Cond_op with dynamic if-else checked-in#4088
Merged
zchen0211 merged 19 commits intoPaddlePaddle:developfrom Sep 14, 2017
Merged
Cond_op with dynamic if-else checked-in#4088zchen0211 merged 19 commits intoPaddlePaddle:developfrom
zchen0211 merged 19 commits intoPaddlePaddle:developfrom
Conversation
Fix CI test
qingqing01
reviewed
Sep 14, 2017
paddle/framework/tensor.h
Outdated
| inline T* mutable_data(DDim dims, platform::Place place); | ||
|
|
||
| /*! Size of a single element in data() */ | ||
| inline size_t element_size() const { return holder_->element_size(); } |
Contributor
There was a problem hiding this comment.
Actually, I think there is no need to add this function. A function like Variable::IsType() is more useful: https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/variable.h#L41
| tensor_child->mutable_data<float>(dim, platform::CPUPlace()); | ||
|
|
||
| Gather<float>(dev_ctx.GetPlace(), tensor_parent, &index_tensors[i], | ||
| tensor_child); |
Contributor
There was a problem hiding this comment.
The recurrent operator also uses explicit type float in the implementation. CondOp has the same problem. Maybe the RecurrentOp and CondOp should like:
template <T>
class CondOp : public framework::OperatorBase {
}Then specialized the class.
Superjomn
reviewed
Sep 14, 2017
paddle/operators/cond_op.cc
Outdated
|
|
||
| for (auto& output : Outputs("Outs")) { | ||
| Tensor* tensor_t_out = sub_scopes[0]->FindVar(output)->GetMutable<Tensor>(); | ||
| PADDLE_ENFORCE_NOT_NULL(tensor_t_out, "True output should be NULL"); |
paddle/operators/cond_op.cc
Outdated
| Tensor* tensor_t_out = sub_scopes[0]->FindVar(output)->GetMutable<Tensor>(); | ||
| PADDLE_ENFORCE_NOT_NULL(tensor_t_out, "True output should be NULL"); | ||
| Tensor* tensor_f_out = sub_scopes[1]->FindVar(output)->GetMutable<Tensor>(); | ||
| PADDLE_ENFORCE_NOT_NULL(tensor_f_out, "True output should be NULL"); |
paddle/operators/cond_op.cc
Outdated
|
|
||
| void CondOp::Run(const Scope& scope, | ||
| const platform::DeviceContext& dev_ctx) const { | ||
| auto sub_scopes = scope.FindVar("SubScopes")->Get<std::vector<Scope*>>(); |
Contributor
There was a problem hiding this comment.
enforce scope.FindVar("SubScopes") is not null first
then Get.
paddle/operators/cond_op.cc
Outdated
| const platform::DeviceContext& dev_ctx) const { | ||
| auto sub_scopes = scope.FindVar("SubScopes")->Get<std::vector<Scope*>>(); | ||
| auto index_tensors = | ||
| scope.FindVar("IndexTensors")->Get<std::vector<Tensor>>(); |
paddle/operators/cond_op.cc
Outdated
| } | ||
|
|
||
| // Step 3: run | ||
| for (int i = 0; i < 2; ++i) sub_net_op_[i]->Run(*sub_scopes[i], dev_ctx); |
Contributor
There was a problem hiding this comment.
add { }
every cond operator should be wrapped with {}
paddle/operators/cond_op.cc
Outdated
|
|
||
| AddComment(R"DOC( | ||
| Sample dependent Cond Operator: | ||
| The equation is: Out[i] = subnet_t[i], if Cond[i] == true |
paddle/operators/cond_op.h
Outdated
| */ | ||
| void InferShape(const framework::Scope& scope) const override; | ||
|
|
||
| // Set True Block |
Contributor
There was a problem hiding this comment.
comment format should be unified.
use
/*
* some comment
*/
paddle/operators/cond_op.h
Outdated
| } | ||
|
|
||
| // Set False Block | ||
| void set_falsenet(std::unique_ptr<OperatorBase> net) { |
Contributor
There was a problem hiding this comment.
std::unique_ptr &&net)
global reference
|
|
||
| namespace paddle { | ||
| namespace operators { | ||
|
|
Superjomn
approved these changes
Sep 14, 2017
paddle/operators/cond_op.h
Outdated
| * | ||
| * if cond == 0, it will run false_net, which is another NetOp. | ||
| */ | ||
|
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Implemented the dynamic condition (if/else) op. The result matches my python results.
TODO: Will implement the backward part soon.