Albert DeFusco

"Prompt Engineering" by Lee Boonstra. This paper covers everything from basic prompting techniques to advanced strategies like Chain of Thought (CoT), Tree of Thoughts (ToT), and ReAct, along with practical tips for code-related prompting. Key sections include: - Configuring output with temperature, top-K, and top-P settings - Using zero-shot, one-shot, and few-shot prompting effectively - System, role, and contextual prompting techniques - Best practices, including experimenting with input formats and adapting prompts to model updates https://lnkd.in/gH5xrivh

13

Now available on edX, CS50's new Introduction to Programming with R, a popular language for statistical computing and graphics in data science and other domains, with CS50's own Carter Zenke. Overview at https://lnkd.in/eY-Rw4hk. Enroll for free at https://cs50.edx.org/r. #education #community #r #statistics 

1,576

34 Comments

Did you miss this awesome demo Hugo Lu did with Youssef Mrini over at Databricks 🏘 In Orchestra you can trigger and monitor Databricks Workbooks, Workflows, data quality tests, and lots more. Our users also leverage Orchestra's overarching monitoring and alerting capabilities to link upstream jobs like data ingestion (e.g. ADF) through Databricks into downstream jobs like Reverse ELT with Hightouch or Power BI dataset refreshes. Check it out here 🔗 https://lnkd.in/dU2grpum #dataengineering #databricks

7

1 Comment

Why wait for slow queries when you can speed things up? On the Apache Arrow blog, Ian Cook, David Li, and Matt Topol, break down how Arrow's format accelerates query result transfers by eliminating inefficiencies. Key takeaways: ✅ Columnar format reduces costly transpositions ✅ Zero-copy architecture minimizes overhead ✅ Universal language support ensures seamless integration across platforms Read the full post to learn how Arrow boosts query performance and discover real-world speedups from tools like Theseus, DuckDB, and Snowflake. 🔗 Read more: https://lnkd.in/eBbjz5uZ #DataAnalytics #ApacheArrow #BigData #OpenSource

18

An underrated component of RAG evals is retrieval evals - is potentially more important than LLM evals! Some requirements for performing retrieval evals is 1) identifying the right retrieval metrics, and 2) having a unified dataset representation. This article by Ross Ashman takes you through tour of retrieval/ranking metrics (e.g. precision@K, NDCG, and more), and also shows you how to convert a dataset into the BEIR format in order to run these evals. Once retrieval evals is setup, run it over your LlamaIndex retrievers. https://lnkd.in/gfha8kiX

172

3 Comments

Regardless of your specific role in data science or machine learning, data parsing and processing are likely part of your regular workflow — and Ilija Lazarevic explains how you can leverage open-source library Pydantic to execute them more smoothly and effectively.

23

Commonly used terms in #data preprocessing prior to machine learning are rescaling, normalizing, and standardizing.  These terms are used more or less interchangeably depending on the customs within various fields. 'Normalizing' a vector most often means dividing by a norm of the vector, for example, to make the Euclidean length of the vector equal to one (unit norm). 'Normalizing' also refers to rescaling by the range of the vector (minimum & maximum), to make all the elements lie between 0 and 1. ⭐ https://lnkd.in/dsXS_sey 'Rescaling' a vector means adding or subtracting a constant, then multiplying or dividing by a constant, as one would do to change the units of measurement of the #data. 'Rescaling' also refers to scaling with median and quantiles, which is called robust scaling (rescaled #data becomes robust to outliers). ⭐ https://lnkd.in/deSb2yb8 'Standardizing' a vector most often means subtracting a measure of location and dividing by a measure of scale. For example, if the vector contains random values (with Gaussian distribution), one subtracts the mean and divides by the standard deviation, thereby obtaining a "standard normal" random variable with zero mean and unity standard deviation. ⭐ https://lnkd.in/dEpyrsGx And these essential-to-know feature scaling methods have been nicely elaborated by Soledad Galli in one of her blogs at Train in Data. Thanks to her! Other methods apart from above for #data preprocessing with #scikit-learn are available here. ➡ https://lnkd.in/dtnbhpbt

29

🚨 Is your data engineering workflow holding you back? 🚨 MotherDuck and Dagster Labs are teaming up to discuss the secrets of modern data engineering. We'll cover: - Crafting an end-to-end data platform with Bluesky data, MotherDuck, and Dagster - Proactively solving problems with "shift-left" testing to catch issues early - Practical automation strategies to streamline your workflow and boost productivity 📅 January 14th ⏰ 9:00 AM PT ☑️ Registration link in the comments

13

1 Comment

EK's Lulit Tesfaye and Jeff Jonas of Senzing teamed up on EK's latest blog, in which they discuss typical problems that would necessitate Entity Resolution, how to unify enterprise data with a Semantic Layer, and the role of entity-resolved knowledge graphs. #semanticlayer #knowledgegraph #entityresolution https://lnkd.in/ePujrbve

9

Use MLflow with Azure Machine Learning jobs submitted with CLI (v2) - https://lnkd.in/gUkxrC9p Use MLflow to track model metrics and artifacts when training and registering models with the Azure Machine Learning CLI (v2). Learning objectives In this module, you'll learn how to: Automatically track model metrics with MLflow when using common machine learning libraries. Track custom metrics with MLflow. Use MLflow model assets to register a model in the Azure Machine Learning workspace. Prerequisites Some familiarity with the Azure Machine Learning service. Some experience in training machine learning models using frameworks such as scikit-learn. https://lnkd.in/gUkxrC9p

3

🤝 Data Exchange Strategies between Airflow DAGs 🖋️ Author: Shivananda D 🔗 Read the article: https://lnkd.in/e_HaSKz7 ------------------------------------------- ✅ Follow Data Engineer Things for more insights and updates. 💬 Hit the 'Like' button if you enjoyed the article. ------------------------------------------- #dataengineering #airflow #orchestration #data

25

- SwiftKV is a novel model transformation and distillation procedure (by folks at Snowflake) designed to optimize large language model (LLM) inference. Very cool findings in the paper (as presented in the vLLM meetup on 11/13). - SingleInputKV: Prefills later layers' KV cache using earlier layer outputs, reducing computation on prompt tokens. The key point is that the first few layers of attention in LLMs are similar (nice find), allowing SwiftKV's SingleInputKV mechanism to skip computations in later layers during inference. By using a single earlier layer's output to compute the KV cache for subsequent layers, it significantly reduces both computation (by 50%) and memory usage (by 62.5%) without sacrificing model performance. - AcrossKV: Merges KV caches of neighboring layers to minimize memory usage and increase batch size for higher throughput. - For models like Llama-3.1-8B and 70B, SwiftKV reduces prefill compute by 50%, memory usage by 62.5%, and achieves up to 2× higher throughput and 60% lower time per token. https://lnkd.in/guhEBW5G

62

Models like OpenAI's CLIP combine text and images for search, but they have limitations, particularly with longer texts and complex relationships. This article talks about JINA's new model that improves on CLIP by integrating both text and image embeddings in one model, making it faster and more efficient for multimodal search tasks. Experiments and Findings: → Search with Both Text and Images: Combining both embeddings leads to better and more diverse search results, improving relevance and variety in results. → Image-Driven Diversification: By using images for diversification, Jina-CLIP v1 can present a broader range of results, offering users more varied options. Results Comparison: → Using both text and image embeddings (via averaging) consistently outperformed searches based solely on text or images, providing more precise and diverse retrieval across both broad and fine-grained categories. https://lnkd.in/guNRQpBU

15

As a Machine Learning Engineer, you should understand the math behind Deep Learning. And this course covers the Linear Algebra and Calculus you need to know. You'll learn about determinants, differentiation, & integration – plus practical applications.

330

2 Comments

Attention Snowflake users ! Wouldn't it be great if every time anything anomalous happened in your infrastructure it got flagged? Well in Orchestra that's pretty easy to set-up. Check it out below! https://lnkd.in/eP8bj6MU #dataengineering #anomalydetection #snowflake

7

ModernBERT is recently released by (like BERT, it is also an encoder-only model) by Answer.AI and LightOn, building upon BERT. ModernBERT incorporates several advancements from recent LLM research: - Modernized Architecture: It adopts elements from Transformer++ (as used in Llama 2), including rotary positional embeddings (RoPE) for improved handling of long sequences, GeGLU activation functions in the MLP layers, removal of unnecessary bias terms, and an additional normalization layer after embeddings. - Efficiency Optimizations: ModernBERT leverages Flash Attention 2 for speed improvements. It also employs "Alternating Attention" (global attention every 3 layers, local attention with a sliding window in between) to reduce computational complexity for long sequences. "Unpadding" and "Sequence Packing" are used to minimize wasted computation on padding tokens, maximizing hardware utilization. - Enhanced Training Data: ModernBERT was trained on 2 trillion tokens from a diverse set of English sources, including web documents, code, and scientific articles, with minimal repetition of training examples. This diverse training data particularly enhances performance on code-related tasks. - Training Process: The training uses a three-phase approach: initial training at a sequence length of 1024, a long-context adaptation phase at 8192, and final annealing. The stable checkpoints from the first two phases are publicly released to facilitate further research and fine-tuning. Training optimizations include batch size warmup and weight initialization via tiling (for the larger model), leveraging pre-trained weights from the base model. ModernBERT achieves state-of-the-art performance across various benchmarks (GLUE, ColBERT for retrieval, and StackOverflow-QA for code). https://lnkd.in/gJzgxxvW

17