While op forward for sentimental analysis #6140
Merged
Conversation
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
A v2 API like data feeder for book demos. We can feed data directly from reader.
reyoung
changed the title
…op_sentiment_analysis
…op_sentiment_analysis
QiJune
reviewed
Dec 4, 2017
| sentence = fluid.layers.data( | ||
| name='word', shape=[1], dtype='int64', lod_level=1) | ||
| sent_emb = fluid.layers.embedding( | ||
| input=sentence, size=[65535, 32], dtype='float32') |
reyoung
force-pushed
the
feature/while_op_sentiment_analysis
branch
2 times, most recently
from
c00349c to
6aae3ad
Compare
reyoung
force-pushed
the
feature/while_op_sentiment_analysis
branch
from
6aae3ad to
bf38d85
Compare
QiJune
reviewed
Dec 4, 2017
paddle/framework/backward.cc
Outdated
| return false; | ||
| if (ctrl_flow_ops.find(ops[op_index]->Type()) != | ||
| ctrl_flow_ops.end()) { | ||
| if (block_desc->HasVarRecursive(grad_var_name)) { |
There was a problem hiding this comment.
In fact, variable in Control operator will be used in nested blocks. It's created in parent block, but used in children block. So it should be find recursively.
| // FIXME: Hack code here | ||
| auto& ctrl_flow_ops = CtrlFlowOps(); | ||
| if (ctrl_flow_ops.find(op_desc->Type()) == ctrl_flow_ops.end()) { | ||
| // Only computational op need drop input's gradient. |
There was a problem hiding this comment.
In fact, the backward of Control Operator is not calculating gradient. So no_grad_vars does no sense to Control Operator.
| @@ -466,7 +466,12 @@ DDim CompileTimeInferShapeContext::GetDim(const std::string &name) const { | |||
| auto var = block_.FindVarRecursive(name); | |||
| PADDLE_ENFORCE(var != nullptr, "Cannot find variable %s", name); | |||
| try { | |||
| return framework::make_ddim(var->Shape()); | |||
| auto shape = var->Shape(); | |||
| if (shape.empty()) { | |||
There was a problem hiding this comment.
In which case, the shape is empty?
There was a problem hiding this comment.
The final step of RNN network. Memory's gradient could be empty.
QiJune
reviewed
Dec 4, 2017
| framework::CopyFrom(x_array[offset], dev_ctx.GetPlace(), dev_ctx, | ||
| out_tensor); | ||
| out_tensor->set_lod(x_array[offset].lod()); | ||
| } else { |
There was a problem hiding this comment.
In which case will offset > x_array.size() occur? Why not throw exception here?
There was a problem hiding this comment.
Final timestep, memory
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Fix #6139
Also, Fix #6134, Fix #6135
Base on PR #6102