Matthieu BARRUE--BRAGADE

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

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

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

Everyone is diving into learning about LLMs, and this is great because it's leading to the creation of exceptional resources. . . . #mlops #ai #LanguageModel #MachineLearning #DeepLearning #ArtificialIntelligence #TechInnovation #LLM

8

My Solution In SQL QUERY :- CREATE TABLE ##WITHDAYNAME( SENTENCE_ID INT, SENTENCE VARCHAR (500)) INSERT INTO ##WITHDAYNAME VALUES  (1,'MY SCHOOL RE-OPENS ON 21-05-2024'),  (2,'WE WENT FOR A TRIP ON ON 18-08-2023'),  (3,'THERE WAS NO EVENT ON 12-12-2022'),  (4,'NO DATES ARE MENTIONED HERE')  SELECT SENTENCE_ID ID,  CASE WHEN ISDATE(SUBSTRING(DTECOL,7,4)+'-'+SUBSTRING(DTECOL,4,2)+'-'+SUBSTRING(DTECOL,1,2))=1 THEN   SENTENCE+' ('+DTENAME +').' ELSE SENTENCE+'.' END UPDATED_SENTENCE FROM (  SELECT *, SUBSTRING(DTECOL,7,4)+'-'+SUBSTRING(DTECOL,4,2)+'-'+SUBSTRING(DTECOL,1,2)DTECOL1,  CASE WHEN ISDATE(SUBSTRING(DTECOL,7,4)+'-'+SUBSTRING(DTECOL,4,2)+'-'+SUBSTRING(DTECOL,1,2))=1 THEN   DATENAME(DW,SUBSTRING(DTECOL,7,4)+'-'+SUBSTRING(DTECOL,4,2)+'-'+SUBSTRING(DTECOL,1,2))   ELSE '' END DTENAME  FROM (  SELECT *, RIGHT(SENTENCE,10)DTECOL FROM ##WITHDAYNAME  )AA  )BB

10

Question : where do I store my vector embeddings ! Answer : np.array , for the scale at which you are. Hesitate to go fancy so fast 😂 #vectordb #rag #embeddings #nlp #datascience

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

These days I’ve been in Montpellier, France, to present our last research results on combinatorial optimization problems over temporal graphs to the 31st symposium on temporal representation and reasoning (TIME2024). Check out our paper at https://lnkd.in/da66ref4 Here we discuss a new heuristic-based approach for minimum timeline cover over temporal graphs. The paper provides theoretical results to a problem largely adopted in social network analysis as interpretable network summarization framework to find important events. Particularly, we used it to analyze job postings, to find the most required skills by companies and understand how the market is evolving. We are already working on evolutions of this approach with Deep Learning. Stay tuned for next publications. #knowledgegraph #combinatorialoptimization #networksummarization

15

1 commentaire

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

19

Interesting and the best part is that in the comment section, people share other useful packages other than pipreqs, such as pip-tool, Poetry, uv and others.

5

1 commentaire

Create table ##stock (stock_id varchar(10), date date, price int) Insert into ##stock values  ('A', '2024-01-01', 100),  ('A', '2024-01-02', 105),  ('A', '2024-01-03', 104),  ('B', '2024-01-01', 200),  ('B', '2024-01-02', 200),  ('B', '2024-01-03', 201)  with cte as  (  select * , LAG(price)over(partition by stock_id order by date)lprice from ##stock  )  select stock_id,date,price ,   case when lprice IS null then 'NULL'  when price>lprice then 'UP'  when price<lprice then 'Down'  else 'Same' end price_change  from cte

5

1 commentaire

📚 Another conference, another article! The International Conference on Machine Learning (ICML) took place last week, from July 21 to 27, 2024, in Vienna. In this article I explore some exciting advances in Machine Learning presented at this year’s edition: 🔍 For knowledge transfer, I highlight LLM distillation, 2-bit quantization, and transfer learning, which aim to create smaller and efficient models. 📈 In time series analysis, I explore zero-shot forecasting and metadata-enhanced time series generation, which is especially useful in energy field. Discover more in my medium article!

21

Question : Why is #AI modeling just the tip of the iceberg❓ Many aspiring product managers and entrepreneurs are drawn to AI modeling, believing it to be the primary driver of success for AI-powered solutions. However, seasoned AI experts know that the actual modeling work represents only a small fraction of the overall effort required to build a successful AI product. In reality, AI modeling accounts for roughly 20% of the job 😯 . The remaining 80% involves a complex web of supporting systems and processes that are essential for transforming AI capabilities into tangible business value. From data collection and preprocessing to model infrastructure and deployment, each of these components plays a critical role in the AI product lifecycle. 👉 Data is the lifeblood of any AI system, and the effort required to source, clean, and prepare data for modeling can often exceed the time spent on the modeling itself. 👉 Once the data is ready, the next challenge lies in building the infrastructure to support the AI models. This includes everything from cloud-based compute resources and storage solutions to model versioning, monitoring, and scaling capabilities. Effective model deployment and integration with the broader product ecosystem are also essential for delivering a seamless user experience. 👉 Beyond the technical aspects, AI product managers must also consider the operational and organizational factors that contribute to the success of an AI-powered solution. This includes establishing processes for model governance, managing model drift, and ensuring ongoing model maintenance and retraining. 💡 Quick Tip: Adopt a systems-level perspective when developing your AI-powered products, ensuring that you allocate resources and attention across the full spectrum of requirements, not just the modeling component. #artificialintelligence #productmanagement #datascience #productmanager

4

𝐁𝐫𝐢𝐥𝐥𝐢𝐚𝐧𝐭 𝐬𝐨𝐥𝐮𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐢𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐯𝐞 𝐚𝐧𝐚𝐥𝐲𝐬𝐢𝐬 𝐚𝐧𝐝 𝐌𝐋 𝐚𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬 The architecture is simply brilliant - e2b runtime - LLM middle layer- - notebook style rendering Massive applications. - fully scalable. - open source - customize - build - integrate Next up - prompt-driven iterative analytics and machine learning. Now exploring ways to incorporate similar solution into REX (rex.tigzig.com / realtime.tigzig.com - my open source analytics toolkit), to provide interactive analysis and machine learning. Check out the video demo on X (I actually saw it on X first): https://lnkd.in/gQXqMRiZ App & Code base : https://lnkd.in/eGmyXh-q

2

A nice and simple introduction to pandas. I wrote about it on my Medium. Feel free to check it out here : https://lnkd.in/ejKwWTJT #pandas #MachineLearning #DataScience #DataAnalysis #Python #AI #fundamentals

7

Mastering Vector Embeddings in Snowflake! Are you diving into the world of vector embeddings and exploring how to leverage Snowflake for Natural Language Processing and RAG based GenAI applications ? My latest blog for phData walks you through an essential topic: 👉 How to Split Text for Vector Embeddings in Snowflake 💡 Ready to level up your text processing game? Check it out here: https://lnkd.in/gzzev-mv When working with large text datasets, effective splitting techniques can make or break your embeddings' accuracy. In this blog , you'll discover: ✅ What are the major types of Text Splitting Strategies . ✅ How to use Snowflake to implement vector embeddings. ✅ Practical tips and SQL snippets to get started right away. Whether you're a data engineer, data scientist, or machine learning engineer, this guide equips you with actionable insights to harness the full power of vector embeddings. Let’s unlock the potential of unstructured data together! 💻✨ #DataScience #MachineLearning #Snowflake #VectorEmbeddings #NaturalLanguageProcessing #DataEngineering #VectorDatabases #GenAI #RAG

17