Problem on running with fdtd.set_backend("torch")

[moienmovafagh]

**Hello,

When I execute my code with fdtd.set_backend("numpy"), it functions properly. However, switching to fdtd.set_backend("torch") leads to an error:**

Exception has occurred: TypeError 'torch.dtype' object is not callable File "C:\Users\yo0003\moein\FDTD.py", line 73, in grid[row:row+1,col:col+1,0] = fdtd.AbsorbingObject(permittivity=45.38,conductivity=0.77, name=str(row) + "," + str(col))

TypeError: 'torch.dtype' object is not callable

**this is my code:**  

import torch
import fdtd
import numpy as np
from line_profiler import LineProfiler
import nibabel as nib
import matplotlib.pyplot as plt

device = torch.device("cpu")

# Load the MNC file
mnc_img = nib.load('C:\\Users\\yo0003\\moein\\MATLAB_DRIVE\\Matlab\\subject04.mnc')

# Get the data as a numpy array
data = mnc_img.get_fdata()

pic=20*data[250,:,:]
pic = np.pad(pic, ((10, 10), (10, 10)), mode='constant', constant_values=0)
for i in range(len(pic[:,1])):
    for j in range(len(pic[1,:])):
        if pic[i,j]==160:
          pic[i,j]=20

        if pic[i,j]==80 or pic[i,j]==180 or pic[i,j]==100:
           pic[i,j]=120 

        if pic[i,j]==140 or pic[i,j]==200 : 
           pic[i,j]=220 

pic=pic[::4,::4]
plt.imshow(pic)
plt.show()
pic = np.pad(pic, ((40, 40), (40, 40)), 'constant', constant_values=((0, 0), (0, 0)))

# Set FDTD backend to CPU
fdtd.set_backend("torch")

# Create a 2D grid with a single layer in the z-direction
Lx=len(pic[:,1])
Ly=len(pic[1,:])
freq=1e9
landa=3e8/freq
grid = fdtd.Grid(shape=(Lx, Ly, 1), grid_spacing=0.1*landa)
Brain=np.zeros((Lx, Ly,2))

# # Add a light source
grid[20, 20, 0] = fdtd.PointSource(period=landa/3e8, name="source")

#*********************************************************
#                         avg
#*********************************************************
loc=np.where(pic==120)

try:
    # Your problematic code here, e.g.:
    grid[40:50,50:60,0] = fdtd.Object(permittivity=45)

except Exception as e:
    print("Error:", str(e))

    # Importing the traceback module to print the full stack trace
    import traceback
    print("Full traceback:\n", traceback.format_exc())

for row, col in zip(*loc):
 grid[row:row+1,col:col+1,0] = fdtd.AbsorbingObject(permittivity=45.38,conductivity=0.77, name=str(row) + "," + str(col))

[devbrones]

same issue with the 05-lenses-and-analysing-lensing-actions.ipynb example :/ (using python 3.9, pytorch=2.0.1=py3.9_cuda11.7_cudnn8.5.0_0, pytorch-cuda=11.7=h778d358_5)

[devbrones]

Appears to be caused by pr #63

[flaport]

Hi everyone. My apologies for taking so long to fix this. The fix is now available in fdtd>=0.3.3.

[devbrones]

Sorry, but i still cant get the examples to work properly, using 05-lenses-and-analysing-lensing-actions.ipynb. Setting the backend using fdtd.set_backend("torch.cuda") i get:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
[/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb) Cell 7 line 7
      [5](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#W6sZmlsZQ%3D%3D?line=4) for i, val in enumerate(np.flip(col)):
      [6](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#W6sZmlsZQ%3D%3D?line=5)     if val:
----> [7](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#W6sZmlsZQ%3D%3D?line=6)         grid[30 + i : 50 - i, j - 100 : j - 99, 0] = fdtd.Object(permittivity=1.5 ** 2, name=str(i) + "," + str(j))
      [8](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#W6sZmlsZQ%3D%3D?line=7)         break

File [~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:370](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:370), in Grid.__setitem__(self, key, attr)
    [367](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:367) else:
    [368](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:368)     raise KeyError("maximum number of indices for the grid is 3")
--> [370](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:370) attr._register_grid(
    [371](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:371)     grid=self,
    [372](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:372)     x=self._handle_single_key(x),
    [373](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:373)     y=self._handle_single_key(y),
    [374](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:374)     z=self._handle_single_key(z),
    [375](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:375) )

File [~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:73](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:73), in Object._register_grid(self, grid, x, y, z)
     [70](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:70) if bd.is_array(self.permittivity) and len(self.permittivity.shape) == 3:
     [71](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:71)     self.permittivity = self.permittivity[:, :, :, None]
     [72](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:72) self.inverse_permittivity = (
---> [73](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:73)     bd.ones((self.Nx, self.Ny, self.Nz, 3), dtype=self.permittivity.dtype)
     [74](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:74)     [/](https://file+.vscode-resource.vscode-cdn.net/) self.permittivity
     [75](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:75) )
     [77](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:77) # set the permittivity values of the object at its border to be equal
     [78](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:78) # to the grid permittivity. This way, the object is made symmetric.
     [79](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/objects.py:79) if self.Nx > 1:

TypeError: ones() got an unexpected keyword argument 'dtype'

[flaport]

version 0.3.4 should fix the issue

[devbrones]

please tell me if im doing something wrong here, but it still does not work? are you developing on a system that has cuda support?

TypeError                                 Traceback (most recent call last)
[/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb) Cell 15 line 1
      [7](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#X20sZmlsZQ%3D%3D?line=6)         plt.title(f"{i:3.0f}")
      [8](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#X20sZmlsZQ%3D%3D?line=7)         clear_output(wait=True) # only necessary in jupyter notebooks
---> [10](vscode-notebook-cell:/home/tb/repos/magisim/src/tests/05-lenses-and-analysing-lensing-actions.ipynb#X20sZmlsZQ%3D%3D?line=9) grid.save_data()  # saving detector readings

File [~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:505](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:505), in Grid.save_data(self)
    [503](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:503)     dic[detector.name + " (E)"] = [x for x in detector.detector_values()["E"]]
    [504](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:504)     dic[detector.name + " (H)"] = [x for x in detector.detector_values()["H"]]
--> [505](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/fdtd/grid.py:505) savez(path.join(self.folder, "detector_readings"), **dic)

File <__array_function__ internals>:180, in savez(*args, **kwargs)

File [~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:612](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:612), in savez(file, *args, **kwds)
    [528](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:528) @array_function_dispatch(_savez_dispatcher)
    [529](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:529) def savez(file, *args, **kwds):
    [530](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:530)     """Save several arrays into a single file in uncompressed ``.npz`` format.
    [531](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:531) 
    [532](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:532)     Provide arrays as keyword arguments to store them under the
   (...)
    [610](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:610) 
    [611](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:611)     """
--> [612](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/numpy/lib/npyio.py:612)     _savez(file, args, kwds, False)
...
--> [970](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/torch/_tensor.py:970)     return self.numpy()
    [971](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/torch/_tensor.py:971) else:
    [972](https://file+.vscode-resource.vscode-cdn.net/home/tb/repos/magisim/src/tests/~/anaconda3/envs/msimdev/lib/python3.9/site-packages/torch/_tensor.py:972)     return self.numpy().astype(dtype, copy=False)

TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.```

[flaport]

Hi, my apologies... I don't have my computer with cuda GPU anymore so before I thought I could just debug by using the torch.cpu backend, but it seems that this was a wrong assumption...

I spun up a VM with a GPU to debug this further. I can now confirm that all example notebooks now succesfully run with torch.cuda backend for fdtd>=0.3.5.

My apologies again for making you loose your time 🙂

[devbrones]

still appears broken in 0.3.5, when using fdtd.set_backend("torch.cuda") i get

python3.9/site-packages/fdtd/objects.py", line 80, in _register_grid
    self.inverse_permittivity[-1, :, :, 0] = self.grid.inverse_permittivity[
TypeError: can't assign a numpy.ndarray to a torch.cuda.DoubleTensor

[flaport]

which notebook is this in? Or is this in a custom notebook?

[devbrones]

this is in the lenses example 05-lenses-and-analysing-lensing-actions.ipynb all the path prefixes are just because i copied them to my project