Benjamin Dubreu

#DataEngineering 𝗣𝗿𝗼 𝗧𝗶𝗽𝘀 𝗳𝗿𝗼𝗺 𝗙𝗮𝗯𝗿𝗶𝗰𝗲 𝗗𝗲𝘀𝗲𝘆𝗻, 𝗗𝗮𝘁𝗮𝗯𝗿𝗶𝗰𝗸𝘀 𝗖𝗵𝗮𝗺𝗽𝗶𝗼𝗻 Join our colleague, Fabrice Deseyn, for a deep dive into the world of Databricks. In a recent interview with Youssef Mrini on the #NextGenLakehouse YouTube channel, Fabrice shares his insights and journey of becoming a Databricks champion. Beyond his champion status, Fabrice provides expert guidance on navigating Unity Catalog migrations. Discover practical tips and strategies to streamline the process and ensure a successful transition. Watch the video now 👉 https://lnkd.in/e2zFw53p #Databricks

11

1 commentaire

Want to test Meta, Google, and Mistral models all on one platform ? **Groq.com** Why Groq? Because speed matters. GroqCloud supports a diverse array of models, enabling you to select the best for your needs: ➡ llama3-8b-8192 by Meta (Context Window: 8,192 tokens) ➡ llama3-70b-8192 by Meta (Context Window: 8,192 tokens) ➡ mixtral-8x7b-32768 by Mistral (Context Window: 32,768 tokens) ➡ gemma-7b-it by Google (Context Window: 8,192 tokens) Groq's commitment to GenAI inference speed is transforming real-time AI applications with its LPU. What is an LPU? ➡ Language Processing Unit, is a revolutionary processing system designed specifically for LANGUAGE applications. Unlike traditional CPUs (Central Processing Units) and GPUs (Graphics Processing Units), which handle a broad range of tasks. LPUs are optimized for the rapid processing demands of LLMs This specialized architecture allows for more efficient handling of sequential data, significantly reducing the time per word processed. ➡ That means you get faster outputs You can run Llama-2 70B at over 300 tokens per second per user!

24

Data Engineers ! This one's for you: 🧠 The Economics of Generative AI Stephanie Kirmer 🧠 Connecting ML Models in Sagemaker to Snowflake and dbt Hugo Lu 🧠 Introduction to Interpretable Clustering NAKUL UPADHYAY 🧠 Snowflake automatic clustering cost estimation Tim Sander 🧠 Death to Tech Debt? Ari Joury, PhD Check it here; https://lnkd.in/ebdFKNZ5 #dataengineering #dataroundup #startup

5

Check out my latest talk on "Peak Performance in Python" at this year's PyConDE Open to feedback!

19

🚀 Mistral AI: Pixtral Large and Le Chat Platform Introducing Pixtral Large: Mistral AI, the French AI startup, has launched Pixtral Large, a groundbreaking 124-billion parameter multimodal model designed to process both text and visual data. - 🧠 Core Architecture: Combines a 123-billion parameter text decoder with a 1-billion parameter vision encoder. - 📖 Expanded Context Window: Supports up to 128K tokens, enabling it to handle 30 high-resolution images or a 300-page book. - 🏆 Benchmark Leader: Outperforms GPT-4o, Gemini 1.5 Pro, and Claude-3.5 Sonnet on key benchmarks like MathVista (69.4% accuracy), ChartQA, and DocVQA. Le Chat Platform: Unlocking New Possibilities Mistral’s Le Chat platform integrates Pixtral Large, introducing innovative features for diverse applications: -🔍 Web Search with Citations: Real-time data retrieval with source transparency. -🖋️ Canvas: Tools for live document creation, collaborative editing, and version control. -📄 Advanced OCR: Efficiently processes PDFs, tables, and equations to extract insights. - 🎨 Image Generation: Powered by Flux Pro from Black Forest Labs. - 🤖 Task Agents: Automates tasks like summarization and invoice processing. - 🌟 Performance Highlights Pixtral Large sets new standards on benchmarks like ChartQA, DocVQA, and VQAv2, showcasing cutting-edge capabilities in document analysis and visual data interpretation. The model is available under the Mistral Research License for non-commercial use, with commercial licenses for enterprise adoption. 📢 Additional Updates: Mistral AI also unveiled Mistral Large 2.1, a 123-billion parameter model optimized for general-purpose tasks, along with enhanced APIs to streamline developer integration.

6

𝐃𝐚𝐲 7 𝐈𝐧 𝐆𝐞𝐧𝐀𝐈: 𝐄𝐱𝐩𝐥𝐨𝐫𝐢𝐧𝐠 𝐁𝐄𝐑𝐓'𝐬 𝐂𝐨𝐫𝐞 𝐚𝐧𝐝 𝐅𝐢𝐧𝐞-𝐓𝐮𝐧𝐢𝐧𝐠 Day 7 of my Generative AI journey was packed with insights as I wrapped up the BERT paper, "𝐁𝐄𝐑𝐓: 𝐏𝐫𝐞-𝐭𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝐨𝐟 𝐃𝐞𝐞𝐩 𝐁𝐢𝐝𝐢𝐫𝐞𝐜𝐭𝐢𝐨𝐧𝐚𝐥 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐞𝐫𝐬 𝐟𝐨𝐫 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐔𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠" by Devlin et al. I also delved into Umair Jamil’s video, "𝐁𝐄𝐑𝐓 𝐞𝐱𝐩𝐥𝐚𝐢𝐧𝐞𝐝: 𝐓𝐫𝐚𝐢𝐧𝐢𝐧𝐠, 𝐈𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐞, 𝐁𝐄𝐑𝐓 𝐯𝐬 𝐆𝐏𝐓/𝐋𝐋𝐌, 𝐅𝐢𝐧𝐞-𝐭𝐮𝐧𝐢𝐧𝐠, [𝐂𝐋𝐒] 𝐭𝐨𝐤𝐞𝐧", which added another layer of clarity to my understanding. ( Check it out: https://lnkd.in/dGZg6mQw ) 𝐔𝐧𝐩𝐚𝐜𝐤𝐢𝐧𝐠 𝐭𝐡𝐞 𝐁𝐄𝐑𝐓 𝐏𝐚𝐩𝐞𝐫 The architecture and pre-training strategies in BERT are a masterclass in optimization and versatility. Here’s what stood out: → [𝐂𝐋𝐒] 𝐓𝐨𝐤𝐞𝐧: Used for sentence-level tasks, the token aggregates information from the input sequence and is pivotal for classification models. → 𝐌𝐚𝐬𝐤𝐞𝐝 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐢𝐧𝐠 (𝐌𝐋𝐌): This clever approach of masking random tokens teaches BERT to predict context-sensitive words, enabling a bidirectional understanding of language. → 𝐍𝐞𝐱𝐭 𝐒𝐞𝐧𝐭𝐞𝐧𝐜𝐞 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧 (𝐍𝐒𝐏): A task that helps BERT understand sentence relationships, making it effective for tasks like Q&A and summarization. 𝐈𝐧𝐬𝐢𝐠𝐡𝐭𝐬 𝐟𝐫𝐨𝐦 𝐔𝐦𝐚𝐢𝐫 𝐉𝐚𝐦𝐢𝐥’𝐬 𝐕𝐢𝐝𝐞𝐨 Umair Jamil’s explanation bridged the gap between theory and practice. He not only demystified BERT’s architecture but also touched on: → 𝐓𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝐓𝐞𝐜𝐡𝐧𝐢𝐪𝐮𝐞𝐬: How large-scale datasets and compute are utilized for pre-training. → 𝐅𝐢𝐧𝐞-𝐓𝐮𝐧𝐢𝐧𝐠: The ease of adapting BERT for tasks like sentiment analysis and named entity recognition by tweaking the final layers. → 𝐂𝐨𝐦𝐩𝐚𝐫𝐢𝐬𝐨𝐧𝐬: Distinguishing BERT from GPT and other LLMs, emphasizing its pre-training methodology and bidirectional nature. And that was it for today 🤷‍♂️ 𝐖𝐡𝐚𝐭’𝐬 𝐍𝐞𝐱𝐭? Now it’s time to put theory into practice. I'll try implementing the BERT tomorrow. #GenAI #BERT_learnings

10

3 commentaires

Hugging Face x LangChain : A new partner package in LangChain - We are thrilled to announce the launch of langchain_huggingface, a partner package in LangChain jointly maintained by Hugging Face and LangChain. This new Python package is designed to bring the power of the latest development of Hugging Face into LangChain and keep it up to date. langchain-huggingface integrates seamlessly with LangChain, providing an efficient and effective way to utilize Hugging Face models within the LangChain ecosystem. This partnership is not just about sharing technology but also about a joint commitment to maintain and continually improve this integration. Getting started with langchain-huggingface is straightforward. Here’s how you can install and begin using the package: pip install langchain-huggingface #llm #huggingface #langchain #genai #generativeai #artificialintelligence

25

𝐍𝐞𝐰 𝐀𝐈 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧𝐚𝐥𝐢𝐭𝐲 𝐟𝐫𝐨𝐦 𝐬𝐧𝐨𝐰𝐟𝐥𝐚𝐤𝐞 ❄️ No I don’t mean the new #snowflakeartic #LLM. It’s even way much cooler than that. Take a look at #classification!! Heuhhhh, classification? Yes classification!! Now in public preview in #snowflake, classification functionality that you can directly use in SQL code to train gradient boosting algorithms for binary or multi-class classification. In my opinion, companies are still not making full use of classification use cases. For example, response modeling, churn modeling, propensity to {filler}. I also think these use cases are far more present at companies than #LLM or #genai use cases. The data for these use cases are often already in data warehouses like snowflake. I.e. customer details, customer transactions, canceled subscriptions, etc… The new snowflake classification functionality allows you to train a #MLmodel directly on data in snowflake without the need for separate python environments. Moreover, the model that is trained is just a snowflake object that can be used to score / predict new data. #snowflake_is_cool See for detail and examples 👉

29

4 commentaires

📢 New to Version Control on Databricks? Start strong with our beginner’s guide! In this one, our Head of AI, Gavi, is joined by Stuart from Databricks, diving into everything you need to get started with Version Control - setting up with Azure DevOps, mastering feature branches, navigating pull requests, and more. Packed with practical tips, this video is perfect for anyone looking to implement version control effectively! Check it out - we’d love to hear your thoughts! 👇 https://hubs.la/Q02W6d4m0 #Databricks #VersionControl #AzureDevOps #DataScience #AI

39

2 commentaires

So many insightful informations through Databricks' system tables : https://lnkd.in/e3pP3Pre. Even if you enabled the preview feature on Unity, don't forget to enable each schema through API: curl -v -X PUT -H "Authorization: Bearer <PAT Token>" "https://<workspace>.https://lnkd.in/eeZPjAX3<​metastore-id>/systemschemas/<SCHEMA_NAME>" #unity #governance #databricks

5

📢 Hugging face is directly integrated with GenAI developers favorite tool "langchain"🦜 Get set go!🚀 💡 pip install langchain-huggingface https://lnkd.in/ddHreVrs #huggingface #GenerativeAI #langchain

22

Sometimes I just can't share all my thoughts on LinkedIn. That's why I've launched my own newsletter : https://lnkd.in/eVukhBqB I've just published my first post, where I dive into what TL;DR is and how to craft them effectively : https://lnkd.in/eKTNkeiA Give it a read, and if you enjoy it, be sure to subscribe for more exciting content in the future! #newsletter #datascience

18

SingleStore Announces Bi-Directional Integration with Apache Iceberg Learn more & get our take 👇 https://lnkd.in/gNTBCWAE "Our vision has always been to provide one single data store for all companies to be able to take advantage of speed, scale and simplicity. Our data platform is designed to unlock all types of enterprise data — including data that is frozen in data lakes." — Raj Verma, CEO at SingleStore "SingleStore’s data platform is indispensable for analyzing our data and delivering real-time insights to our customers. This latest update continues to push the boundaries of what we can accomplish by harnessing all of our enterprise data in real time." — Mauricio Aristizabal, Principal Data Architect and Founder at Giist Social "The integration with Iceberg, plus enhancing our platform’s scaling and processing capabilities will allow users to access and harness their data in real time, and build intelligent apps." — Nadeem Asghar, Head of Engineering at SingleStore #apache #product #SoTNews

1

Counter Analytics in PostgreSQL: Rethinking Denormalization Denormalized counters are often championed for OLTP workloads, trading write overhead for faster reads. But is it always the best approach? 🤔 For high-frequency updates, like tracking post likes in time-series data, denormalization can lead to: - VACUUM overhead: Frequent updates bloat tables. - Lock contention: Concurrent updates fight for the same row. Enter TimescaleDB. 🌟 Instead of maintaining counters, leverage: - Hypertables for automatic time-based partitioning. - Continuous aggregates to pre-compute metrics, avoiding constant updates. - Retention policies to manage data growth seamlessly. This approach improves concurrency, analytics, and scalability while reducing maintenance overhead. 💡 Key takeaway: For time-series workloads, rethink traditional OLTP patterns and embrace TimescaleDB's time-first mindset. 👉 Scale smarter: https://lnkd.in/gK5SB5uP

6

Netflix have mentioned about their internal workflow orchrestrator since 2022 “‘Maestro” And they just opensource it Last week! Some key highlight ✨: - It can scale horizontally and suitable for large scale data processing - Support various kind of run script (Docker images, notebooks, bash script, SQL, Python, and more) - Support acyclic, cyclic workflows and come with built-in complex run strategy/pattern - Workflow definition is JSON format (Unlike Airflow that is python base) - Support dynamically parameterized workflows with Simple Expression Language (SEL) For those who use cloud data platform like Databricks, Azure Data Factory and other mighy feel that their is nothings new. All of these features can be done easily (with some extra pay 💸) But Maestro is opensource like Apache Airflow. Repository : https://lnkd.in/g4dT4VN5 Blog from Netflix : https://lnkd.in/gkTPhFfz

14

Fine-tuning LLMs: Aligning models with specific customer needs is crucial for real-world success, yet it requires precision to ensure the model adapts effectively without compromising quality. Fine-tuning often introduces risks, such as underfit or overfit layers, which can lead to unpredictable behavior in production. Tools like WeightWatcher are essential for addressing these challenges, offering a data-free diagnostic to evaluate model quality by analyzing the spectral properties of weight matrices. Grounded in the Heavy-Tailed Self-Regularization (HT-SR) theory, WeightWatcher provides insights into layer performance and potential issues, helping developers optimize instruction-tuned models like Llama3.1 and Falcon2 for robust, production-ready use cases. #InstructionFineTuning #LLMEngineering #ModelQuality #CustomerFocusedAI

2

"If we're handling big data, we need to be as efficient as possible with memory. PySpark's lazy evaluation helps with this because it plans the best execution. PySpark does the optimization for you." – Justine Paul Padayao, Senior Data Engineer at 3Cloud Solutions. In his talk 𝗪𝗼𝗿𝗸𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗕𝗶𝗴 𝗗𝗮𝘁𝗮 𝗨𝘀𝗶𝗻𝗴 𝗣𝘆𝗦𝗽𝗮𝗿𝗸 at PyConf Mini Davao 2024, Justine highlighted how PySpark’s lazy evaluation and built-in optimization help developers manage big data more efficiently. By automatically optimizing execution plans, PySpark ensures memory is used effectively, making it a powerful tool for scalable data processing.
 #PythonConference #PyConfMiniDavao #Python #DSW2024 #PySpark #BigData
 Organized by: DurianPy - Davao Python User Group 
 Sponsored by: Python Philippines (PythonPH) Stace CoDev Mugna Tech JetBrains 
 Contributing Partners: DEVCON Davao IDEAS Davao Ateneo de Davao Ventures Innovation Center
 Community Partners: PizzaPy - Cebu Python Users Group DICE - Davao Interschool Computer Enthusiasts AWS User Group Davao Society of Programmers And Refined Computer Scientists (SPARCS) Ateneo Circle of Computer Enthusiasts for Study and Success SAMAHAN Systems Development Computer Studies Student Executive Council - ADDU CS AWS Learning Club - UP Mindanao

21