RNN & LSTM input error

nlp
1

Hi, I am new to pytorch and meet a problem using LSTM. Maybe it is because of my input shape, but I don’t know how to do.
Here is my code (some lines omitted)

class LSTMModel(nn.Module):
	def __init__(self, in_dim, n_layer):
		super(LSTMModel, self).__init__()
		self.n_layer = n_layer
		self.hidden_dim = in_dim
		self.lstm = nn.LSTM(in_dim, self.hidden_dim, n_layer, batch_first=True)

	def forward(self, x):
		out, h = self.lstm(x)
		return h[0]


class MyModel(nn.Module):
	def __init__(self, config):
		super(TATransEModel, self).__init__()
		...

		self.lstm = LSTMModel(self.embedding_size, 1)

		..

	def unroll(self, data, unroll_len = 4):
		result = []
		for i in range(len(data) - unroll_len):
			result.append(data[i: i+unroll_len])
		return result

	def forward(self, pos_h, pos_t, pos_r, pos_tem, neg_h, neg_t, neg_r, neg_tem):
		pos_h_e = self.ent_embeddings(pos_h)
		pos_t_e = self.ent_embeddings(pos_t)
		pos_r_e = self.rel_embeddings(pos_r)
		pos_tem_e = []
		for tem in pos_tem:
			tem_e = []
			for token in tem:
				token_e = self.tem_embeddings(token)
				tem_e.append(token_e)
			pos_tem_e.append(tem_e)

		pos_rseq_e = []
		# add LSTM
		pos_input_r = self.unroll(pos_r_e)
		pos_hidden_r = self.lstm(pos_input_r)

The error message is:

Traceback (most recent call last):
  File "train_new.py", line 237, in <module>
    pos, neg = model(pos_h_batch, pos_t_batch, pos_r_batch, pos_time_batch, neg_h_batch, neg_t_batch, neg_r_batch, neg_time_batch)
  File "/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)
  File "/Users/genius/dynamic-KG-basic/model.py", line 92, in forward
    pos_hidden_r = self.lstm(pos_input_r)
  File "/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)
  File "/Users/genius/dynamic-KG-basic/model.py", line 30, in forward
    out, h = self.lstm(x)
  File "/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)
  File "/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py", line 165, in forward
    max_batch_size = input.size(0) if self.batch_first else input.size(1)
AttributeError: 'list' object has no attribute 'size'

Thank you for having a look!

2

input is a list, not tensor.

1 Like
3

Thank you so much! I wonder if I do torch.cat operation of two autograd tensors, will the result still be a autograd tensor?

4 
>>> import torch
>>> from torch.autograd import Variable
>>> a = torch.randn(2, 3)
>>> a_var = Variable(a, requires_grad=True)
>>> b = torch.randn(2, 3)
>>> b_var = Variable(b, requires_grad=True)
>>> a.requires_grad
False
>>> b.requires_grad
False
>>> a_var.requires_grad
True
>>> b_var.requires_grad
True
>>> c = torch.cat([a, b], 1)
>>> c.requires_grad
False
>>> c_var = torch.cat([a_var, b_var], 1)
>>> c_var.requires_grad
True

Note that :

Merging Tensor and Variable classes

torch.autograd.Variable and torch.Tensor are now the same class.
More precisely, torch.Tensor is capable of tracking history and behaves like the old Variable;
Variable wrapping continues to work as before but returns an object of type torch.Tensor.
This means that you don’t need the Variable wrapper everywhere in your code anymore.

1 Like
5

Thank you so much! It helps me a lot!