[ Layers ] Bug Fix for NHWC Support #2686
Conversation
jijoongmoon
requested review from
myungjoo,
again4you,
jaeyun-jung,
leemgs,
wooksong,
helloahn,
kparichay,
gichan-jang,
anyj0527,
zhoonit,
lhs8928,
songgot,
jihochu,
DonghakPark,
SeoHyungjun,
baek2sm,
skykongkong8,
djeong20,
EunjuYang and
a team
as code owners
|
📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2686. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/. |
jijoongmoon
force-pushed
the
flatten_nhwc
branch
from
814d0a3
to
5b43d99
Compare
There was a problem hiding this comment.
@jijoongmoon, 💯 All CI checkers are successfully verified. Thanks.
This PR fixes the bug about the inplace layers when the channel last is enabled. Previously, the input var_grad tensor's format is not changed in inplace layers. **Self evaluation:** 1. Build test: [X]Passed [ ]Failed [ ]Skipped 2. Run test: [X]Passed [ ]Failed [ ]Skipped Signed-off-by: jijoong.moon <[email protected]>
jijoongmoon
force-pushed
the
flatten_nhwc
branch
from
5b43d99
to
01fe193
Compare
There was a problem hiding this comment.
@jijoongmoon, 💯 All CI checkers are successfully verified. Thanks.
| @@ -779,17 +779,21 @@ NetworkGraph::finalizeContext(const std::shared_ptr<LayerNode> &lnode, | |||
| TensorSpecV2::RequestType::READ_ONLY_VIEW; | |||
| if (lnode->getType() == IdentityLayer::type) { | |||
| s.variable_spec.reference_name = inputs[i]->getName(); | |||
| s.variable_spec.dim.setFormat(inputs[i]->getDim().getFormat()); | |||
There was a problem hiding this comment.
how about using getFormat() directly instead of calling getDim()?
| s.variable_spec.dim.setFormat(inputs[i]->getDim().getFormat()); | |
| s.variable_spec.dim.setFormat(inputs[i]->getFormat()); |
using Tensor::getFormat() could also have a memory advantage since getDim() returns a copy of the TensorDim.
| target_shape = "target_shape=" + std::to_string(target_dim[1]) + ":" + | ||
| std::to_string(target_dim[2]) + ":" + | ||
| std::to_string(target_dim[3]); |
There was a problem hiding this comment.
How about adding a getter like getShape() to tensor_dim?
| target_shape = "target_shape=" + std::to_string(target_dim[1]) + ":" + | |
| std::to_string(target_dim[2]) + ":" + | |
| std::to_string(target_dim[3]); | |
| target_shape = "target_shape=" + target_dim.getShape() |
In tensor_dim, MAXDIM is set to 4.
It would be beneficial if the getShape function could create and return a string that matches the dimensions.
This PR fixes the bug about the inplace layers when the channel last is enabled.
Previously, the input var_grad tensor's format is not changed in inplace layers.
Self evaluation: