Conv1d does not accept DoubleTensor

1

I have a minimal example to reproduce the issue:

It looks like Conv1d only accepts FloatTensor, and when it is fed DoubleTensor it errors out.

Here is a short example

import torch
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F

x_stub = Variable(torch.DoubleTensor(100, 15, 12).normal_(0, 1))
conv_1 = nn.Conv1d(15, 15, 3)
y = conv_1(x_stub)

so show what’s going on, I added the following line to the source

def conv1d(input, weight, bias=None, stride=1, padding=0, dilation=1,
           groups=1):
    f = ConvNd(_single(stride), _single(padding), _single(dilation), False,
               _single(0), groups, torch.backends.cudnn.benchmark, torch.backends.cudnn.enabled)
    =this line=> print(input, weight, bias)
    return f(input, weight, bias)

when running code, it gives me the following error message:

Variable containing:
( 0 ,.,.) = 
   1   1   1  ...    0   0   0
   0   0   0  ...    0   0   0
   0   0   0  ...    0   0   0
     ...       ⋱       ...    
   0   0   0  ...    0   0   0
   0   0   0  ...    0   0   0
   0   0   0  ...    0   0   1
... 

(127,.,.) = 
   1   1   1  ...    0   0   0
   0   0   0  ...    0   0   0
   0   0   0  ...    0   0   0
     ...       ⋱       ...    
   0   0   0  ...    0   0   0
   0   0   0  ...    0   1   0
   0   0   0  ...    0   0   1
[torch.DoubleTensor of size 128x12x15]

 Parameter containing:
(0 ,.,.) = 
  0.1286 -0.1301
 -0.0871  0.0397
  0.0317  0.0072
 -0.0406  0.0803
 -0.1885  0.1544
  0.1090 -0.1772
 -0.1818  0.0865
 -0.1696 -0.0973
 -0.1179 -0.0781
 -0.0745 -0.1268
  0.1303 -0.0950
  0.0804 -0.1008
...
(11,.,.) = 
 -0.1984 -0.1655
 -0.0531 -0.0365
 -0.1009  0.2038
 -0.0382  0.1492
 -0.1048 -0.1378
  0.0774  0.0515
 -0.0548 -0.1791
 -0.1805  0.0558
 -0.1805 -0.0603
 -0.1938  0.0465
 -0.1470 -0.0298
 -0.1597 -0.1718
[torch.FloatTensor of size 12x12x2]
 

Parameter containing:
 0.2023
 0.0939
-0.2037
 0.1501
-0.0270
 0.0494
 0.0637
-0.1420
 0.1512
-0.1538
-0.1828
 0.0366
[torch.FloatTensor of size 12]

Traceback (most recent call last):
  File "/Users/usr/projects/deep_learning_notes/pytorch_playground/grammar_variational_autoencoder/grammar_vae.py", line 69, in <module>
    losses += sess.train(train_loader, epoch)
  File "/Users/usr/projects/deep_learning_notes/pytorch_playground/grammar_variational_autoencoder/grammar_vae.py", line 26, in train
    recon_batch, mu, log_var = self.model(data)
  File "/Users/usr/anaconda/envs/deep-learning/lib/python3.6/site-packausrs/torch/nn/modules/module.py", line 206, in __call__
    result = self.forward(*input, **kwargs)
  File "/Users/usr/projects/deep_learning_notes/pytorch_playground/grammar_variational_autoencoder/model.py", line 90, in forward
    mu, log_var = self.encoder(x)
  File "/Users/usr/anaconda/envs/deep-learning/lib/python3.6/site-packausrs/torch/nn/modules/module.py", line 206, in __call__
    result = self.forward(*input, **kwargs)
  File "/Users/usr/projects/deep_learning_notes/pytorch_playground/grammar_variational_autoencoder/model.py", line 47, in forward
    h = self.conv_1(x)
  File "/Users/usr/anaconda/envs/deep-learning/lib/python3.6/site-packausrs/torch/nn/modules/module.py", line 206, in __call__
    result = self.forward(*input, **kwargs)
  File "/Users/usr/anaconda/envs/deep-learning/lib/python3.6/site-packausrs/torch/nn/modules/conv.py", line 143, in forward
    self.padding, self.dilation, self.groups)
  File "/Users/usr/anaconda/envs/deep-learning/lib/python3.6/site-packausrs/torch/nn/functional.py", line 69, in conv1d
    return f(input, weight, bias)
RuntimeError: expected Double tensor (got Float tensor)
2 
import torch
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F

x_stub = Variable(torch.DoubleTensor(100, 15, 12).normal_(0, 1))
conv_1 = nn.Conv1d(15, 15, 3).double()
y = conv_1(x_stub)
1 Like