Tensilica DNA 100 Architecture and Programming Training
版本 | 区域 | |
---|---|---|
9.0 | Online | ENROLL |
Other Versions | Online | EXPRESS INTERESTINQUIRE |
Length: 1 day (8 Hours)
Course Description
This course teaches how to run applications on the Tensilica® DNA 100 Neural Network Accelerator. You will learn about the DNA 100 architecture, programming model, and using the Xtensa® Neural Network Compiler (XNNC).
Learning Objectives
After completing this course, you will be able to:
- Use XNNC to map Neural Networks to DNA 100
- Assess accuracy, performance and bandwidth at network level and layer level granularity
- Understand how to use XNNC optimization features like Pruning, Quantization & Batching to improve network accuracy, performance and memory bandwidth
- Review ResNet50 (Classification) network on DNA 100.
Software Used in This Course
- Tensilica Xtensa Xplorer 8.0
- Tensilica Xtensa Software Tools RI-2019.1
- Xtensa Neural Network Compiler (XNNC v1.5)
Software Release(s)
RI-2019.1
Modules in this Course
- DNA 100 Deliverables
- DNA 100 Architecture Overview
- DNA 100 XNNC Programming
- DNA 100 Modeling, Simulation & Emulation Platforms
- DNA 100 Profiling & Debugging
- DNA 100 Optimization Techniques
Audience
- DNA 100 users and programmers
- Machine Learning Architects, Engineers & Scientists .
Prerequisites
You must have experience with or knowledge of the following:
- Machine Learning, Computer Vision & Imaging Concepts
- Neural Network Frameworks (specifically Caffe & Tensorflow)
- Some familiarity with DSP Processors and Embedded Programming
You must have completed the following courses:
You must have reviewed the following tutorial video:
Tensilica Xtensa Neural Network Compiler Basic flow
Related Courses
Please see course learning maps at this link for a visual representation of courses and course relationships. Regional course catalogs may be viewed here.
ONLINE TRAINING
Genus Synthesis Solution v16.1
This online class features the Cadence® Genus™ Synthesis Solution with next generation synthesis capabilities and how SoC design productivity gap is filled by Genus