Fantastic Models and Where to Find Them

Transcript

1. Fantastic Models and Where to Find Them Rishit Dagli High School TEDx, TED-Ed Speaker rishit_dagli Rishit-dagli
2. Fantastic Models and Where to Find Them Rishit Dagli
3. TensorFlow Hub rishit_dagli A repository of trained machine learning models ready for fine-tuning and deployable anywhere
4. TensorFlow Hub rishit_dagli
5. • High School Student • TEDx and 2xTed-Ed Speaker • I contribute to TF Hub • I ❤ ML Research • My coordinates - www.rishit.tech $whoami rishit_dagli Rishit-dagli
6. • Devs who have worked on Deep Learning Models (Keras) Ideal Audience rishit_dagli
7. rishit_dagli • Off the shelf inference with SoTA machine learning models Could be super helpful rishit_dagli
8. rishit_dagli • • You do need to ask yourself: ◦ Is there enough information in P to represent Q? ◦ how much new data? ◦ can unlabelled data do it? ◦ what’s the right mix of P and Q data? ◦ How far is from P from Q? Transfer Learning rishit_dagli
9. rishit_dagli • • You do need to ask yourself: ◦ Is there enough information in P to represent Q? ◦ how much new data? ◦ can unlabelled data do it? ◦ what’s the right mix of P and Q data? ◦ How far is from P from Q? • Samory Kpotufe: Some Recent Insights on Transfer-Learning Transfer Learning rishit_dagli
10. rishit_dagli • • You do need to ask yourself: ◦ Is there enough information in P to represent Q? ◦ how much new data? ◦ can unlabelled data do it? ◦ what’s the right mix of P and Q data? ◦ How far is from P from Q? • You can do this easily with TF Hub too! Transfer Learning rishit_dagli
11. rishit_dagli • Less data Could be super helpful rishit_dagli
12. rishit_dagli • Less data • Less domain expertise Could be super helpful rishit_dagli
13. rishit_dagli • Less data • Less domain expertise • for different domains including image, video, audio etc. Could be super helpful rishit_dagli
14. rishit_dagli • Less data • Less domain expertise • for different domains including image, video, audio etc. • Take advantage of SoTA machine learning models Could be super helpful rishit_dagli
15. rishit_dagli Ready to Use rishit_dagli Credits: @TheSandeepGupta
17. Style Transfer
18. Style Transfer rishit_dagli
19. Style Transfer rishit_dagli
20. Style Transfer rishit_dagli
21. Classification Prediction: Goldfish, Carassius_auratus
22. Image Classification rishit_dagli
23. rishit_dagli bit.ly/hub-classification
24. Fine tune BERT
25. Fine tune BERT rishit_dagli Classify Text with BERT
26. Fine tune BERT rishit_dagli Classify Text with BERT
27. Fine tune BERT rishit_dagli Classify Text with BERT
28. Fine tune BERT rishit_dagli Classify Text with BERT
29. Fine tune BERT rishit_dagli Classify Text with BERT
30. Fine tune BERT rishit_dagli Classify Text with BERT
31. Fine tune BERT rishit_dagli Classify Text with BERT
32. rishit_dagli And more... TF Hub for other tasks rishit_dagli
33. rishit_dagli tensorflow.org/hub/publish Publish your own models
34. rishit_dagli bit.ly/tfhub-deck Slides here
35. Thank You rishit_dagli Rishit-dagli