fix dynamic shape inference in DepthToSpace #880
Merged
Conversation
This file contains 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
|
|
Signed-off-by: Dom Miketa <[email protected]>
Signed-off-by: Dom Miketa <[email protected]>
| @@ -40,7 +40,8 @@ def _common(cls, node, **kwargs): | |||
| if mode == "CRD": | |||
| # need native computation | |||
| bsize = attrs.get("blocksize") | |||
| batch, channel, height, width = x.shape | |||
| x_shape = tf.shape(x) | |||
There was a problem hiding this comment.
First, thanks very much for making the contribution! If you don't mind, please consider use the utility method tf_shape, which takes care of both fully defined and dynamic shapes, and provides slight optimization for static shape.
There was a problem hiding this comment.
Thanks for the suggestion, switched to tf_shape.
test/backend/test_dynamic_shape.py
Outdated
Comment on lines
251
to
252
| def test_depth_to_space(self): | ||
| b,c,h,w = shape = [2, 48, 5, 6] |
There was a problem hiding this comment.
The code looks fine. Please run yapf as described in https://github.com/onnx/onnx-tensorflow#code-standard to keep the code format and style consistent. Thanks.
There was a problem hiding this comment.
Thank you for pointing out that document, done on both files.
d-miketa
force-pushed
the
master
branch
2 times, most recently
from
c9ac540 to
fff017b
Compare
Signed-off-by: Dom Miketa <[email protected]>
Signed-off-by: Dom Miketa <[email protected]>
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Outdated suggestions cannot be applied.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.
Nets using DepthToSpace are currently incompatible with dynamic axes. It's enough to switch from
x.shapetotf.shape(x)along with manual unrolling of the resulting 4-dimensional tensor (as one is not allowed to directly iterate over a tf.Tensor).Related to #543 and #771
Edit: Added a unittest. It's my first one and my second PR ever, so fingers crossed..!