The repo is for the ONNX Dojo training course. It is organized with presentations, participants, and materials for the lab exercises in sub-folders.
Agenda:
Day 1 (Morning) 8am - 11am, Pacific
- Opening - Chin Huang
- Open Source at IBM - 15 min. https://ibm.ent.box.com/file/656007169789
- Git hands-on - 60 min. Chin Huang
- Introduction to ONNX - 45 min. Chin Huang
- ONNX training - 30 min. Svetlana Levitan
Day 1 (Afternoon) 11:30am - 2:30pm, Pacific
- ONNX Community and Contributor Resources - 15 min. Chin Huang
- Overview of ONNX GitHub repositories - 30 min. Chin Huang
- ONNX Development Environment with hands-on - 30 min. Chin Huang
- High level code walk-through of major components ONNX-Tensorflow converter - 45 min. Winnie Tsang
- Open Lab: Run ONNX ResNet model in Tensorflow - 45 min. Chin Huang
Day 2 (Morning) 8am - 11am, Pacific
- PR creation walk-through in ONNX-Tensorflow - 30 min. Winnie Tsang
- Open Lab: Develop support for a new/updated operator - 30 min. Winnie Tsang
- ONNX-MLIR overview - 60 min. Tian Jin
- ONNX-MLIR tutorials/code walkthrough - 45 min. Tian Jin
- Wrap up - Chin Huang
Lab exercises:
System requirements: Mac, Linux, or Windows 10 with Subsystem for Linux (Ubuntu 18.0.4)
Prereqs: python3, git, cmake, protobuf-compiler libprotoc-dev
- Exercise 1: git - Get familiar with git from fork to PR
- Exercise 2: resnet - Run ResNet model in Tensorflow
- Exercise 3: constant_pr - Develop support for a new/updated operator