Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[GPU] Fix static_impl != nullptr assertion for reorder node #27786

Open
wants to merge 11 commits into
base: master
Choose a base branch
from
Open

[GPU] Fix static_impl != nullptr assertion for reorder node #27786

wants to merge 11 commits into from
+113 −0

Conversation

yuanxion
Copy link
Contributor

Details:

  • Fix ”Assertion `static_impl != nullptr' failed“ issue for reorder (Convolution_0_reorder_1) node by skipping mark node in shape flow.

Tickets:

@github-actions github-actions bot added the category: GPU OpenVINO GPU plugin label Nov 28, 2024
@yuanxion
Copy link
Contributor Author

For a transformer quantized model, reorder node is added for convolution, and is marked in shape_of subgraph. Thus it cannot get available (onednn & ocl) implementations and failed to run inference.
image (3)

Skip marking shape_of subgraphs for reorder node in this case.

@yuanxion yuanxion marked this pull request as ready for review November 29, 2024 01:14
@yuanxion yuanxion requested review from a team as code owners November 29, 2024 01:14
@yuanxion yuanxion force-pushed the fix-choose-conv-impl branch 3 times, most recently from 4cb8b68 to 710f965 Compare December 2, 2024 06:36
@yuanxion yuanxion force-pushed the fix-choose-conv-impl branch 2 times, most recently from 3e86c67 to ccb4ce5 Compare December 9, 2024 05:14
@yuanxion yuanxion force-pushed the fix-choose-conv-impl branch from ccb4ce5 to 457a67a Compare December 12, 2024 06:04
@yuanxion yuanxion marked this pull request as draft December 12, 2024 06:30
@yuanxion yuanxion marked this pull request as ready for review December 12, 2024 07:36
@sshlyapn sshlyapn added this to the 2025.0 milestone Dec 13, 2024
@yuanxion
Skip mark_node in shape flow for reorder node if dependency node is b…
f60e065 
…roadcase

Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Change to skip mark_node in shape flow for broadcast node if dependen…
73b4e49 
…cy node is shape_of

Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Change to skip mark reorder node (added for conv) in shape_of flow fo…
3ed8bf6 
…r quantized model

Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Check dependencies of reorder node all the way to input_layout node f…
157cc5b 
…or quantized model

Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Skip mark_node in shape flow for broadcast node if dependency nodes a…
32fcc41 
…re data and shape_of

Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Fix typo
611f9ba 
Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Check broadcast node in can_mark_node
8e6399f 
Signed-off-by: yuan.xiong <[email protected]>
@yuanxion
Extend mark_shape_of_subgraphs_test for broaccast with shape_of for q…
f25e82e 
…uantized models

Signed-off-by: yuan.xiong <[email protected]>
@yuanxion yuanxion force-pushed the fix-choose-conv-impl branch from 1fd12ac to f25e82e Compare December 17, 2024 09:37
@isanghao
Copy link
Contributor

no perf issue from static dgpu daily test
mingyuki@DUT1382DG2FRD:~/gpu-tools$ ./check_performance.py --this /home/mingyuki/log/pr/2024-12-19--18-49-PR27786.v2_report --ref ~/log/2024-12-19--09-36.v2_report --no_cldnn --version 2

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@yeonbok yeonbok yeonbok left review comments

@sshlyapn sshlyapn sshlyapn approved these changes

Assignees
No one assigned
Labels
category: GPU OpenVINO GPU plugin
Projects
None yet
Milestone
2025.0
Development

Successfully merging this pull request may close these issues.
4 participants
@yuanxion @isanghao @sshlyapn @yeonbok