (ну почти). Данная статья - первая в цикле статей по созданию и обучению GPT-like LLM с нуля. В них я постараюсь рассказать, как работают Decoder-only модели, как обучить небольшую LLM, сколько это заняло у меня времени.

https://habr.com/ru/articles/1056212/

#llm #gpt #transformers #pytorch #nlp #bpe #huggingface #dataset #pretraining #python

Inside the CrewAI GitHub repository: role-playing agents meet event-driven control, and observability heats up vs LangSmith. What that

https://aistory.news/open-source-ai/inside-the-crewai-github-repository-and-its-agent-strategy/

#AmazonAI #FineTuning #HuggingFace

Hugging Face Models on Foundry Managed Compute

https://huggingface.co/blog/microsoft/foundry-managed-compute

#ai #foundry #huggingface #aimodels #microsoftfoundry #cloud #azure

Hugging Face Models on Foundry Managed Compute

https://huggingface.co/blog/microsoft/foundry-managed-compute

#ai #foundry #huggingface #aimodels #microsoftfoundry #cloud #azure

CrewAI Flows add event-driven control to role-playing agents, pairing autonomous Crews with production tracing. See how it stacks up to

https://aistory.news/open-source-ai/crewai-flows-bring-event-driven-order-to-ai-role-play/

#AmazonAI #FineTuning #HuggingFace

أصدرت OpenClaw تحديثات رئيسية لتطبيقها المحمول، حيث قدمت لمستخدمي iOS إعادة تصميم شاملة تضمنت ميزات جديدة للدردشة والمكالمات والصور. في المقابل، استفاد مستخدمو أندرويد من خيارات توطين محسّنة ودعم كامل لبوابة .local. الأهم من ذلك، أصبحت OpenClaw قابلة للنشر كتطبيق محلي على منصة Hugging Face، مما يتيح للمستخدمين تشغيل وكلاء استدعاء أدوات خاصة بالكامل ومحلية، مستفيدين من نماذج GGUF أو MLX لتحقيق أقصى درجات الخصوصية والكفاءة.

#OpenClaw #HuggingFace

أصدرت Mistral تحديثاً رئيسياً لوكيلها البرمجي مفتوح المصدر Leanstral 1.5، المصمم لمستخدمي Lean 4 لتحسين أدوات التحقق الرسمي. يتميز الإصدار الجديد بـ 119 مليار معلمة ويقدم دفعة ملحوظة في أداء التحقق من الكود، محققاً نتائج متطورة في معايير إثبات متعددة. كما أثبت فعاليته في اكتشاف الأخطاء وتأكيد خصائص الكود المعقدة. يتوفر Leanstral 1.5 بالكامل كمصدر مفتوح للتنزيل على Hugging Face وعبر واجهة برمجة تطبيقات مجانية، مما يسهل تبنيه.

#Leanstral #Mistral #HuggingFace

The CrewAI multi-agent framework formalizes role-playing agents and event-driven Flows, signaling a shift to production tooling and ops.

https://aistory.news/open-source-ai/inside-the-crewai-multi-agent-frameworks-push-to-prod/

#AmazonAI #FineTuning #HuggingFace

New week, new slides: Run LLMs Locally

Added OpenAI's Privacy-Filter model using privacy-filter.cpp.
New slides about economics and hardware used by the Hugging Face community.

https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf

#ai #llm #llamacpp #wllama #stablediffusion #qwen #localai #gemma4 #webgpu #opencode #huggingface #privacy #digitalsovereignty

New week, new slides: Run LLMs Locally

Added OpenAI's Privacy-Filter model using privacy-filter.cpp.
New slides about economics and hardware used by the Hugging Face community.

https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf

#ai #llm #llamacpp #wllama #stablediffusion #qwen #localai #gemma4 #webgpu #opencode #huggingface #privacy #digitalsovereignty

RT @RoundtableSpace: Die Hugging-Bucht, die Pirate-Bucht für offene LLMs, Modellgewichte zum Download mit Torrents. Das ist ein Game-Changer. https://huggingbay.xyz

mehr auf Arint.info

#AI #HuggingFace #LLM #MachineLearning #OpenSource #Torrents #arint_info

https://x.com/RoundtableSpace/status/2073840584955994527#m

RT @RoundtableSpace: Die Hugging-Bucht, die Pirate-Bucht für offene LLMs, Modellgewichte zum Download mit Torrents. Das ist ein Game-Changer. https://huggingbay.xyz

mehr auf Arint.info

#AI #HuggingFace #LLM #MachineLearning #OpenSource #Torrents #arint_info

https://x.com/RoundtableSpace/status/2073840584955994527#m

CrewAI AMP Suite brings an enterprise control plane to its open-source agent framework. See how it compares with LangSmith and why ops

https://aistory.news/open-source-ai/crewai-amp-suite-brings-an-enterprise-control-plane/

#AmazonAI #FineTuning #HuggingFace

HuggingFace Space のデモをローカルに移行してみた。コード改修はたった1行だった
https://qiita.com/engchina/items/f9d45a1581d7de62eaf2?utm_campaign=popular_items&utm_medium=feed&utm_source=popular_items

#qiita #AI #PyTorch #huggingface #gradio #生成AI

The European AI approach pairs the AI Act with new compute, capital and skills programs. Here’s how the two-track plan changes the playbook

https://aistory.news/open-source-ai/inside-the-european-ai-approach-rules-tools-and-scale/

#AmazonAI #FineTuning #HuggingFace

RT @MiaAI_lab: Ich werde die neue @NVIDIAAI Nemotron-3-Nano-30B-A3B ausprobieren und sie mit Qwen 3.6 35B in agentic Workflows vergleichen. huggingface.co/nvidia/Nemotr… Link nvidia/Nemotron-Labs-TwoTower-30B-A3B-Base-BF16 · Hugging Face Wir sind auf einer Reise, um künstliche Intelligenz durch Open Source und Open Science voranzubringen und zu demokratisieren. huggingface.co

mehr auf Arint.info

#AI #HuggingFace #MachineLearning #Nemotron #NVIDIA #OpenSource #arint_info

https://x.com/MiaAI_lab/status/2072402489186414623#m

RT @MiaAI_lab: Ich werde die neue @NVIDIAAI Nemotron-3-Nano-30B-A3B ausprobieren und sie mit Qwen 3.6 35B in agentic Workflows vergleichen. huggingface.co/nvidia/Nemotr… Link nvidia/Nemotron-Labs-TwoTower-30B-A3B-Base-BF16 · Hugging Face Wir sind auf einer Reise, um künstliche Intelligenz durch Open Source und Open Science voranzubringen und zu demokratisieren. huggingface.co

mehr auf Arint.info

#AI #HuggingFace #MachineLearning #Nemotron #NVIDIA #OpenSource #arint_info

https://x.com/MiaAI_lab/status/2072402489186414623#m

The AI Continent Action Plan signals Europe’s shift from rules to build: AI Factories, InvestAI, CADA, and an AI Skills Academy aimed at

https://aistory.news/open-source-ai/inside-the-ai-continent-action-plan-europe-shifts-to-build/

#AmazonAI #FineTuning #HuggingFace

RT @MiaAI_lab: Ich werde die neue @NVIDIAAI Nemotron-3-Nano-30B-A3B ausprobieren und sie mit Qwen 3.6 35B in agentic Workflows vergleichen. huggingface.co/nvidia/Nemotr… Link nvidia/Nemotron-Labs-TwoTower-30B-A3B-Base-BF16 · Hugging Face Wir sind auf einer Reise, um künstliche Intelligenz durch Open Source und Open Science zu fördern und demokratisieren. huggingface.co

mehr auf Arint.info

#AI #HuggingFace #MachineLearning #Nemotron #NVIDIA #OpenSource #arint_info

https://x.com/MiaAI_lab/status/2072402489186414623#m

RT @MiaAI_lab: Ich werde die neue @NVIDIAAI Nemotron-3-Nano-30B-A3B ausprobieren und sie mit Qwen 3.6 35B in agentic Workflows vergleichen. huggingface.co/nvidia/Nemotr… Link nvidia/Nemotron-Labs-TwoTower-30B-A3B-Base-BF16 · Hugging Face Wir sind auf einer Reise, um künstliche Intelligenz durch Open Source und Open Science zu fördern und demokratisieren. huggingface.co

mehr auf Arint.info

#AI #HuggingFace #MachineLearning #Nemotron #NVIDIA #OpenSource #arint_info

https://x.com/MiaAI_lab/status/2072402489186414623#m

🔭 80TB of astrophysics data from 30+ surveys just dropped on Hugging Face — and it fits on a laptop.

Inspired by a post from Georgia Channing (@cgeorgiaw on X), we reproduced the demo in Groovy 6's #groovysh: galaxy imagery + SDSS spectra from the Multimodal Universe, pulled over HTTP and rendered inline in the terminal with the /img command.

#groovylang #JVM #Astronomy #HuggingFace #DataScience @TheASF
https://groovy.apache.org/blog/multimodal-universe-with-groovy

🔭 80TB of astrophysics data from 30+ surveys just dropped on Hugging Face — and it fits on a laptop.

Inspired by a post from Georgia Channing (@cgeorgiaw on X), we reproduced the demo in Groovy 6's #groovysh: galaxy imagery + SDSS spectra from the Multimodal Universe, pulled over HTTP and rendered inline in the terminal with the /img command.

#groovylang #JVM #Astronomy #HuggingFace #DataScience @TheASF
https://groovy.apache.org/blog/multimodal-universe-with-groovy

RT @Ali_TongyiLab: Du liebst Qwen3.6-27B? NVIDIA hat gerade die ultra-effiziente NVFP4-Version veröffentlicht! Für alle, die Qwen3.6-27B nutzen möchten, aber nicht genügend VRAM zur Verfügung haben, hat NVIDIA seinen offiziellen NVFP4-quantisierten Build auf Hugging Face veröffentlicht. Das große Upgrade ist das NVFP4-4-Bit-Float-Format, das enorme VRAM-Einsparungen durch eine Verkleinerung des Modells auf etwa ein Drittel der Größe der standardmäßigen BF16-Gewichte bietet. URL👇

mehr auf Arint.info

#AI #HuggingFace #MachineLearning #NVFP4 #NVIDIA #Qwen3 #arint_info

https://x.com/Ali_TongyiLab/status/2072162006359240908#m

Ornith-1.0, a new open-source coding model family from DeepReinforce is not just about coding, Instead the model writes its own scaffold. At every training step, it looks at the task in front of it and the scaffold it used last time, then proposes a better version of that scaffold before even attempting an answer. The reward doesn’t just grade the solution. It grades the scaffold that produced it.
https://firethering.com/ornith-1-0-ai-model-for-agentic-coding/
#AI #technews #Claude #ornith #trending #opensource #huggingface #genai

DeepSeek has documented a new inference acceleration framework that it claims greatly increases the efficiency of how LLMs are run.

https://www.computing.co.uk/news/2026/ai/deepseek-claims-new-technique-boosts-llm-serving-efficiency-by-up-to-85?utm_source=mastodon_org&utm_medium=post&utm_campaign=June_Deepseek

#deepseek #ai #technews #Dspark #llm #huggingface #opensource

So, I’ve been re-evaluating my local #LLM #AI implementation on my computer…and I’m realizing that I never configured #JanAI to use my #GPU! Now responses are way faster, so I’ll probably go out and get a more powerful model from #Huggingface.

RT @jun_song: Wir präsentieren SuperGemma4-12b-abliterated 🚀 Das beste Modell für kleinere Hardware 🔥 abliterated (ungefiltert) nachtrainiert mit Super-Tune verbesserte allgemeine Intelligenz. Verfügbar in BF16, GGUF, MLX, NVFP4. Hugging Face ⬇️

mehr auf Arint.info

#HuggingFace #KünstlicheIntelligenz #LLM #MachineLearning #OpenSourceAI #SuperGemma #arint_info

https://x.com/jun_song/status/2070068065757544601#m