ONNX talk and workshop details for WIDS 2020.
- Vinitra Swamy
- Cecilia Liu
- Ashwini Khade
Optimize BERT with ONNX Runtime
- Overview of ONNX / ONNX Runtime (10 minutes)
- Setup: Create Microsoft accounts, clone Azure Notebooks repo, install dependencies (10 minutes)
- Demo
- Convert BERT from PyTorch to ONNX (10 minutes)
- Inference inputs / outputs with ONNX Runtime (5 minutes)
- Optimizations with ONNX Runtime (5 minutes) + Demo: OLive Tool (5 minutes)
- Discuss Azure ML deployment (5 minutes)
- Questions (5 minutes)