facebookresearch/active-mri-acquisition

Active Acquisition for fastMRI

active-mri-acquisition's Issues

Check model.eval() and model.train()

In the current version of the code we always work in .train() mode. Is this correct?

Functions in fft_utils.py does not match version of torch>=1.8

Thanks for this repository.

In the file activemri/experimental/cvpr19_models/models/fft_utils.py, the functions of torch.fft, torch.ifft and torch.rfft of torch<=1.7 are used for the implementation of the Fast Fourier Transform. However, these functions are dropped in the new version of torch>=1.8 and replaced with torch.fft.fft, torch.fft.ifft and torch.fft.rfft. What's more, the packages this code depends on include fastmri which depends on torch>=1.8. The above problems will lead to a bug when running the intro notebook

It would be nice to update fft_utils.py to the new version.

Thanks,
Pu Yang

Shape mismatch issue in FTPASGANRAWModel for raw data

Shape mismatch issue in FTPASGANRAWModel for raw data
https://github.com/fairinternal/Active_Acquisition/blob/master/models/ft_pasganraw_model.py#L56

Traceback (most recent call last):
  File "/private/home/sumanab/.pycharm_helpers/pydev/pydevd.py", line 1741, in <module>
    main()
  File "/private/home/sumanab/.pycharm_helpers/pydev/pydevd.py", line 1735, in main
    globals = debugger.run(setup['file'], None, None, is_module)
  File "/private/home/sumanab/.pycharm_helpers/pydev/pydevd.py", line 1135, in run
    pydev_imports.execfile(file, globals, locals)  # execute the script
  File "/private/home/sumanab/.pycharm_helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile
    exec(compile(contents+"\n", file, 'exec'), glob, loc)
  File "/private/home/sumanab/code/fastMRI/train.py", line 39, in <module>
    model.optimize_parameters()
  File "/private/home/sumanab/code/fastMRI/models/ft_pasganraw_model.py", line 471, in optimize_parameters
    self.backward_D()
  File "/private/home/sumanab/code/fastMRI/models/ft_pasganraw_model.py", line 452, in backward_D
    self.loss_D_fake = self.criterionGAN(pred_fake, False, self.mask, degree=degree, pred_gt=(fake,self.real_B.permute(0,3,1,2)))
  File "/private/home/sumanab/code/fastMRI/models/ft_pasganraw_model.py", line 73, in __call__
    target_tensor = self.get_target_tensor(input, target_is_real, degree, mask, pred_gt)
  File "/private/home/sumanab/code/fastMRI/models/ft_pasganraw_model.py", line 62, in get_target_tensor
    target_tensor[:] = energy
RuntimeError: expand(torch.cuda.FloatTensor{[2, 640]}, size=[1280]): the number of sizes provided (1) must be greater or equal to the number of dimensions in the tensor (2)

Process finished with exit code 1

Remove use of fft_functions argument to preprocess_inputs

Move the functions to fft_utils.py