Openning07

I've tried it and it still cannot works. Specifically, when I used 5e-4 as the initial learning-rate, LeViT-128S came up with *nan loss* at Epoch 6 (I ran 3 times...

https://arxiv.org/abs/2005.09973

Can you also release the training log files of the released shunted ViT models? The log files will help us check the training details to ensure the reproduce procedure. Thank...

Primarily, you should change the data directory of AVA and make sure the environment is prepared to reproduce the claimed results in my paper. Please read the **Note** and one...

The core difference is the *weights_JensenFactor* in function [compute_loss_and_error](https://github.com/Openning07/MPADA/blob/fd438b26764c1fad2ed4d6d3cc78fdb2c181a970/imagenet_utils.py#L460). If you want to change MPada to other two modes, please change the implementation of *weights_JensenFactor*: for MPavg, *weights_JensenFactor* is a...

Yes. I used *0.5* for *JensenFactor* when MPada mode is selected.

Thanks for your comments. 1. Since the project is built on tensorpack lib, and we need to create a new data loader to load the AVA benchmark. The file of...