When attempting to run 'feddg', the operation failed

[HoTuanLong]

I am attempting to run feddg using the following command:

python fed_run.py --mode fedavg --dg_method 'feddg' --source art_painting cartoon sketch --target photo --log

Unfortunately, I have encountered an error which states that there are missing files called img_{sample_name}.npy. I was wondering if you could kindly provide me with instructions on how to create the npy file for the data and assist me in resolving this issue.

[JeremyCJM]

Could you copy-paste the complete error message?

[HoTuanLong]

Of course, this is the output of my issue. Please keep in mind that the labels' names have been changed from PACS to 0123, respectively. 

=============Global iter is 0 =============== Traceback (most recent call last): File "/home/ubuntu/long.ht/CCST/federated/fed_run.py", line 703, in result = future.result() File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/concurrent/futures/_base.py", line 451, in result return self.__get_result() File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/concurrent/futures/_base.py", line 403, in __get_result raise self._exception File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/concurrent/futures/thread.py", line 58, in run result = self.fn(*self.args, **self.kwargs) File "/home/ubuntu/long.ht/CCST/federated/fed_run.py", line 487, in process_client train_loss, train_acc = train_meta(models[client_idx], train_loaders[client_idx], optimizers[client_idx], loss_fun, client_num, device_client, args, iter_idx, logger) File "/home/ubuntu/long.ht/CCST/federated/fed_run.py", line 100, in train_meta for it, data in enumerate(train_loader): File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 628, in next data = self._next_data() File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/site-packages/torch/utils/data/dataloader.py", line 671, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 58, in fetch data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/site-packages/torch/utils/data/_utils/fetch.py", line 58, in data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/ubuntu/long.ht/CCST/data/ImageLoader.py", line 317, in getitem tar_freq = self.amp_loader.get_amp() File "/home/ubuntu/long.ht/CCST/data/ImageLoader.py", line 238, in get_amp tar_freq = np.load(os.path.join(tar_freq_path)) File "/home/ubuntu/miniconda3/envs/ccst/lib/python3.10/site-packages/numpy/lib/npyio.py", line 405, in load fid = stack.enter_context(open(os_fspath(file), "rb")) FileNotFoundError: [Errno 2] No such file or directory: '/home/ubuntu/long.ht/CCST/datasets/PACS/1/fit/0/pic_006.npy

In addition, my data structure will be like this:

[JeremyCJM]

Sorry for getting back late for you. This is the pre-computed FFT amplitude of the images for the FedDG method. You can refer to the Official FedDG Repo to see how to compute the amplitude.

[HoTuanLong]

Thank you for your reply.

But maybe the prepare_dataset.py in the FedDG repo has not been completed. How do you generate the .npy file? Could you share your source or guide me to compute, please?

The following is their prepare_dataset.py, im_trg is not defined and where is amp will be used?

def extract_amp_spectrum(img_np):
    # trg_img is of dimention CxHxW (C = 3 for RGB image and 1 for slice)
    
    fft = np.fft.fft2( img_np, axes=(-2, -1) )
    amp_np, pha_np = np.abs(fft), np.angle(fft)

    return amp_sample

for client_idx in range(client_number):
    for data_idx, each_data_path in enumerate(client_data_list[client_idx]):
        img = Image.open(each_data_path)
        img = img.resize( (384,384), Image.BICUBIC )
        img_np = np.asarray(im_trg)

        amp = extract_amp_spectrum(img_np)
        np.save('./client{}/freq_amp_npy/amp_sample{}'.fprmat(client_idx, data_idx))