Hướng dẫn tinh chỉnh mô hình Qwen2.5-Coder-1.5B cho phân tích cảm xúc tiếng Trung. Có thể chạy trên Google Colab miễn phí trong 20-30 phút. Độ chính xác tăng từ 91,6% lên 97,8%. #AI #MachineLearning #Qwen2.5 #PhânTíchCảmXúc #GoogleColab #TinhChỉnhMôHình #TríTuệNhânTạo #HọcMáy

https://i.redd.it/7xx856mftfzf1.png

Ubuntu erleichtert Einsatz von KI Modellen mit neuen Inference Snaps https://fosstopia.de/ubuntu-ki-modelle-inference-snaps/ #DeepSeekR1 #Qwen2.5VL #Ubuntu #UbuntuAI #UbuntuDeepSeekR1 #UbuntuKI #UbuntuKIModelle #UbuntuQwen2.5VL

Ch peque nhá! Tôi vừa chuyển sang dùng Qwen2.5 Code Instruct bản tự-host thành công! M المقابل với Claude đầu tiên (lần nào 1h phải chờ), Qwen2.5 có thể xử lý comuni code, debug, và nhiếp ý nhanh lùi ởстром đường công việc. Ưbrochen ở máy MBook Pro 48GB và PC 2x RTX 5060TI 16GB (không cần quantize). Cài đặt đơn giản, chất lượng tốt cho công việc lẻ lậu.
Tham khảo GitHub: @reliableJARED/qwen_coder
Tags: #AI #Qwen2.5 #CodeAssistant #LocalTech #MáyTínhLâu
#TechTips #OfflineAI #DevelopersCommu

🧠 #ByteDance ha rilasciato UI-TARS-1.5, un agente multimodale basato su #Qwen2.5-VL-7B che unisce visione e linguaggio con "reasoning". 

👉 I dettagli: https://www.linkedin.com/posts/alessiopomaro_bytedance-qwen2-claude-activity-7321413516488286208-YtOI

___ 

✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: https://bit.ly/newsletter-alessiopomaro 

#AI #GenAI #GenerativeAI #IntelligenzaArtificiale #LLM 

​Qwen2.5-VL & QVQ-Max: Neue Maßstäbe in der visuellen KI

Fortschrittliche Bild- und Videoanalyse
Präzise Objekterkennung
Verbesserte Dokumentenverarbeitung

#ai #ki #artificialintelligence #kuenstlicheintelligenz #Qwen2.5-VL #QVQ-Max

Jetzt lesen und folgen!

https://kinews24.de/qwen2-5-vl-qvq-max/

Alibaba Cloud shakes up the AI scene with **Qwen2.5-Omni-7B!** This cutting-edge multimodal model processes text, images, audio, and video, making it perfect for mobile devices. It's designed for cost-effective AI agents, especially in voice applications for the visually impaired. With a hefty **$53 billion** investment in AI and cloud infrastructure, Alibaba is positioning itself for success in the booming AI market—don’t miss the full story. [Read more](https://www.cnbc.com/2025/03/27/alibaba-launches-open-source-ai-model-for-cost-effective-ai-agents.html) #ArtificialIntelligence #AlibabaCloud #Qwen2 #TechInnovation

Qwen2.5-VL-32B: because nothing says "cutting-edge" like moaning about parameter scales and reinforcement learning 🙄. Apparently, this 32B thing is "smarter" and "lighter" – sounds like a diet ad for AI models. 😂🍩 #Innovation!
https://qwenlm.github.io/blog/qwen2.5-vl-32b/ #Qwen2.5VL32B #AIModels #ReinforcementLearning #CuttingEdge #TechHumor #HackerNews #ngated

Qwen2.5-VL-32B: Smarter and Lighter

https://qwenlm.github.io/blog/qwen2.5-vl-32b/

#HackerNews #Qwen2.5VL32B #Smarter #Lighter #AI #Technology #Innovation

OLMo 2 32B offers unprecedented transparency in #LLM development:

• 🚀 State-of-the-art results: Outperforms GPT3.5, GPT4o-mini, matches top open-weight models like #Qwen2.5 and approaches #Llama3

#AI2 releases OLMo 2 32B, trained on 6T tokens with #Tulu3.1 post-training. Matches or exceeds GPT3.5 Turbo while using just 1/3 the compute of #Qwen2.5 32B. Complete open recipe includes data, code, weights and training methodology.

Testing Deepseek-r1 on Ollama:
https://www.glukhov.org/post/2025/02/deepseek-r1-on-ollama/
#Deepseek-r1 #Ollama #qwen2.5 #llama31

Un grupo de #investigadores dicen haber creado una #IA #AI tan #buena como las de #OpenAI y #DeepSeek por 50 #dólares , inspirada de l modelo #S1 de #Qwen2.5-32B. Y el #dato es #real

https://www.xataka.com/robotica-e-ia/tormenta-desencadenada-deepseek-va-a-entrenamiento-modelo-s1-ha-costado-solo-50-dolares

Really going down the #Ollama rabbit hole. #VScode with #qwen2.5-coder, #deepseek-r1 and #localaipilot for fully contained local chat and code completion. Interested in hearing what others are doing for locally hosted ai.

Alibaba's nieuwe ai-model qwen2.5-max: een krachtige uitdager voor deepseek-v3 en chatgpt-4o https://www.trendingtech.news/trending-news/2025/01/51946/alibaba-s-nieuwe-ai-model-qwen2-5-max-een-krachtige-uitdager-voor-deepseek-v3-en-chatgpt-4o #Alibaba #Qwen2.5-Max #AI-model #DeepSeek-V3 #ChatGPT-4o #Trending #News #Nieuws

Alibaba's nieuwe ai modellen controleren pc's en mobiele apparaten https://www.trendingtech.news/trending-news/2025/01/51242/alibaba-s-nieuwe-ai-modellen-controleren-pc-s-en-mobiele-apparaten #Alibaba #AI-modellen #Qwen2.5-VL #softwarecontrole #beeldanalyse #Trending #News #Nieuws

🎯 #OpenSource Language Model Platform Launch

🔧 Leverages #vLLM technology with custom #GPU scheduler for running various #LLM models
🤖 Supports major models: #Llama3 (405B/70B/8B), #Qwen2 72B, #Mixtral, #Gemma2, #Jamba15, #Phi3

https://glhf.chat/

Advanced Reasoning Model: #ai #llm Marco-o1 Pushes Boundaries in Problem-Solving 🧠

🔬 Built on #Qwen2, focusing on open-ended reasoning beyond traditional tasks

💡 Key Innovations:
🤔 #ChainOfThought fine-tuning for structured reasoning
🌳 Monte Carlo Tree Search (#MCTS) for solution space exploration
🔄 Novel reflection mechanisms for self-improvement
🎯 Multiple action granularities for complex problem-solving

📊 Performance Highlights:
📈 +6.17% accuracy on MGSM English dataset
📈 +5.60% accuracy on MGSM Chinese dataset
🌐 Excels in translation tasks, especially with colloquial expressions

🛠️ Technical Features:
• Fine-tuned on 60,266 training samples
• Implements step & mini-step MCTS strategies
• Utilizes confidence scoring for path selection
• Incorporates self-reflection mechanisms

⚡️ Project Status: Research work in progress with continuous optimization
https://github.com/AIDC-AI/Marco-o1

Edge-Ready #Vision Language Model Advances Visual #AI Processing 🌟

🧠 #OmniVision (968M params) sets new benchmark as world's smallest #VisionLanguageModel

🔄 Architecture combines #Qwen2 (0.5B) for text & #SigLIP (400M) for vision processing

💡 Key Innovations:
• 9x token reduction (729 → 81) for faster processing
• Enhanced accuracy through #DPO training
• Only 988MB RAM & 948MB storage required
• Outperforms #nanoLLAVA across multiple benchmarks

🎯 Use Cases:
• Image analysis & description
• Visual memory assistance
• Recipe generation from food images
• Technical documentation support

Try it now: https://huggingface.co/spaces/NexaAIDev/omnivlm-dpo-demo
Source: https://nexa.ai/blogs/omni-vision

New Cloud Platform for Large Language Model Deployment 🚀

🔧 Run any #opensource #LLM supported by #vLLM on autoscaling #GPU clusters, supporting models up to 640GB VRAM

🤖 Compatible with major models: #Llama3 405B/70B/8B, #Qwen2 72B, #Mixtral 8x22B, #Gemma2 27B, #Phi3, and more

💻 Features include:
- #OpenAI compatible #API
- Custom-built #GPU scheduler
- Support for full-weight and 4-bit AWQ repos
- Multi-tenant architecture for cost efficiency

🆓 Currently free during beta phase, promising competitive pricing post-launch

https://glhf.chat/landing/home

Pamēģināju jauno #AI #LLM #Qwen2.5 #coder, tas, kurš 32B. Nav slikti, bet izskatās, ka manām / www.MediaBox.lv vajadzībām pieteik arī ar 14B. Smuki izlaboju dažas nepilnības, kas nebija īsti kļūdas, bet labās prakses ievērošana strukturējot #PHP kodu.
Bail domāt par dāvanu sev - jaunu grafisko karti un visu ar to saistīto, lai šajā mestos iekšā vēl dziļāk...
#programmēšana #EsNoLaukiem