[C++ API] RNN / GRU / LSTM layer refactoring #34322

yf225 · 2020-03-05T22:12:48Z

This PR refactors RNN / GRU / LSTM layers in C++ API to exactly match the implementation in Python API.

BC-breaking changes:

Instead of returning RNNOutput, RNN / GRU forward method now returns std::tuple<Tensor, Tensor>, and LSTM forward method now returns std::tuple<Tensor, std::tuple<Tensor, Tensor>>, matching Python API.
RNN / LSTM / GRU forward method now accepts the same inputs (input tensor and optionally hidden state), matching Python API.
RNN / LSTM / GRU layers now have forward_with_packed_input method which accepts PackedSequence as input and optionally hidden state, matching the forward(PackedSequence, ...) variant in Python API.
RNN / LSTM / GRU layers no longer have these fields: w_ih / w_hh / b_ih / b_hh. Instead, to access the weights and biases of the gates, users should do e.g. rnn->named_parameters()["weight_ih_l0"], which mirrors the Python API rnn.weight_ih_l0.
In RNNOptions
- tanh() / relu() / activation are removed. Instead, nonlinearity is added which takes either torch::kTanh or torch::kReLU
- layers -> num_layers
- with_bias -> bias
In LSTMOptions
- layers -> num_layers
- with_bias -> bias
In GRUOptions
- layers -> num_layers
- with_bias -> bias

The majority of the changes in this PR focused on refactoring the implementations in torch/csrc/api/src/nn/modules/rnn.cpp to match the Python API. RNN tests are then changed to reflected the revised API design.

dr-ci · 2020-03-05T23:16:28Z

💊 CircleCI build failures summary and remediations

As of commit 358fc46 (more details on the Dr. CI page):

1/1 failures introduced in this PR

🕵️ 1 new failure recognized by patterns

The following build failures do not appear to be due to upstream breakages:

pytorch_linux_backward_compatibility_check_test (1/1)

Step: "Test" (full log | pattern match details)

Mar 15 01:12:26 The PR is introducing backward incompatible changes to the operator library. Please contact PyTorch team to confirm whether this change is wanted or not.

Mar 15 01:12:26 processing existing schema:  aten::t(Tensor(a) self) -> (Tensor(a)) 
Mar 15 01:12:26 processing existing schema:  aten::to.other(Tensor self, Tensor other, bool non_blocking=False, bool copy=False, int? memory_format=None) -> (Tensor) 
Mar 15 01:12:26 processing existing schema:  aten::to.dtype(Tensor self, int dtype, bool non_blocking=False, bool copy=False, int? memory_format=None) -> (Tensor) 
Mar 15 01:12:26 processing existing schema:  aten::to.device(Tensor self, Device device, int dtype, bool non_blocking=False, bool copy=False, int? memory_format=None) -> (Tensor) 
Mar 15 01:12:26 processing existing schema:  aten::to.dtype_layout(Tensor self, *, int dtype, int layout, Device device, bool pin_memory=False, bool non_blocking=False, bool copy=False, int? memory_format=None) -> (Tensor) 
Mar 15 01:12:26 processing existing schema:  aten::to(Tensor(a) self, Device? device, int? dtype=None, bool non_blocking=False, bool copy=False) -> (Tensor(b|a)) 
Mar 15 01:12:26 processing existing schema:  aten::to(Tensor(a) self, int? dtype=None, bool non_blocking=False, bool copy=False) -> (Tensor(b|a)) 
Mar 15 01:12:26 processing existing schema:  aten::to(Tensor(a) self, bool non_blocking=False, bool copy=False) -> (Tensor(b|a)) 
Mar 15 01:12:26 processing existing schema:  aten::topk(Tensor self, int k, int dim=-1, bool largest=True, bool sorted=True) -> (Tensor values, Tensor indices) 
Mar 15 01:12:26 processing existing schema:  aten::trace(Tensor self) -> (Tensor) 
Mar 15 01:12:26 The PR is introducing backward incompatible changes to the operator library. Please contact PyTorch team to confirm whether this change is wanted or not.  
Mar 15 01:12:26  
Mar 15 01:12:26 Broken ops: [ 
Mar 15 01:12:26 	aten::_linear_packed(Tensor packed_weight, Tensor input) -> (Tensor) 
Mar 15 01:12:26 	aten::_linear_prepack(Tensor weight, Tensor? bias=None, float? output_min=None, float? output_max=None) -> (Tensor) 
Mar 15 01:12:26 	aten::_conv2d_packed(Tensor packed_weight, Tensor input) -> (Tensor) 
Mar 15 01:12:26 	aten::_conv2d_prepack(Tensor weight, Tensor? bias=None, int[2] stride=[1, 1], int[2] padding=[0, 0], int[2] dilation=[1, 1], int groups=1, float? output_min=None, float? output_max=None) -> (Tensor) 
Mar 15 01:12:26 ] 
Mar 15 01:12:26 + cleanup 
Mar 15 01:12:26 + retcode=1 
Mar 15 01:12:26 + set +x

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions on the GitHub issue tracker.

This comment has been revised 90 times.

facebook-github-bot

@yf225 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

torch/csrc/api/src/nn/modules/rnn.cpp

test/cpp/api/rnn.cpp

torch/csrc/api/src/nn/modules/rnn.cpp

torch/csrc/api/include/torch/nn/options/rnn.h

Co-Authored-By: Pavel Belevich <[email protected]>

facebook-github-bot

@yf225 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

facebook-github-bot

@yf225 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

facebook-github-bot · 2020-03-14T20:05:38Z

@yf225 merged this pull request in e23a9dc.

facebook-github-bot

@yf225 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

facebook-github-bot

@yf225 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

facebook-github-bot · 2020-03-16T04:09:44Z

@yf225 merged this pull request in bdd7dbf.

facebook-github-bot · 2020-03-16T04:09:54Z

@yf225 merged this pull request in bdd7dbf.

yf225 requested review from ebetica and goldsborough as code owners March 5, 2020 22:12

yf225 removed request for ebetica and goldsborough March 5, 2020 22:12

yf225 force-pushed the cpp_rnn_layers branch from a10e0a8 to 953bb95 Compare March 5, 2020 22:45

yf225 force-pushed the cpp_rnn_layers branch 23 times, most recently from 6e6cec2 to e71781b Compare March 6, 2020 17:04

yf225 changed the title ~~[WIP] C++ RNN layers refactoring~~ C++ RNN / GRU / LSTM layer refactoring Mar 6, 2020

facebook-github-bot reviewed Mar 12, 2020

View reviewed changes

pbelevich suggested changes Mar 14, 2020

View reviewed changes

Will Feng and others added 7 commits March 13, 2020 22:14

Update torch/csrc/api/src/nn/modules/rnn.cpp

c64b847

Co-Authored-By: Pavel Belevich <[email protected]>

Update test/cpp/api/rnn.cpp

2d3679c

Co-Authored-By: Pavel Belevich <[email protected]>

Update test/cpp/api/rnn.cpp

1424d9c

Co-Authored-By: Pavel Belevich <[email protected]>

Update test/cpp/api/rnn.cpp

ad0e8d1

Co-Authored-By: Pavel Belevich <[email protected]>

Update torch/csrc/api/src/nn/modules/rnn.cpp

61f32cc

Co-Authored-By: Pavel Belevich <[email protected]>

remove RNNOptionsBase virtual destructor

2541a7a

Merge branch 'master' into cpp_rnn_layers

820a0d7

pbelevich approved these changes Mar 14, 2020

View reviewed changes

facebook-github-bot reviewed Mar 14, 2020

View reviewed changes

Merge branch 'master' into cpp_rnn_layers

61de223

facebook-github-bot reviewed Mar 14, 2020

View reviewed changes

facebook-github-bot closed this in e23a9dc Mar 14, 2020

facebook-github-bot added the merged label Mar 14, 2020

yf225 reopened this Mar 14, 2020

yf225 removed the merged label Mar 14, 2020

facebook-github-bot reviewed Mar 14, 2020

View reviewed changes

fix lint

358fc46

facebook-github-bot reviewed Mar 15, 2020

View reviewed changes

facebook-github-bot closed this in bdd7dbf Mar 16, 2020

facebook-github-bot added the merged label Mar 16, 2020

facebook-github-bot deleted the cpp_rnn_layers branch July 13, 2020 17:54

mruberry added the Merged label Oct 28, 2020

dhern023 mentioned this pull request Jun 3, 2023

Get Chapter 11 to build using the latest libtorch PacktPublishing/Hands-On-Machine-Learning-with-CPP#11

Open

[C++ API] RNN / GRU / LSTM layer refactoring #34322

[C++ API] RNN / GRU / LSTM layer refactoring #34322

Uh oh!

Conversation

yf225 commented Mar 5, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dr-ci bot commented Mar 5, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CircleCI build failures summary and remediations

🕵️ 1 new failure recognized by patterns

pytorch_linux_backward_compatibility_check_test (1/1)

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Mar 14, 2020

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Mar 16, 2020

Uh oh!

facebook-github-bot commented Mar 16, 2020

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

yf225 commented Mar 5, 2020 •

edited

Loading

dr-ci bot commented Mar 5, 2020 •

edited

Loading