Description
Master LLM Inference (Phase 1: Foundations & Optimization). This course is a live intensive workshop on deep learning, infrastructures, and optimization of large language model inference systems (LLM Inference). This 4-week intensive workshop is led by Dr. Raj Dandekar (PhD graduate from MIT) and features 9 leading experts from Anthropic, NVIDIA, Apple, and other reputable companies. In this specialized training, learners will not only learn theoretical topics, but will also work in a fully practical way with the same frameworks and tools that are used in large technology companies for real production systems. The main goal of the course is to enable engineers to move beyond the simple use of large language models and achieve the ability to design, build, and optimize complex inference infrastructures. Along the way, participants will implement a practical and valuable final project that connects all the concepts and techniques learned. Dedicated lab days are also included in the program so that programmers can run real-world models on different hardware and evaluate their performance. Since each hardware device has different processing bottlenecks, learners learn how to measure and analyze each hardware bottleneck live. This training provides a balance between scientific theories and applied systems engineering, and offers a unique opportunity to gain career insights directly from senior engineers in the IT industry.
What you will learn
- Designing Inference Systems: Answering Interview Questions Designing inference systems from start to finish to ace ML job interviews.
- Service Optimization: Providing and serving large language models with low latency, high processing power, and optimized cost at large scale.
- Deployment on real hardware: Deploy large real language models on personal laptops, Raspberry Pi 4, Android devices, and Jetson Orin Nano.
- Building a professional portfolio: Implementing industry-level portfolio projects based on lab days and working with a variety of hardware.
- Answer key interview questions: Ability to answer complex questions from technology companies like Anthropic, NVIDIA, Microsoft, Meta, and Google DeepMind about designing low-latency, high-throughput inference systems.
- Receive career insights: Gain first-hand knowledge and insight about career paths from leading engineers at leading companies around the world.
This course is suitable for people who:
- IT Engineers: People who want to transition into ML infrastructure or AI engineering.
- Students and researchers: Those aiming to obtain job positions at companies such as Anthropic, NVIDIA, Microsoft, Apple, and Amazon.
- Developers and Engineers: Programmers who want to go beyond simple implementation of LLMs and build advanced inference systems.
- Artificial Intelligence Researchers: Researchers who, in addition to scientific theory, require in-depth knowledge of real-world systems engineering.
Course details
- Publisher: Vizuara AI Labs
- Lecturer: Dr. Raj Dandekar , Yash Dixit
- Training level: Beginner to advanced
- Number of lessons: 13
- Training duration: 30 hours and 31 minutes
Course headings
- The Vizuara Inference Book Phase 1 (Also your lecture notes!)
- Hardware Labs
- All Visual Walkthroughs in one place
- Lecture 1 Introduction to Inference Engineering
- Lecture 2 Good and Evil of KV Cache
- Bonus Lecture Inference and YCombinator
- Lecture 3 Attention Variants Part 1 (MHA, MQA, GQA, Latent Attention, Sparse Attention)
- Lecture 4 Attention Variants Part 2 (Sliding Window, State Space Models and Mamba)
- Lecture 5 All about Flash Attention 1, 2, 3
- Lecture 6 The anatomy of a vLLM Step
- Lecture 7 All about Quantization
- Lecture 8 Speculative Decoding
- Phase 1 Capstone Project!
Course images
Sample course video
Installation Guide
After Extract, view with your favorite player.
Subtitles: None
Quality: 1080p
Download link
Rapidgator link
File(s) password: www.downloadly.ir
File size
4.9 GB