Ruochen Wang
Ruochen Wang
Hello. That is correct; for S1-S4 subspaces we report the best result following the original RobustDARTS, to encourage fair comparison with its baselines. (For the reference of other fellows that...
The CIFAR10 evaluations are identical to DARTS, i.e. under --init_channels 36 --layer 20. The evaluations on C100 and SVHN follow the original RobustDARTS paper (App C), which is different from...
No problem!
Hi, we didn't search for RNN. I think it's been a long time since I saw new papers working on the DARTS' RNN space.
The Conda environment used in this project was set up in Feb 2020; at the time APIv1.0/FILEv1.0 was the newest version. For the differences, I think 201's author would be...
Hi zhengjian, sorry for the late reply as I needed some time to rerun the exp you mentioned. I reran the darts-201 experiment using the current repo with 2 seeds...
Thanks thuangb for the comment. Yes, it looks like a mismatch of tensor locations. Some part of the model seems to be on CPU (either loaded parameter, original model parameter,...
Can you provide detailed steps to help us reproduce the issue you encountered?
From your reply, it seems that you have changed the task. In this case, I might only be able to provide limited help since I could not run the adapted...
You are welcome. The input dimension of your task is much larger than cifar10/cifar100 we used. So the memory consumption might exceed your GPU limit. It could help if try...