Skip to content

Add rnn args check#3925

Merged
soumith merged 3 commits intopytorch:masterfrom
zou3519:rnn-args
Dec 13, 2017
Merged

Add rnn args check#3925
soumith merged 3 commits intopytorch:masterfrom
zou3519:rnn-args

Conversation

@zou3519
Copy link
Copy Markdown
Contributor

@zou3519 zou3519 commented Nov 28, 2017 •
edited
Loading

Fixes #3851, #3259

Added a high-level check for arguments to RNNBase (these were moved from arg checks in cudnn/rnn.py)

Test Plan

python test/test_nn.py

apaszke
apaszke reviewed Nov 28, 2017
View reviewed changes
Copy link
Copy Markdown
Contributor

@apaszke apaszke left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good, but please check everything for LSTM.

Comment thread torch/nn/modules/rnn.py Outdated
mini_batch, self.hidden_size)
if self.mode == 'LSTM':
hidden = hidden[0]
if tuple(hidden.size()) != expected_hidden_size:

This comment was marked as off-topic.

Sign in to view

@apaszke apaszke closed this Nov 29, 2017
@apaszke apaszke reopened this Nov 29, 2017
@zou3519 zou3519 mentioned this pull request Dec 4, 2017
Closed
@zou3519 zou3519 mentioned this pull request Dec 12, 2017
Closed
apaszke
apaszke approved these changes Dec 12, 2017
View reviewed changes
Copy link
Copy Markdown
Contributor

@apaszke apaszke left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Didn't this PR have a test for these cases as well? It would be nice to add it back before merging.

@soumith soumith merged commit 28890b2 into pytorch:master Dec 13, 2017
@zou3519 zou3519 deleted the rnn-args branch January 3, 2018 19:59
@soumith soumith added the 0.3.1 label Feb 4, 2018
soumith pushed a commit that referenced this pull request Feb 7, 2018
@zou3519 @soumith
Add rnn args check (#3925)
a120a00 
* Add rnn args check

* Check both hidden sizes for LSTM

* RNN args check test
@Shandilya21
Copy link
Copy Markdown

Shandilya21 commented Jun 10, 2019

@zou3519 I have similar problem,

RuntimeError: Expected hidden[0] size (1, 64, 256), got (64, 256), i tried different ways but unable to get it.
here is the code snippet.:
def forward(self, input_src, input_trg, courteous_template, ctx_mask=None, trg_mask=None):
src_emb = self.src_embedding(input_src)
trg_emb = self.trg_embedding(input_trg)
temp_emb = self.temp_embedding(courteous_template)

    self.h0_encoder, self.c0_encoder = self.get_state(input_src)
    self.h1_encoder, self.c1_encoder = self.get_courteous(courteous_template)

    src_h, (src_h_t, src_c_t) = self.encoder(
        src_emb, (self.h0_encoder, self.c0_encoder)
    )
    

    temp_h, (tmp_h_t, tmp_c_t) = self.temp_encoder(
        temp_emb, (self.h1_encoder, self.c1_encoder)
    )

    out = torch.cat((src_h, temp_h),1)


    out = out.reshape(out.size(1),
    	out.size(0),out.size(2))

    # print(out.size())

    h_t = out[-1]
    h = h_t.view(h_t.size(0),h_t.size(1))



    trg_h, (_,_) = self.decoder(
    	trg_emb, h_t.view(
    		h_t.size(0),
    		h_t.size(1))
    	)

RuntimeError: Expected hidden[0] size (1, 64, 256), got (64, 256), i tried different ways but unable to get it.
please help to rectify this issues

@zou3519
Copy link
Copy Markdown
Contributor Author

zou3519 commented Jun 10, 2019 •
edited
Loading

@Shandilya21 Please ask a question on the forums or open a (new) issue if you think there is a bug.

laurentdupin pushed a commit to laurentdupin/pytorch that referenced this pull request Apr 24, 2026
@zou3519 @soumith
Add rnn args check (pytorch#3925)
fe31896 
* Add rnn args check

* Check both hidden sizes for LSTM

* RNN args check test
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@apaszke apaszke apaszke approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Improve error message for RuntimeError: inconsistent tensor size in RNN

4 participants

@zou3519 @Shandilya21 @apaszke @soumith