sink reshape for reorder+reshape+permute pattern opt #28183
+320
−0
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: fishbell <[email protected]>
Signed-off-by: fishbell <[email protected]>
Signed-off-by: fishbell <[email protected]>
songbell
changed the title
songbell
force-pushed
the
bell/fuse_reorder_reshape_transpose
branch
2 times, most recently
from
8362209 to
5c077b4
Compare
songbell
changed the title
songbell
force-pushed
the
bell/fuse_reorder_reshape_transpose
branch
from
5c077b4 to
58aafd5
Compare
Signed-off-by: fishbell <[email protected]>
Signed-off-by: fishbell <[email protected]>
songbell
force-pushed
the
bell/fuse_reorder_reshape_transpose
branch
2 times, most recently
from
68ba5b1 to
d233d75
Compare
Signed-off-by: fishbell <[email protected]>
songbell
force-pushed
the
bell/fuse_reorder_reshape_transpose
branch
from
3f765c1 to
98e39cc
Compare
peterchen-intel
pushed a commit
to peterchen-intel/openvino
that referenced
this pull request
Jan 2, 2025
From openvinotoolkit#28183 sink reshape for further optimization relaxed pattern match to allow reorder not in place use random values in test Signed-off-by: fishbell <[email protected]>
src/plugins/intel_gpu/src/graph/graph_optimizer/reshape_transfer.cpp
Outdated
Show resolved
Hide resolved
| auto itr = p.get_processing_order().begin(); | ||
| while (itr != p.get_processing_order().end()) { | ||
| auto& node = *itr++; | ||
| if (!node->is_type<permute>()) |
There was a problem hiding this comment.
Seems this pass doesn't support dynamic shape, why not check it at the beginning?
src/plugins/intel_gpu/src/graph/graph_optimizer/reshape_transfer.cpp
Outdated
Show resolved
Hide resolved
src/plugins/intel_gpu/tests/unit/passes/reorder_reshape_permute.cpp
Outdated
Show resolved
Hide resolved
…er.cpp Co-authored-by: River Li <[email protected]>
Signed-off-by: fishbell <[email protected]>
…com/songbell/openvino into bell/fuse_reorder_reshape_transpose
Draft
songbell
requested review from
sshlyapn and
vladimir-paramuzov
and removed request for
sshlyapn
yeonbok
reviewed
Jan 14, 2025
src/plugins/intel_gpu/src/graph/graph_optimizer/reshape_transfer.cpp
Outdated
Show resolved
Hide resolved
Signed-off-by: fishbell <[email protected]>
…com/songbell/openvino into bell/fuse_reorder_reshape_transpose
Signed-off-by: fishbell <[email protected]>
|
But why don't we do this at ngraph transform? |
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.
Details:
convolution ([N,W,C,H], byxf) -> reorder ([N,W,C,H), bfyx) -> reshape [N, W, CxH], bfyx -> permute order (0, 2,1) to [N, CxH, W] bfyx, which is equivalent/alias to ([N,W,C,H] byxf), and this execution chain can be optimized
Tickets: