Machine Learning Systems (Spring 2022)

Abstraction: UC Berkeley graduate seminar on ML systems and AI-hardware co-design

Key points:

• CS294 graduate course at UC Berkeley taught by Joseph E. Gonzalez and Amir Gholami covering systems to support AI (GPUs, TPUs, TensorFlow, PyTorch, Horovod, Ray) and AI to optimize systems (scheduling, circuit layout, program synthesis)
• Format is a program-committee-style paper review meeting each week; students submit detailed paper reviews and lead discussions alongside invited researchers from each area
• Grading: 60% projects, 20% weekly summaries, 20% class participation; projects encouraged to mix AI and systems students
• Third offering of the course; previous versions in Spring 2019 and Fall 2019
• Reflects the thesis that recent AI success is largely attributable to hardware/software systems advances, not just algorithmic breakthroughs

Connections: Uc Berkeley · Machine Learning Systems · Hardware Accelerators · Mlops

Source: https://ucbrise.github.io/cs294-ai-sys-sp22/