The Case for Hardware-ML Model Co-design with Diana Marculescu - #391
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
00:00
Profiling for Power and Latency Models in Neural Networks
This chapter delves into the methodology for collecting and profiling data to create power and latency models for neural networks across multiple GPU platforms. It highlights the extensive assessments conducted on power, latency, and memory usage, while assuring listeners of the adaptability of the methodology for various platforms.
Play episode from 26:31
Transcript
