RuntimeError: shape '[1000, 1024, 1]' is invalid for input of size 8192000

[dbl001]

I tried running the file fourier_1d.py:

$ python fourier_1d.py 
Traceback (most recent call last):
  File "fourier_1d.py", line 161, in <module>
    x_train = torch.cat([x_train.reshape(ntrain,s,1), grid.repeat(ntrain,1,1)], dim=2)
RuntimeError: shape '[1000, 1024, 1]' is invalid for input of size 8192000

This statement is generating the exception:

x_train.reshape(ntrain,s,1)

From the documentation of view() method, 'The returned tensor shares the same data and must have the same number of elements, but may have a different size.'

Ideas?

[iamcalvinlau]

I made the assumption that he meant to have it as

sub = 2**3 #subsampling rate
h = 2**13 // sub
s = h

but I don't know if that is really the case.

[zongyi-li]

That's right. The raw data of the Burgers equation has resolution 8192. You can downsample it to the desired resolution.