Anderson Chaves

⭐️ New video release 📺: Refactoring Large Programs Watch Dr. Kristian Rother guide you through refactoring a large, messy, and undocumented Python program, offering valuable insights and strategies for your next big refactoring challenge. 📺 Watch the video on YouTube: https://lnkd.in/emjn58cb Kristian, a freelance Python trainer, presented a tutorial on refactoring large programs at a recent event. He addressed the challenge of cleaning up complex software with 10,000-100,000 lines of code, particularly in the case of legacy code. The tutorial focused on practical procedures such as code review, writing minimal tests, adding type annotations, and isolating cleanable parts from problematic sections. Kristian emphasized the systematic approach to large-scale refactoring, highlighting that while it is difficult and costly, it can be tackled methodically. The session, aimed at those with fluency in basic Python, included interactive elements and practical exercises for attendees to enhance their refactoring skills. Participants were provided with the messy code and refactoring guidelines through GitHub for hands-on learning.

1

VentureBeat reports Meta's AI according to Yann LeCunn quickly becoming the most used AI assistant https://lnkd.in/eHWPev5w. #meta #aiassistant #openai #chatgpt #yannlecunn #venturebeat

1

Forget about LLMs Ranking Leaderboards🥇, Welcome to the Era of Model Routing 🔀! This shift in the market is truly transformative. The idea of dynamically selecting the most suitable model for each specific task is a game-changer. It's like having a garage full of specialized vehicles, each ready to take on a different terrain. Whether it's the speed of a Ferrari 🏎 for complex technical analyses or the reliability of a Range Rover 🚗 for varied tasks, model routing ensures the optimal performance for every scenario. Amazon Q for Business is a perfect example, dynamically choosing the best model based on query complexity to maximize user experience. This approach heralds a future where model selection is flexible, task-specific, and efficient, marking a significant departure from static model use. Exciting times ahead in the world of generative AI! #AI #MachineLearning #ModelRouting #Innovation #TechRevolution #FutureOfAI #GenerativeAI #SmartTech 🚀🔍🧠

1

📰 Enabling Diffusion Model for Conditioned Time Series Generation ✍ Frédéric Montet, Benjamin Pasquier, Beat Wolf and Jean Hennebert Synthetic #timeseries generation is an emerging field of study in the broad spectrum of data science, addressing critical needs in diverse fields such as finance, meteorology, and healthcare. In recent years, #diffusion methods have shown impressive results for image synthesis thanks to models such as Stable Diffusion and DALL·E, defining the new state-of-the-art methods. In time series generation, their potential exists but remains largely unexplored. In this work, we demonstrate the applicability and suitability of diffusion methods for time series generation on several datasets with a rigorous evaluation procedure. Our proposal, inspired from an existing diffusion model, obtained a better performance than a reference model based on generative adversarial networks (GANs). We also propose a modification of the model to allow for guiding the generation with respect to conditioning variables. This conditioned generation is successfully demonstrated on meteorological data. 🔗 Read the paper at https://lnkd.in/gMtseAuQ. #mdpiproceedings #ITISE2024 #forecasting

1

[INTRODUCTION TO GENERATIVE AI] 🚀 Thrilled to share the key insights from our recent Generative AI presentation! Our Data Scientist khadija hader delivered an engaging session, focusing on: 🌟 Generative AI Overview: Unpacking core principles and innovative capabilities. 🔍 Technical Deep Dive: Exploring the architecture and mechanics of AI models. 🏥 Healthcare Impact: Showcasing AI's role in personalized treatments, enhanced imaging, and Annotations progress. 🎯 The presentation concluded with a forward-looking perspective on the future of Generative AI in healthcare. We discussed emerging trends, ethical considerations, and the collaborative efforts needed to harness AI’s full potential while ensuring patient safety and data privacy. 🙏A huge thank you to everyone who attended and participated in the discussion! Your insights and questions made the session truly enriching. 💬 If you're passionate about AI and its applications in healthcare, let's connect and continue the conversation! Feel free to share your thoughts and experiences in the comments below. #GenerativeAI #HealthcareInnovation #TechTalk #FutureOfHealthcare

30

Unlocking the Power of Llama 3: Efficient Fine-Tuning with FSDP QDoRA ... In a new blog article, "Efficient finetuning of Llama 3 with FSDP QDoRA," the team at Answer.AI has introduced a novel method for fine-tuning large language models (LLMs) that is both memory-efficient and highly scalable. This development has the potential to democratize access to state-of-the-art AI, enabling more developers and researchers to improve and customize LLMs for specific tasks. 👉 QDoRA: Making AI Accessible The key innovation, QDoRA (Quantized Weight-Decomposed Low-Rank Adaptation), allows for the fine-tuning of massive models like Llama 3 on less powerful hardware, such as gaming GPUs. This is a significant step forward in making cutting-edge AI more accessible to a broader range of developers and institutions. QDoRA's memory efficiency is truly impressive. It enables the fine-tuning of models with hundreds of billions of parameters on commonly available hardware, a feat that was previously out of reach for many in the AI community. 👉 Performance That Rivals Full Fine-Tuning Not only is QDoRA more memory-efficient, but it also delivers performance that matches or exceeds full fine-tuning. The paper shares compelling results demonstrating QDoRA's competitive performance when fine-tuning Llama 3. 👉 Llama-Pro: Enhancing Model Specialization In addition to QDoRA, the paper also introduces Llama-Pro, a method for progressively expanding LLMs by adding new Transformer blocks. This technique enhances model specialization without losing existing capabilities, making it particularly useful for domain-specific applications where deep expertise is required. 👉 The Future of AI Model Training QDoRA and Llama-Pro represent a significant step forward in the field of AI model training. As more developers and researchers adopt these techniques, we can expect to see a surge in AI innovation and the development of highly specialized, task-specific models. Moreover, this research opens up exciting possibilities for the continuous improvement of open-source models. As the AI community collaborates and builds upon these advancements, we may witness an acceleration in the pace of AI development and deployment. 👉 Explore the Possibilities I invite you to delve deeper into this blog article and consider how it could be applied in your own work. The potential for QDoRA and Llama-Pro to transform various industries is immense, and I am excited to see how the AI community will build upon these advancements. To stay updated on the latest developments in AI research, I recommend following the authors and institutions behind this paper. And if you found this post informative, please share it with your network to foster broader discussions about the future of AI model training.

56

5 commentaires

Hey everyone 👋 Just shared two new blog posts on key topics in AI and data extraction: 1️⃣ Simple Techniques for Extracting Table Data from PDFs (part one)📤📊 Medium : https://lnkd.in/efWknuPz My webbook : https://lnkd.in/evGxTYAv 2️⃣ Serving Large models (part two): Ollama and TGI 🦙🚀 Medium : https://lnkd.in/ebABt_f7 My Webbook : https://lnkd.in/ecEsHy8b #Ollama #TGI #RAG #LLMs #AI

38

3 commentaires

𝗨𝗻𝗹𝗼𝗰𝗸 𝗻𝗲𝘅𝘁-𝗹𝗲𝘃𝗲𝗹 𝗔𝗜 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲𝘀𝗲 𝘁𝗼𝗽 𝗟𝗟𝗠 𝘁𝗼𝗼𝗹𝘀! 🤖 Multi-agent applications: LangGraph. ✨ Advanced Prompt Engineering: DSPy. 🔧 LLMOps: LangSmith. 🛠️ Fine-Tuning: LlamaIndex, HF Sentence Transformers. 🚀 LLM serving & inference: vLLM. 📊 Vector DB: QDrant. 💬 Chat model: GPT-4o-mini. 🧠 Embedding model: text-3-embedding-small. 🗣️ Open chat model: Llama 3 8B Instruct. 🔍 Open embeddings: Arctic Embed M. Follow AI Makerspace for insights and a deep dive into these tools! #MachineLearning #GenAI #LLMs #RAG #TechInnovation #AICommunity #LLMOps

12

4 commentaires

Finding Diamonds in the Rough: How Unexpected Data Can Spark Innovation In many companies, design projects are increasingly "data-driven," meaning they are based on insights gleaned from large datasets. Antoine Bordas; Pascal Le Masson; and Weil, Benoit from Mines Paris, PSL University explore how seemingly anomalous or unexpected data points can be used to guide the design process (via ICED23 Proceedings: DS 122). https://lnkd.in/gdDFBCJp Join the "Design is a Team Sport" movement at ICED25 Dallas. How can we effectively use data anomalies to drive innovation? What are the challenges of incorporating data science into the design process? Share your thoughts on the future of data-driven design. Don't miss out!  · Register for ICED25 Dallas: iced.designsociety.org/ The University of Texas at Dallas, The Design Society #DataScience #DesignThinking #Innovation #DataAnomalies #DesignProcess #ICED25

7

Absolutely fantastic news! 🎉 The launch of the Hugging Face Embedding Container for Amazon SageMaker opens up a world of possibilities for embedding creation, particularly for Retrieval-Augmented Generation (RAG) applications. It's incredible to see the support for popular architectures like BERT, RoBERTa, XLM-RoBERTa, Nomic AI Bert, and Jina AI Bert, ensuring flexibility and compatibility. Powered by Text Embedding Inference (TEI) for lightning-fast inference and available for both CPU and NVIDIA GPU instances, this container truly takes embedding creation to the next level. And the ability to seamlessly deploy open embedding models like Snowflake Arctic, BAAI BGE, and Jina AI further expands the horizon. Can't wait to dive into the blog post detailing how to deploy these models on Amazon SageMaker! #MachineLearning #DeepLearning #AWS #SageMaker #HuggingFace #TechInnovation 🚀🤖🔍

2

1 commentaire

Want to master LLM fine-tuning with HuggingFace? Check out our latest video with 5 quick, essential tips to elevate your models! For more insights like these, don’t miss the 𝐆𝐫𝐚𝐝𝐢𝐞𝐧𝐭 𝐁𝐨𝐨𝐬𝐭 newsletter—your go-to for cutting-edge data science guidance 👉 https://lnkd.in/epDfMFv8 #DataScienceCommunity #LargeLanguageModels #LLM #DrDataScience #GradientBoost

1

"Stay on topic with Classifier-Free Guidance" by Guillaume Sanchez will be published at [ICML] Int'l Conference on Machine Learning 2024 ! Classifier-Free Guidance is an inference time technique which gives more importance to the user intent, expressed through the prompt. The hypothesis is: focusing more on the prompt at inference-time will result in generations that better align with expected behavior. CFG is improving alignment on a wide range of prompting approaches including zero-shot prompting, Chain-of-Thought prompting, long-form generative prompting and complex chatbot-style prompting. Guillaume's paper shows that for the same inference cost, one can train a model that is half the size and obtain similar performance. As our mission is delivering custom-made models, powerful yet precisely tailored to the customer needs, the integration of the CFG seemed obvious. This is why we proposed to Guillaume to join LightOn as Head of R&D. Congratulation Guillaume Sanchez, Stella Biderman, Alexander Spangher, Honglu F. and EleutherAI for the paper's publication at ICML 2024 🔗 ArXiv link in comment 👇 #LightOn #ArtificialIntelligence #GenAI #LLM #ICML2024

39

3 commentaires

🤖 Deep Dive Workshop D: Collaboration Between Agents – Techniques for Multi-Agent LLMs and Their Operational Impact Join Donatien Chedom Fotso, Ph.D., AI & ML Team Lead at Deutsche Bank, at Generative AI EUROPE 2024 for an in-depth exploration of how multi-agent LLM systems can revolutionize business operations. Key highlights: *Advanced techniques for improving AI agent collaboration and interaction *Real-world case studies showcasing the operational impact of multi-agent systems *Interactive group discussions and scenario-based activities to explore potential applications *Ideal for AI researchers, developers, and tech strategists looking to push the boundaries of AI capabilities. 🚀 👉 Register now and enhance your AI solutions with multi-agent systems: https://lnkd.in/e783QTaV #GenAIEUROPE #AIConference

8

𝐆𝐄𝐍𝐃𝐄𝐑 𝐑𝐄𝐂𝐎𝐆𝐍𝐈𝐓𝐈𝐎𝐍 | 𝐀𝐈 | 𝐃𝐄𝐄𝐏 𝐋𝐄𝐀𝐑𝐍𝐈𝐍𝐆 | 𝐈𝐌𝐀𝐆𝐄 𝐂𝐋𝐀𝐒𝐒𝐈𝐅𝐈𝐂𝐀𝐓𝐈𝐎𝐍 Excited to share the details of my new artificial intelligence project focused on gender classification, which identifies whether an individual is male or female with high accuracy. The project includes several key components: Data Collection: Gathered diverse datasets from multiple sources to ensure a wide range of images for model training. Image Preprocessing: Utilized libraries like Numpy and OpenCV2 to preprocess images for consistent quality and format. Transfer Learning: Implemented a pretrained VGG16 model from Keras as a base for feature extraction. Data Augmentation: Applied random rate and contrast adjustments to enhance the training dataset and improve model robustness. Model Architecture: Added a binary output layer with a sigmoid activation function to classify images into male or female categories. Model Training: Trained the model on the prepared dataset, achieving a training accuracy of 98% and a testing accuracy of 94%. Real-Time Detection: Integrated OpenCV2 for video capture, where the model analyzes live video feed to detect and label individuals with appropriate text annotations. #ImageClassification #ImagePreprocessing #GenderRecognition #ComputerVision #ArtificialIntelligence #DeepLearning #ConvolutionalNeuralNetwork ITSOLERA PVT LTD Joinrs Italia Joinrs Italia Joinrs UK Joinrs US Sixlogics WebX Google Huawei Microsoft ZAPTA Technologies (Pvt.) Limited IBM Coursera Codebasics Dhaval Patel Hemanand Vadivel Generative AI OpenAI Digital Empowerment Network CognoRise InfoTech CodeAlpha SkyElectric Pvt. Ltd

34

4 commentaires

🎉 𝐄𝐱𝐜𝐢𝐭𝐢𝐧𝐠 𝐁𝐫𝐞𝐚𝐤𝐭𝐡𝐫𝐨𝐮𝐠𝐡 𝐢𝐧 𝐀𝐈 𝐈𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭𝐚𝐛𝐢𝐥𝐢𝐭𝐲! Last few weeks have seen a surge in AI Interpretability work: Anthropic researchers work on understanding Claude 3 models [Scaling Monosemanticity in AI](https://lnkd.in/eaYkgSNM) OpenAI released work on understanding GPT-4 [Extracting concepts from GPT-4] (https://lnkd.in/gFzijQwM) What's even more fascinating is their use of 𝘚𝘱𝘢𝘳𝘴𝘦 𝘈𝘶𝘵𝘰𝘌𝘯𝘤𝘰𝘥𝘦𝘳𝘴 (𝘚𝘈𝘌𝘴). Curious why SAEs are so effective and scalable? Our research at Martian provides some compelling insights: https://lnkd.in/gkvN2cXG At Martian, we leverage category theory—a mathematical approach focusing on relationships rather than object internals—to understand why SAEs perform so well. This theory underpins our broader efforts in "model mapping," a series of methods that promise enhanced interpretability without intensive manual analysis. 🔬 𝐌𝐨𝐝𝐞𝐥 𝐌𝐚𝐩𝐩𝐢𝐧𝐠: 𝐀 𝐂𝐨𝐦𝐩𝐮𝐭𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐌𝐢𝐜𝐫𝐨𝐬𝐜𝐨𝐩𝐞 𝐟𝐨𝐫 𝐀𝐈 Our model mapping initiative acts like a microscope, revealing how AI models operate without dissecting their inner workings. This scalable approach benefits from increased computational power, paving the way for more efficient AI development. 🚀 𝐉𝐨𝐢𝐧 𝐌𝐚𝐫𝐭𝐢𝐚𝐧 𝐨𝐧 𝐎𝐮𝐫 𝐉𝐨𝐮𝐫𝐧𝐞𝐲! We're pushing the boundaries of AI interpretability and alignment. Learn more and get involved at [withmartian.com](https://withmartian.com).

32