experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

seq2seq.py: Fixed the documentation to be consistent with the code (#3802)

Browse Source 
* Fixed the documentation to be consistent with the code

* Changed documentation according to @lukaszkaiser's comments

* Fixed indentation of comments as pointed out by Lukas
This commit is contained in:
Sahil Sharma 2016-09-01 21:00:03 +05:30 committed by Vijay Vasudevan
parent 1f681d207f
commit a8be586394
1 changed files with 13 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
tensorflow/python/ops/seq2seq.py
View File

@ -248,8 +248,11 @@ def embedding_rnn_decoder(decoder_inputs,
Returns: Returns:
A tuple of the form (outputs, state), where: A tuple of the form (outputs, state), where:
outputs: A list of the same length as decoder_inputs of 2D Tensors with outputs: A list of the same length as decoder_inputs of 2D Tensors. The
shape [batch_size x output_size] containing the generated outputs. output is of shape [batch_size x cell.output_size] when
output_projection is not None (and represents the dense representation
of predicted tokens). It is of shape [batch_size x num_decoder_symbols]
when output_projection is None.
state: The state of each decoder cell in each time-step. This is a list state: The state of each decoder cell in each time-step. This is a list
with length len(decoder_inputs) -- one item for each time-step. with length len(decoder_inputs) -- one item for each time-step.
It is a 2D Tensor of shape [batch_size x cell.state_size]. It is a 2D Tensor of shape [batch_size x cell.state_size].
@ -317,9 +320,11 @@ def embedding_rnn_seq2seq(encoder_inputs,
Returns: Returns:
A tuple of the form (outputs, state), where: A tuple of the form (outputs, state), where:
outputs: A list of the same length as decoder_inputs of 2D Tensors with outputs: A list of the same length as decoder_inputs of 2D Tensors. The
shape [batch_size x num_decoder_symbols] containing the generated output is of shape [batch_size x cell.output_size] when
outputs. output_projection is not None (and represents the dense representation
of predicted tokens). It is of shape [batch_size x num_decoder_symbols]
when output_projection is None.
state: The state of each decoder cell in each time-step. This is a list state: The state of each decoder cell in each time-step. This is a list
with length len(decoder_inputs) -- one item for each time-step. with length len(decoder_inputs) -- one item for each time-step.
It is a 2D Tensor of shape [batch_size x cell.state_size]. It is a 2D Tensor of shape [batch_size x cell.state_size].
@ -1079,7 +1084,9 @@ def model_with_buckets(encoder_inputs, decoder_inputs, targets, weights,
Returns: Returns:
A tuple of the form (outputs, losses), where: A tuple of the form (outputs, losses), where:
outputs: The outputs for each bucket. Its j'th element consists of a list outputs: The outputs for each bucket. Its j'th element consists of a list
of 2D Tensors of shape [batch_size x num_decoder_symbols] (jth outputs). of 2D Tensors. The shape of output tensors can be either
[batch_size x output_size] or [batch_size x num_decoder_symbols]
depending on the seq2seq model used.
losses: List of scalar Tensors, representing losses for each bucket, or, losses: List of scalar Tensors, representing losses for each bucket, or,
if per_example_loss is set, a list of 1D batch-sized float Tensors. if per_example_loss is set, a list of 1D batch-sized float Tensors.