_1722956449156001fz1R.png "Anyscale Logo")
Enjoy a light breakfast and coffee while mingling with Ray attendees
8:00 AM PDT - Monday, Sep 30
- Lunch/BreakSeptember 30: TrainingOctober 1: ConferenceOctober 2: Conference8:00 a.m. Monday, Sep 30Monday, Sep 30Session Type: Lunch/Break
- Monday, Sep 308:00 AM - 9:00 AM PDTTraining 1
9:00 AM PDT - Monday, Sep 30
- TrainingHands-on TrainingRay Use CasesBeginnerSeptember 30: Training9:00 a.m. Monday, Sep 30Monday, Sep 30Ray AI Libraries provide a high-level, easy-to-use set of APIs for the full AI/ML lifecycle, designed with MLEs and SWEs in mind. Consume all sorts of data, train models, tune them, serve them, and use them for batch inference or semantic featurization (like embeddings) -- all with minimal code, using patterns you already know. If you'd like to simplify a large-scale workflow, this session is for you.
In this workshop, we'll introduce patterns and APIs for using the Ray AI Libraries, including Ray Data, Ray Train, and Ray Serve. We'll use running code examples that demonstrate end-to-end workflows with Ray and show examples featuring both tabular data modeling (e.g., XGBoost) and deep learning models (PyTorch).
By the end of the class, you'll know when and how to replace local, small-scale, or heterogeneous data pipelines with the unified Ray AI Libraries. You'll have working notebooks to use later. And you'll gain experience using Ray to deliver the fastest performance and the largest scales required by today's projects.
Prerequisites:
This is beginner friendly training, you will learn about Ray fundamentals.
Familiarity with basic ML concepts and workflows.
Intermediate-level experience with Python.
Ray Libraries:
Ray Data
Ray Train
Ray Serve, Technical Trainer, Anyscale
Level of Expertise: BeginnerSession Tracks: Ray Use CasesSession Type: Hands-on Training- Monday, Sep 309:00 AM - 12:00 PM PDTTraining 1
- TrainingHands-on TrainingBeginnerSeptember 30: Training9:00 a.m. Monday, Sep 30Monday, Sep 30Retrieval-Augmented Generation (RAG) is transforming how AI systems interact with data, offering a powerful approach to enhance information retrieval and generation. This workshop is designed for AI practitioners who want to build and deploy RAG-powered applications from the ground up, leveraging Ray's capabilities for scalability and efficiency.
You'll be guided through the entire lifecycle of a RAG system, starting with constructing and encoding a knowledge base using Ray Data. From there, you'll learn to deploy retrieval and generation endpoints with Ray Serve, ensuring your application can handle real-world queries effectively.
By the end of this training, you will have a comprehensive understanding of the RAG application lifecycle and the skills to design, and deploy RAG systems. Join us to enhance your AI toolkit and master the art of creating scalable, high-quality RAG applications with Ray.
Prerequisites:
This is beginner friendly training, you will learn about RAG from first principles.
Basic understanding of LLMs-enabled use cases.
Intermediate-level experience with Python.
Ray Libraries:
Ray Data
Ray Serve, Engineering Manager, Anyscale
, Engineering Manager, Anyscale
Level of Expertise: BeginnerSession Type: Hands-on Training- Monday, Sep 309:00 AM - 12:00 PM PDTTraining 3
- TrainingHands-on TrainingIntermediateIntermediateSeptember 30: Training9:00 a.m. Monday, Sep 30Monday, Sep 30Ray Core allows developers to leverage a small number of core primitives for building and scaling distributed AI applications.
This training provides you with an extensive and practical understanding of Ray Core concepts. You will review Ray's architecture, API components (tasks, actors, object store), core services, and observability tooling. You will learn about performance and scalability optimizations, as well as best practices validated by industry experts.
By the end of this training, you will have a comprehensive understanding of Ray Core, and you'll be able to apply this knowledge to tackle challenging scaling problems and deliver robust solutions that meet your organization's needs.
Prerequisites:
Some experience with Ray Core or distributed systems.
Intermediate-level experience with Python.
Ray Libraries:
Ray Core, Anyscale
Level of Expertise: IntermediateSession Type: Hands-on Training- Monday, Sep 309:00 AM - 12:00 PM PDTTraining 2
12:00 PM PDT - Monday, Sep 30
- Lunch/BreakSeptember 30: TrainingOctober 1: ConferenceOctober 2: Conference12:00 p.m. Monday, Sep 30Monday, Sep 30Grab lunch and meet other attendees eager to learn at Ray TrainingSession Type: Lunch/Break
- Monday, Sep 3012:00 PM - 1:00 PM PDTTraining 1
1:00 PM PDT - Monday, Sep 30
- TrainingHands-on TrainingIntermediateSeptember 30: Training1:00 p.m. Monday, Sep 30Monday, Sep 30In the era of AI, fine-tuning large language models (LLMs) for specific tasks has become essential for delivering high-quality, tailored solutions. This workshop will guide you through the process of building end-to-end workflows for LLMs at scale, focusing on MLOps best practices and leveraging Anyscale's LLMForge for fine-tuning.
You'll start by fine-tuning a large language model for a specific functional representation task, using Ray Data for batch evaluation of the fine-tuned model. You'll also learn to deploy the model using Ray Serve, ensuring efficient and scalable deployment. The workshop will conclude with a deep dive into adopting MLOps best practices, including automated retraining and evaluation to maintain and improve model performance.
By the end of this session, you'll have a comprehensive understanding of how to manage the full lifecycle of LLM workflows, from fine-tuning and evaluation to deployment and continuous improvement. You'll gain practical skills to implement scalable, efficient, and automated workflows for large language models, ensuring they remain accurate and effective over time.
Prerequisites:
Familiarity with MLOps and LLM use cases.
Intermediate-level experience with Python.
Ray Libraries:
Ray Data
Ray Train
Ray Serve, Software Engineer, Anyscale
, Software Engineer, Anyscale
Level of Expertise: IntermediateSession Type: Hands-on Training- Monday, Sep 301:00 PM - 4:00 PM PDTTraining 2
- TrainingHands-on TrainingvLLMIntermediateSeptember 30: Training1:00 p.m. Monday, Sep 30Monday, Sep 30Traditional search systems often struggle with handling unstructured data, especially non-text data like images. In this workshop, you'll learn how to enhance legacy search systems using generative and embedding models for richer data representation.
You will build a scalable multi-modal data indexing pipeline and a hybrid search backend using Anyscale and MongoDB. The training will cover practical applications of Ray Data for scalable batch inference, Ray Serve for deploying and scaling the search application, vLLM for integrating large language models and MongoDB Atlas for both lexical and vector search.
By the end of this session, you'll have the skills to implement advanced AI tooling for creating a scalable and efficient search system capable of handling diverse data types in enterprise applications.
Prerequisites:
Familiarity with large scale data processing.
Prior experience with Ray Data or Ray Serve is not required, but participants with some experience in these frameworks will have an advantage in understanding the more advanced topics that will be covered in the training.
Intermediate-level experience with Python.
Ray Libraries:
Ray Data
Ray Serve, Anyscale
Level of Expertise: IntermediateSession Tracks: vLLMSession Type: Hands-on Training- Monday, Sep 301:00 PM - 4:00 PM PDTTraining 3
- TrainingHands-on TrainingAdvancedSeptember 30: Training1:00 p.m. Monday, Sep 30Monday, Sep 30Text-to-image generative AI models, like Stable Diffusion, are transforming the creative industry by generating high-quality images from text descriptions. In this workshop, you'll learn how to build scalable systems for these models by creating an end-to-end pipeline for pre-training a Stable Diffusion model on a billion-scale dataset using Ray Data and Ray Train.
You'll explore how to efficiently stream data preprocessing and conduct distributed training across multiple GPUs. Additionally, you'll gain practical experience in deploying and scaling a Stable Diffusion model using Ray Serve, allowing you to deliver real-time, high-quality generative outputs.
By the end of this session, you'll have a solid understanding of how to implement a complete generative AI pipeline with Ray. You'll leave with the skills to pre-train, deploy, and scale Stable Diffusion models, ready to handle large-scale generative tasks and produce impressive results.
Prerequisites:
Basic familiarity with computer vision tasks, including common challenges with data processing, training, and inference.
Intermediate programming skills with Python.
Basic understanding of text-to-image use cases.
Ray Libraries:
Ray Data
Ray Train Ray Serve, Technical Trainer, Anyscale
Level of Expertise: AdvancedSession Type: Hands-on Training- Monday, Sep 301:00 PM - 4:00 PM PDTTraining 1
