An output with one or more elements was resized since it had shape []

autograd
1

Hi there,
I’m using pytorch to run a custom differentiable “forward” routine I’m writing myself.
Essentially the routine is devised to synthesize audio by using a bank of sine oscillators (like Magenta’s DDSP). The audio works fine and the error seems to backpropagate through the net (I scrapped a code from the internet to plot the errors using, look for plot_grad_flow on the internet).

However, when autograd runs I get this warning, that tells me something’s not good:

<some path>/torch/autograd/__init__.py:251: UserWarning: An output with one or more elements was resized since it had shape [], which does not match the required output shape [4, 48001]. This behavior is deprecated, and in a future PyTorch release outputs will not be resized unless they have zero elements. You can explicitly reuse an out tensor t by resizing it, inplace, to zero elements with t.resize_(0). (Triggered internally at ../aten/src/ATen/native/Resize.cpp:28.)
  Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass

Providing a min working example is not easy, but this is the synthesis part in the forward method:

x = self.fullyLayers.forward(x)
# x is the output from a neural net, is a tensor of (15,)

freqs = x[:NOSC] # get frequency values from the NN output
... # I get a decay vector and an amplitude vector for all oscillators from x

freqs = freqs.reshape((-1,1)) # transpose of a 1D vector (will become Nx1)
... # doing the same for decays and amps

# DENORM FREQUENCY
self.freqs_denorm = 20.0 * SOME_CONSTANT_PITCH * (freqs + 0.5 + 1E-6) # treat them as multiples of F0 (start from F0/2 by adding +1)
self.freqs_denorm[0] = SOME_CONSTANT_PITCH

... # I also apply some denormalization to decay and amp to translate from the NN output range to a physical quantity

# HERE IS THE SYNTHESIS PART
t = torch.linspace(0, 1, 48000+1)      # shape of t is (48001,)
e = torch.exp(torch.tensor(1.0))
decay_matrix = self.amps_denorm * e ** (-t*self.decays_denorm)    # all of the following tensors will be (NOSC,48001)
freq_matrix = self.freqs_denorm.repeat((1,48000+1))
omegas = freq_matrix * (2.0 * torch.pi)
omegas = omegas / float(SAMPLING_FREQ)
phases = torch.cumsum(omegas,1)
audio = decay_matrix * torch.sin(phases+torch.pi/2)
audioS = torch.sum(audio, 0)     # here a sum together all the oscillators outputs so that audioS is (48001,)

So one of the last tensors, which is (4,48001) is not working as I would expect. But I don’t know where to start from. Backprop is done in C++ so I can’t debug.
Please note my batchsize is 1.

BTW: the issue has been raised previously but without reply

2

The error happens when you use out= which materializes the output of an operator in-place into an existing specified tensor, do you use that anywhere in your code?

>>> a = torch.tensor([1, 2, 3])
>>> b = torch.tensor([1, 2, 3, 4])
>>> torch.mul(a, 2, out=b)
<stdin>:1: UserWarning: An output with one or more elements was resized since it had shape [4], which does not match the required output shape [3]. This behavior is deprecated, and in a future PyTorch release outputs will not be resized unless they have zero elements. You can explicitly reuse an out tensor t by resizing it, inplace, to zero elements with t.resize_(0). (Triggered internally at /Users/jw3468/local/pytorch3/aten/src/ATen/native/Resize.cpp:35.)
tensor([2, 4, 6])
>>>

To get the stack trace in cpp you can use TORCH_SHOW_CPP_STACKTRACES=1.

3

Thank you, but nope… I don’t think I’m doing that. The only line where I’m reusing a variable is:
omegas = omegas / float(SAMPLING_FREQ)

I am also trying to use more tensors instead of reusing the same for many code lines.
I’m quite sure the error is over there, because the size (4,48001) is used only by tensors “decay_matrix”, “freq_matrix”, “omegas”, “phases”, and “audio”.

I can’t be doing anything wrong in other places, where other sizes are used.

One last question: where should I put TORCH_SHOW_CPP_STACKTRACES ?

4

Here’s the problem: computing e and using it to elevate power is not ok.

I solved the problem by replacing these lines:

t = torch.linspace(0, 1, 48000+1)      # shape of t is (48001,)
e = torch.exp(torch.tensor(1.0))
decay_matrix = self.amps_denorm * e ** (-t*self.decays_denorm)

with these one:

t = torch.linspace(0, 1, 48000+1)
decay_matrix = self.amps_denorm * torch.exp(-t*self.decays_denorm)

Not sure why, if someone can explain clearly it would be appreciated!

5

It’s an environment variable, so you can do TORCH_SHOW_CPP_STACKTRACES=1 python blah.py