flâneur

lec05.pdf

dropbox.com · 736 words · saved by 2 readers

lec05 PDF File Edit View Help 1 of 1 Share Log in Sign up Page 2 of 730 Edit PDF Draw Highlight Add text 72% MIT 6.5940 : TinyML and E ffi cient Deep Learning Computing https://e ffi cientml.ai EfficientML.ai Lecture 05 Quantization Part I Song Han Associate Professor, MIT Distinguished Scientist, NVIDIA @SongHan_MIT MIT 6.5940 : TinyML and E ffi cient Deep Learning Computing https://e ffi cientml.ai Lecture Plan Today we will: 1. Review the numeric data types used in the modern computing systems, including integers and fl oating- point numbers. 2. Learn the basic concept of neural network quantization 3. Learn three types of common neural network quantization: 1. K-Means-based Quantization 2. Linear Quantization 3. Binary and Ternary Quantization 2 1 1 0 0 1 1 1 1 -2 7 2 6 2 5 2 4 2 3 2 2 2 1 2 0 = -49 × × × × × × × × + + + + + + + Continuous Signal Quantized Signal time Signal MIT 6.5940 : TinyML and E ffi cient Deep Learning Computing https://e ffi cientml.ai Low Bit-Width Operation

lec05pdf Page 1 of 730 Song Han shared this file. Want to do more with it? Note: There may be limited functionality for screen reader users for this commenting tool. MIT 6.5940: TinyML and Efficient Deep Learning Computinghttps://efficientml.aiEfficientML.ai Lecture 05QuantizationPart ISong HanAssociate Professor, MITDistinguished Scientist, NVIDIA@SongHan_MIT Note: There may be limited functionality for screen reader users for this commenting tool. MIT 6.5940: TinyML and Efficient Deep Learning Computinghttps://efficientml.aiLecture PlanToday we will:1.Review the numeric data types…

saved by

related reading