I currently have my data defined as:
X = Variable(X_train_tensor)
y = Variable(y_train_tensor)
where X_train_tensor is a sparse FloatTensor and y_train_tensor is just 0’s and 1’s.
If have a module defined as:
class MyModule(nn.Module):
def __init__(self, num_labels, vocab_size, nonlin=F.relu, num_units=10):
super(MyModule, self).__init__()
self.dense0 = nn.Linear(vocab_size, num_units)
self.nonlin = nonlin
self.dropout = nn.Dropout(0.5)
self.dense1 = nn.Linear(num_units, 10)
self.output = nn.Linear(10, num_labels)
def forward(self, X, **kwargs):
X = self.nonlin(self.dense0(X))
X = self.dropout(X)
X = F.relu(self.dense1(X))
X = F.softmax(self.output(X), dim=-1)
return X
and I’m starting the train cycle with:
for epoch in range(1):
optimizer.zero_grad()
outputs = model(X)
But I get the error:
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-258-d820af66121f> in <module>()
1 for epoch in range(1):
2 optimizer.zero_grad()
----> 3 outputs = model(X)
~\Anaconda3\lib\site-packages\torch\nn\modules\module.py in __call__(self, *input, **kwargs)
355 result = self._slow_forward(*input, **kwargs)
356 else:
--> 357 result = self.forward(*input, **kwargs)
358 for hook in self._forward_hooks.values():
359 hook_result = hook(self, input, result)
<timed exec> in forward(self, X, **kwargs)
~\Anaconda3\lib\site-packages\torch\nn\modules\module.py in __call__(self, *input, **kwargs)
355 result = self._slow_forward(*input, **kwargs)
356 else:
--> 357 result = self.forward(*input, **kwargs)
358 for hook in self._forward_hooks.values():
359 hook_result = hook(self, input, result)
~\Anaconda3\lib\site-packages\torch\nn\modules\linear.py in forward(self, input)
53
54 def forward(self, input):
---> 55 return F.linear(input, self.weight, self.bias)
56
57 def __repr__(self):
~\Anaconda3\lib\site-packages\torch\nn\functional.py in linear(input, weight, bias)
833 if input.dim() == 2 and bias is not None:
834 # fused op is marginally faster
--> 835 return torch.addmm(bias, input, weight.t())
836
837 output = input.matmul(weight.t())
RuntimeError: size is not implemented for type torch.sparse.FloatTensor
Is is possible to use sparse Tensors in this way? I’m not able to fit the dense format in memory during training.