High Level Synthesis SDSoC-SDAccel
Deep Learning on FPGA with SDSoC
This course is specially designed for designers new to DeepLearning design or application. The introduction to the theoretical principle, effective comprehension of training and testing flow using Caffe Framwork, demonstrate impressive project demo, understand the design flow with C/C++ coding techniques, use little HDL coding techniques and more C coding techniques to apply Deeplearning demo on edge of FPGA device.
Participants will be taught the necessary knowledge to Comprenhend the theoretical principle, training, testing and process about application on FPGA device. Participants will be able to test their knowledge and apply their skills on Deeplearning immediately. You will be able to face the real challenge to do a primary Deeplearning design flow project with the guidance of our instructor.
• Understanding of Zynq®-7000 architecture (with emphasis on ACP, HP ports, and internal routing)
• Comfort with the C programming language/ Python
• Familiarity with the Vivado® Design Suite, Vivado HLS tool, and Xilinx SDK
Vivado Design Suite and SDx™ development environment 2017.1 or any other latest version based on notes availability.
1.1 Introduction to DeepLearning
Overview of DeepLearning origin and development, propose the process and methodology of Deeplearning design.
1.2 Introduction to Artifical and Convolutional Neural Network
Overview of Convolutional Neural Network and introduces the mathematical model of Artifical Neural Network.
1.3 DeepLearning Design Methodology
Introduces the methodology guidelines covered the DeepLearning training stage and the design flow.
1.4 Introduction to the Architecture of Convolutional Neural Network
Introduces the primary network layers in the architecture of Convolution Neural Network.
1.5 Installation about Caffe framework
Introduces the process of installation about Deeplearning framework Caffe, including the setting of enviroment about caffe, python, etc.
1.6 Introduction to network structure files in Caffe
Introduces the details of network structure about the Cifar10 demo in Caffe and the basic knowledge for training and testing.
1.7 Process on training and testing with the Cifar10 demo
Demostrate the process of training and testing with Cifar10 demo, comprehend the implication of data on testing process.
1.8 Experiments about Mnist Dataset demo
Interactions using experiments about Mnist Dataset demo with Python coding techniques.
1.9 Hands-on operation and experiment
Review the foundations about DeepLearning on Caffe.
2.2 Introduction to Forward Propagation
Introduces the design of Forward Propagation with C coding techniques and demonstrate the Inference demo about Cifar10 demo to identify airplane, automobile, bird, horse, etc.
2.3 Introduction to Backward Propagation
Introduces the theoretical principle and derivation process about Backward Propagation.
2.4 Introduction to training and testing demo
Demonstrate the training and testing demo and introduce structures and codes about the demo with C++ coding techniques.
2.5 Hands-on operation and experiment
4.1 SDSoC Tool Overview
Explain the advantages of using the SDSoC development environment and describe the typical SDSoC development flow.
4.2 Tools Collection
Introduces the Tools:sdsoc IDE involes a lot of tools ,CDT,TCF,sdscc/sds++ ...
4.3 SDSoC Platform Creation
Introduces the SDSoC Platform structure and how to create a new platform
4.4 Understanding Estimations in the SDSoC Tool
Introduces the Estimations tool in the SDSOC:estimations the source of FPGA when generated a new accelarator.
4.5 Pragma Syntax
Understand the pragma syntax and Use Pragma to get better performance.
4.6 Introduction to Deeplearning demo complete with SDSOC Design Suite
Using sdsoc accelerate complex algorithm :complex algorithms are implemented in FPGA .
4.7 Hands-on operation and experiment