Error in Pretrained Version

[wj926]

I was running the Pretrained Version in Pytorch,
by downloading the files in google drive and then run python generate.py --load_path ./maxlen30
the CUDA is 8.0 and python is running in 3.5
However error message RuntimeError: invalid argument 2: dimension 2 out of range of 2D tensor at /pytorch/torch/lib/TH/generic/THTensor.c:24
Came out and I am not sure how to solve this problem.

Did you had similar problems

I will upload the whole error message.

(tensorflow) slcf@slcf:~/ARAE/pytorch$ python3 generate.py --load_path ./maxlen30/
{'noprint': False, 'ngenerations': 10, 'temp': 1, 'ninterpolations': 5, 'seed': 1111, 'outf': './generated.txt', 'steps': 5, 'load_path': './maxlen30/', 'sample': False}
Loading models from./maxlen30/
Traceback (most recent call last):
File "generate.py", line 135, in
main(args)
File "generate.py", line 74, in main
maxlen=model_args['maxlen'])
File "/home/slcf/ARAE/pytorch/models.py", line 325, in generate
sample=sample)
File "/home/slcf/ARAE/pytorch/models.py", line 270, in generate
inputs = torch.cat([embedding, hidden.unsqueeze(1)], 2)
File "/usr/local/lib/python3.5/dist-packages/torch/autograd/variable.py", line 897, in cat
return Concat.apply(dim, *iterable)
File "/usr/local/lib/python3.5/dist-packages/torch/autograd/_functions/tensor.py", line 316, in forward
ctx.input_sizes = [i.size(dim) for i in inputs]
File "/usr/local/lib/python3.5/dist-packages/torch/autograd/_functions/tensor.py", line 316, in
ctx.input_sizes = [i.size(dim) for i in inputs]
RuntimeError: invalid argument 2: dimension 2 out of range of 2D tensor at /pytorch/torch/lib/TH/generic/THTensor.c:24

[kellywzhang]

Hello, so I replicated this issue when using the newest version of PyTorch. The issue seems to be just differences in the default dimensions (whether they're kept or squeezed). I didn't try it with the pre-trained model, but changing the generate method in model.py to the following fixed it for me:

    def generate(self, hidden, maxlen, sample=True, temp=1.0):
            """Generate through decoder; no backprop"""

            batch_size = hidden.size(0)

            if self.hidden_init:
                    # initialize decoder hidden state to encoder output
                    state = (hidden.unsqueeze(0), self.init_state(batch_size))
            else:
                    state = self.init_hidden(batch_size)

            # <sos>
            self.start_symbols.data.resize_(batch_size)
            self.start_symbols.data.fill_(1)

            embedding = self.embedding_decoder(self.start_symbols)
            inputs = torch.cat([embedding, hidden], 1).unsqueeze(1)

            # unroll
            all_indices = []
            for i in range(maxlen):
                    output, state = self.decoder(inputs, state)
                    overvocab = self.linear(output.squeeze(1))

            if not sample:
                    vals, indices = torch.max(overvocab, 1)
            else:
                    # sampling
                    probs = F.softmax(overvocab/temp)
                    indices = torch.multinomial(probs, 1).squeeze(1)

            all_indices.append(indices.unsqueeze(1))

            embedding = self.embedding_decoder(indices)
            inputs = torch.cat([embedding, hidden], 1).unsqueeze(1)

            max_indices = torch.cat(all_indices, 1)

            return max_indices`

I'm thinking to make an update to the code at some point, but not yet as I don't want to change something that would lead to problems when running with the old version of PyTorch.

Let me know if you run into other issues. Best!

[JulesGM]

Hello. I would propose that this (#8 (comment)) be pushed to the main version, if you so desire, because it's been a while.
thanks for the help!