Abstract: The half-day training will train attendees on how to use Hugging Face's Hub as well as the Transformers and Datasets library to efficiently prototype and productize machine learning models.
The training will cover the following topics:
1. Open-Source Philosophy
- Design principles of Transformers and Datasets
- Community Support
- How to contribute ?
2. From Research to Prototyping:
- Find models and datasets for your target task
- Analyse, experiment with models in Transformers
- Training: Pre-processing, Modeling, Post-processing in Transformers
- Set-up training in Transformers
- Compare different models
3. From Prototype to Production:
- Optimize your model
- Experiment with different setups
- XLA, ONNX, Infinity
Python, Numpy or PyTorch or Tensorflow, Transfer Learning in Machine Learning
Bio: Patrick von Platen is a research engineer at Hugging Face and one of the core maintainers of the popular Transformers library.
He specializes in speech recognition, encoder-decoder models and long-range sequence modeling.
Before joining Hugging Face, Patrick conducted research in speech recognition at Uber AI, Cambridge University, and RWTH Aachen University.