Hi,
When I run etempensv1 I got this error message:
pytorch version : 1.7.1
Files already downloaded and verified
Files already downloaded and verified
Tempens-v1 with epoch pseudo labels
------ Training epochs: 0 ------
/home/xps8500/anaconda3/envs/himp/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:136: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
Traceback (most recent call last):
File "main.py", line 165, in
run(config)
File "main.py", line 160, in run
trainer.loop(config.epochs, *loaders, scheduler=scheduler)
File "/home/xps8500/ssdlp/trainer/eTempensv1.py", line 125, in loop
self.train(train_data, self.print_freq)
File "/home/xps8500/ssdlp/trainer/eTempensv1.py", line 99, in train
return self.train_iteration(data_loader, print_freq)
File "/home/xps8500/ssdlp/trainer/eTempensv1.py", line 56, in train_iteration
self.epoch_pslab[idxs] = outputs.clone().detach()
RuntimeError: Index put requires the source and destination dtypes match, got Long for the destination and Float for the source.

The error message from etempensv2 is:

pytorch version : 1.7.1
Files already downloaded and verified
Files already downloaded and verified
Tempens-v2 with epoch pseudo labels
------ Training epochs: 0 ------
/home/xps8500/anaconda3/envs/himp/lib/python3.6/site-packages/torch/cuda/init.py:52: UserWarning: CUDA initialization: Unexpected error from cudaGetDeviceCount(). Did you run some cuda functions before calling NumCudaDevices() that might have already set an error? Error 804: forward compatibility was attempted on non supported HW (Triggered internally at /pytorch/c10/cuda/CUDAFunctions.cpp:100.)
return torch._C._cuda_getDeviceCount() > 0
/home/xps8500/anaconda3/envs/himp/lib/python3.6/site-packages/torch/optim/lr_scheduler.py:136: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
Traceback (most recent call last):
File "main.py", line 165, in
run(config)
File "main.py", line 160, in run
trainer.loop(config.epochs, *loaders, scheduler=scheduler)
File "/home/xps8500/ssdlp/trainer/eTempensv2.py", line 127, in loop
self.train(label_data, unlab_data, self.print_freq)
File "/home/xps8500/ssdlp/trainer/eTempensv2.py", line 101, in train
return self.train_iteration(label_loader, unlab_loader, print_freq)
File "/home/xps8500/ssdlp/trainer/eTempensv2.py", line 55, in train_iteration
uloss = self.mse_loss(unlab_outputs, iter_unlab_pslab)
File "/home/xps8500/ssdlp/utils/loss.py", line 34, in mse_with_softmax
return F.mse_loss(F.softmax(logit1,1), F.softmax(logit2,1))
File "/home/xps8500/anaconda3/envs/himp/lib/python3.6/site-packages/torch/nn/functional.py", line 1512, in softmax
ret = input.softmax(dim)
RuntimeError: "softmax_lastdim_kernel_impl" not implemented for 'Long'

Thanks!

May I ask what the difference is between iFixMatch and eFixmatch? Thanks.

Hi, thanks for sharing. I have been playing with FixMatch lately. Do you find eFixMatch perform better than iFixMatch? The comparison doesn't show up in the readme.

Also, do you have experience deploying FixMatch on imbalanced data? I can't make it work so far. FixMatch performs equivalent to or worse than a naive supervised learning method.

I created an issue on google's repo: google-research/fixmatch#52. Welcome any comments.

I try to apply FixMatch on one class data.
For the unsupervised loss part, I modified the code like that.

self.uce_loss   = torch.nn.BCEWithLogitsLoss(reduction='none')
with torch.no_grad():
    woutputs = self.model(weak_u)
    woutputs = torh.sigmoid(woutput)
      
           
mask = woutputs .ge(self.threshold).float()
soutputs = self.model(strong_u)
uloss  = torch.mean(mask* self.uce_loss(soutputs, mask) )

Is that correct?

Nice repository! What is the difference between epseudo and ipseudo algortihms, and the difference between versions?
Thanks in advance.

it is a really nice work for semi-surpervised leaning. would you mind provide the performance of your code compared with paper performance?

Hello, it's not that semi supervised learning like miamache can't be used on unbalanced data sets. I use it on my own data set. the test set accurate rises by 2 epochs and then decreases continuously. However, the loss of training set is declining. What's the matter? Is my dataset too unbalanced

I found that there are bugs in the code, and the data in the dev set may be mixed in the test set ！