Ashwin Ramesh

https://www.linkedin.com/in/ashwinramesh26/ [email protected]

University of Rochester Rochester, New York ML Intern @ Adobe | MS Computer Science at University of Rochester | Research Assistant at VISTA Lab | Computer Vision, Audio, Multimodal ML

Repositories
Issues
Comments

Results 11 comments of


                                            Ashwin Ramesh

How to obtain differentiable softmax_lse

@tridao 1. Does "Backprop on softmax_lse is not supported" mean that that backprop wouldn't work correctly even if I use softmax_lse only and immediately to merge attention with another kv-set's...