Nandan Kumar Jha

icon indicating a website link https://www.nankj.com

icon company New York University icon location New York, USA icon location PhD candidate at NYU. Research interests: privacy-preserving deep learning and LLM privacy and security.

Results 2 issues of Nandan Kumar Jha

shapes and out_shapes values in ReviewKD

1 comment

To me, it's confusing, how to set the "shapes" and "out_shapes" when the student is ResNet18 and the teacher is ResNet34 on CIFAR-100. Is it shapes = out_shapes = [1,...

Training only on 2B tokens (openwebtext)

3 comment

Hi ! Interesting work on the role of explicit bias! I was wondering what training settings got you an eval PPL ~3.04. The paper mentions that 50K iterations are required...