-
From the code for multi gpu training, you expand the number of samples by the number of gpus (https://github.com/microsoft/Semi-supervised-learning/blob/main/semilearn/core/utils/build.py#L158), but y…
-
Hey
Thank You for responding to previous issues very promptly.
I am using the recommended settings to transfer fairness between UTK and Fairface. However, averaged across multiple runs, we observ…
-
Hi,
During ICLR 2023, I discussed with Yue Fan the possibility of adding a new method named DeFixmatch in the USB repo.
Check out the paper here: https://arxiv.org/pdf/2203.07512.pdf
The idea o…
-
Your model is based on the fixmatch method. I see that the flexmatch method is also mentioned in your paper, so why not use flexmatch as the baseline?
您好，看到您论文中提到flexmatch，想请教一下不以它为baseline的原因。
-
作者您好，请问您在实现FixMatch时，噪声标签的过滤是过滤掉低置信度的像素，只保留高置信度的像素参与训练？还是过滤掉低置信度的一整张图像，只保留高置信度的像素参与训练？
-
Greetings,
I've been studying this wonderful work recently. I wonder why you use `interleave` and `de_interleave` functions in `train.py`? Can `torch.cat` do the same thing?
-
Hi, thank for the nice code base. I have been trying to run the FixMatch on my own dataset, but I have encountered an issue.
When I use the default setting with uratio=1, by counting the indices of …
-
Hi,
Firstly, I'd like to thank you for your dedication and work on this library.
I have been successful in reproducing the results from the [USB CV Classic benchmark](https://github.com/microsoft/Se…
-
Thanks for your great work!
Just one quick question, how many GPUs were used to obtain the results in the paper? I didn't seem to find the specification on this.
Best
-
Hi,
Sorry to bother you again. I tried to reproduce the results reported in your paper on CIFAR100-LT. Here is my running: python train.py --dataset cifar100 --num-max 50 --num-max-u 400 --a…