TinyML¶
Rather than adding more compute power, focus on improving compute efficiency
Will mainly focus on the following applications: Speech, Computer Vision, NLP
Overview¶
- Hardware
- Architecture & Dataflow
- Metrics and Analysis
- Efficiency
- Micro-architecture/Circuits
- Model Optimization
- Quantization
- Pruning
- Knowledge distillation
- AutoML
- Software
- Domain-specific compilers; eg: TVM
- Kernel implementations
- Mapping onto hardware
- Systems
- Distributed training
- Federated learning
- Pre/Post Processing
- Environmental issues
Pre-Requisites¶
- Computer archictecture
- Machine Learning
- Python programming
- PyTorch Basics
Reading¶
References¶
- Machine Learning Hardware and Systems (Cornell Tech, Spring 2022)
- Videos
- Material
- TinyML and Efficient Deep Learning Computing | EfficientML.ai - MIT HAN Lab
- Tiny Machine Learning | UPenn
- AutoDL | Applied Deep Learning
Current Video¶
https://www.youtube.com/watch?v=5_qVob2Vwf8&list=PL0mFAhrXqy9CuopJhAB8GVu_Oy7J0ery6&index=10