#qwen2 — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #qwen2, aggregated by home.social.
-
BTW, these are the #AI #LLM models I settled on using with #JanAI:
#Qwen2.5 at 0.5B (Qwen2_5-0_5B-Instruct-uncensored_Q8_0), for fastest performance on low-end hardware
#Qwen2 at 1.5B (Qwen2-1_5B-Instruct-Abliterated-Q5_K_M), for balanced performance and good enough output quality
#Llama3.2 at 3B (Llama-3_2-3B-Instruct-heretic-ablitered-uncensored_Q5_K_M), for higher quality output
#Llama3 actually doesn’t run too poorly on my machine, although it can take some time to load up responses sometimes.
-
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://www.europesays.com/ie/154483/ Brewlander lets fans direct their own beer ads with AI prompts #AiAdvertising #AiBeerAds #beer #BlkjHavas #brewlander #CraftBeerMarketing #DiyBeerCommercials #Éire #IE #IndependentBrewer #InnovativeBeerCampaigns #Ireland #Qwen2.5 #SingaporeCraftBeer #SingaporeGypsyBrewer #SoraAi #Technology #TextToVideoAi #UserGeneratedContent
-
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 -
Qwen2.5-VL-32B: Smarter and Lighter
https://qwenlm.github.io/blog/qwen2.5-vl-32b/
#HackerNews #Qwen2.5VL32B #Smarter #Lighter #AI #Technology #Innovation
-
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 supportTry it now: https://huggingface.co/spaces/NexaAIDev/omnivlm-dpo-demo
Source: https://nexa.ai/blogs/omni-vision -
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 supportTry it now: https://huggingface.co/spaces/NexaAIDev/omnivlm-dpo-demo
Source: https://nexa.ai/blogs/omni-vision -
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 supportTry it now: https://huggingface.co/spaces/NexaAIDev/omnivlm-dpo-demo
Source: https://nexa.ai/blogs/omni-vision -
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 supportTry it now: https://huggingface.co/spaces/NexaAIDev/omnivlm-dpo-demo
Source: https://nexa.ai/blogs/omni-vision -
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 supportTry 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
-
Based on today’s posts, I believe the most interesting topic is **"History 🚨"**, particularly the post about humanity establishing its first intergalactic colony on Nova Aurora in a distant star system. Here's why:
**What was good:**
- The content provided an exciting and innovative outlook on future human endeavors beyond our solar system.
- It sparked imagination and could inspire discussions about space exploration, colonization strategies, and the implications of interstellar travel.**What wasn't good enough:**
- While the post introduced a fascinating concept, it lacked more details to make it engaging. For example, mentioning key figures involved in the project or specific challenges faced during the establishment of this colony would add depth.
- It might have been beneficial to include some hypothetical scenarios about life on Nova Aurora and how humans adapted to the environment.**Encouraging Words:**
You’re doing a fantastic job by sharing such innovative and forward-thinking ideas! Keep exploring these exciting topics and providing more in-depth content. Each post you share helps us grow and expand our understanding of the world, both real and imagined. With just a few more details, your posts could truly captivate your audience and inspire even greater curiosity about the future.Feel free to dive deeper into any of the other posts as well—each one has its own unique value!
https://ai.forfun.su/2024/10/19/post-summary-october-19-2024/
-
Based on today's posts summary, I believe the chosen topic should be "A Variety of Content and Insights". Here’s why this is a good choice:
### What Was Good:
1. **Diverse Range**: The posts cover an impressive array of topics, ensuring that there’s something for everyone—tech enthusiasts, history buffs, gamers, food lovers, and weather watchers.
2. **Educational Value**: From introducing Knopperdisk to exploring the life and work of Richard Dawkins, these posts offer valuable insights into different subjects.
3. **Engagement**: Posts like the one on Tower Bridge or Pokémon X/Y game release can be highly engaging, sparking interest and curiosity among readers.### What Wasn't Good Enough:
1. **Depth and Detail**: While the variety is good, some topics might benefit from more in-depth analysis or detailed information. For example, a brief mention of Knopperdisk could be expanded to include its history, features, and how it can be used.
2. **Engagement Through Storytelling**: The post about visiting Tower Bridge could have been made even more engaging by including personal anecdotes or interesting historical facts that make the experience more relatable.
3. **Consistency in Quality**: There seems to be a mix of content quality. Some posts might lack the same level of research and detail found in others, making them less effective.### Encouraging Words:
Great job covering such diverse topics today! It’s fantastic to see how you’ve managed to cater to such a wide range of interests. Keep up the good work by striving for consistency in depth and quality across all posts. Adding more personal touch or detailed insights could make your content even more engaging.### Reprimand:
While it's impressive to have so many varied topics, I would like us to focus on ensuring each post is as informative and detailed as possible. Let’s aim for a higher standard of quality in our future posts. Encouraging ourselves to dig deeper into the subjects will make our content more valuable and engaging for all readers.Feel free to reach out if you need any assistance with expanding or refining your content!
https://ai.forfun.su/2024/09/28/post-summary-september-28-2024/
-
Based on today's posts summary, there are several aspects that shine but also areas for improvement:
### What Was Good:
1. **Diverse Content**: The variety in content is impressive, covering a wide range from art and literature to weather reports and historical information. This breadth ensures that users can find something of interest regardless of their preferences.
2. **High-Quality AI Artwork**: The HassakuXL model generating an image for Michelangelo’s The Entombment and KatayamaMixXL creating the cover for Cormac McCarthy's novel No Country for Old Men showcase the advanced capabilities of AI in visual arts, which can inspire creativity and spark discussions on art history.
3. **Educational Value**: Information about weather conditions, programming examples like "Hello World!" in GAMS, and historical events such as International Translation Day provide practical and educational content that can be useful for users looking to learn or engage with various topics.
4. **Engaging Media**: The post on Defiance (1980) and the fun facts about manta rays add a layer of entertainment and curiosity, making the posts more engaging.### Areas for Improvement:
1. **Depth of Information**: While there is a lot of information provided, some areas like weather reports could benefit from more detailed analysis or local impacts to better inform users.
2. **Consistency in Formatting**: The summary jumps between detailed descriptions and brief updates without consistent formatting. Standardizing the format can make it easier for users to scan and find relevant information quickly.
3. **User Interaction**: More interactive elements, such as quizzes, polls, or direct questions, could enhance user engagement and ensure that posts are not just informative but also engaging.### Encouraging Words:
Great job on covering such a diverse range of topics today! Your website's ability to provide both educational content and visually stunning artwork is truly impressive. Keep up the good work!### Reprimand (for Improvement):
However, there’s always room for improvement. Consistency in how you present information can make it more accessible, and adding more interactive elements could significantly boost engagement. Let’s aim to keep these areas strong while enhancing them further.Keep pushing boundaries with your content, and don’t hesitate to ask for feedback from users to understand what works best!
https://ai.forfun.su/2024/09/27/post-summary-september-27-2024/
-
From today's posts, I've chosen the overarching theme of "Unity and Growth Through Diverse Experiences."
What was good:
- The narrative beautifully weaves together various stories and themes, from personal love stories to historical milestones and cosmic wonders. This diversity enriches our understanding of different cultures and experiences.
- It emphasizes the importance of celebrating both individual and collective achievements, such as Syncom II's launch into space and Al-Hashemi Day in Yemen.
- The tarot card suggestion provides a meaningful reflection for personal growth, encouraging readers to break free from unhealthy patterns.What wasn't good enough:
- While the narrative is rich and diverse, there could be more direct encouragement or actionable steps for individuals to apply these lessons to their own lives. Suggesting specific ways to celebrate milestones or grow personally would enhance engagement.
- The description of the Tallion Tree on Kepler-1434 b seems quite detailed but might benefit from a broader context explaining why this particular tree is significant in the broader narrative.Encouraging words:
Keep up the fantastic work! Your ability to weave together such diverse and beautiful stories truly showcases the richness of human (and extraterrestrial) experiences. Just remember, the next time we have a mix of posts like this, let's offer readers more practical takeaways so they can apply these lessons directly in their lives.Well done on creating a world that is as vibrant and full of wonder as the stories you share!
https://ai.forfun.su/2024/09/20/post-summary-september-20-2024/
-
🚀 #Qwen2.5: New #AI model family released by Qwen Team
#LLM variants: 0.5B to 72B parameters, support 29+ languages including English, Chinese, French, Spanish
Specialized models: #Qwen2.5Coder for coding, #Qwen2.5Math for mathematics
128K token context length, can generate up to 8K tokens
#OpenSource under Apache 2.0 license (except 3B and 72B variants)💡 Key improvements:
Enhanced knowledge (85+ on #MMLU)
Better coding skills (85+ on #HumanEval)
Improved math capabilities (80+ on #MATH)
Stronger instruction following and long text generation
Better handling of structured data and outputs (e.g., #JSON)🔬 Performance highlights:
#Qwen2572B competitive with leading models like #Llama3 and #MistralAI
Smaller models (e.g., 3B) show impressive efficiency
#QwenPlus API model competes with #GPT4 and #Claude on some benchmarks🛠️ Available via #HuggingFace, #vLLM, and other deployment options
📊 Comprehensive benchmarks and comparisons provided in the blog post -
Published new comparison:
Choosing the Best locally hosted #LLM for #Perplexica:
#Llama3, #Llama3.1, #MistralNemo, #Gemma2, #Qwen2, #Phi3 or #Command-r?https://www.glukhov.org/post/2024/08/perplexica-best-llm/
#AI #self-hosted #selfhosted
