-
Hi, thank you for making the code available. Could you please be kind enough answer my questions?
> What is the reason behind adding noise when training the text classifier?
> What is the reason beh…
-
Using fixmatch to train on my dataset, LR = 0.03, batchsize = 64. When the unlabeled loss increases gradually, the model gets worse and worse, and the loss of training set will decrease, but the los…
-
## Bug
https://github.com/microsoft/Semi-supervised-learning/blob/main/semilearn/datasets/utils.py#L63
The dataset path is not being considered from args, instead it is hard-coded as 'data'
###…
-
Thanks for the excellent paper - I enjoyed reading it and think it addresses an important task.
Do you have any experience about applying the proposed GJS loss to semantic segmentation?
From my …
-
Greetings!
I've been studying your wonderful work DST, recently. I notice that you set different threshold to various baseline methods, which seems like an unfair experimental setting. I wonder why?
…
-
Hello,
I was reviewing the code for the RandAugment class and I noticed that the `m` parameter, which is set during initialization, does not seem to be utilized anywhere within the class. Additiona…
-
Hi there, I am trying to use fixmatch for training. During training, I set num_warmup_iter in the config, but according to the records of events in tensorboards, warmup does not seem to take effect. I…
-
Hi! As mentioned in [issue68](https://github.com/LiheYoung/UniMatch/issues/68), the img_u_w is to provide clean and informative BN statistics to the model, therefore you set model.train(). I wonder ho…
-
## Bug
Running eval.py produces this bug of dimension mismatch. I changed the net value in the eval.py to 'vit_base_patch16_96' as well.
### Reproduce the Bug
Running the command :
```
python…
-
I think UniMatch is a very innovative approach that builds upon FixMatch with some seemingly simple but actually very effective innovations. However, while FixMatch is a classification method, UniMatc…