fix(autoware_lidar_transfusion): set tensor names by matching with predefined values. #9057
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
github-actions
bot
added
component:perception
|
Thank you for contributing to the Autoware project! 🚧 If your pull request is in progress, switch it to draft mode. Please ensure:
|
amadeuszsz
requested changes
Oct 9, 2024
perception/autoware_lidar_transfusion/lib/network/network_trt.cpp
Outdated
Show resolved
Hide resolved
SamratThapa120
force-pushed
the
fix/transfusion/trt_output_order
branch
from
e2a28e1 to
2dfbe77
Compare
Signed-off-by: Samrat Thapa <samratthapa120@gmail.com>
Signed-off-by: Samrat Thapa <samratthapa120@gmail.com>
Signed-off-by: Samrat Thapa <samratthapa120@gmail.com>
Signed-off-by: Samrat Thapa <samratthapa120@gmail.com>
SamratThapa120
force-pushed
the
fix/transfusion/trt_output_order
branch
from
2dfbe77 to
9a3e6b7
Compare
amadeuszsz
approved these changes
Oct 9, 2024
SamratThapa120
added
the
run:build-and-test-differential
SamratThapa120
merged commit 23bb4a0
into
autowarefoundation:main
38 of 40 checks passed
SamratThapa120
added a commit
to tier4/autoware.universe
that referenced
this pull request
Oct 10, 2024
…edefined values. (autowarefoundation#9057) * set tensor order using api Signed-off-by: Samrat Thapa <samratthapa120@gmail.com> * style(pre-commit): autofix Signed-off-by: Samrat Thapa <samratthapa120@gmail.com> * fix tensor order Signed-off-by: Samrat Thapa <samratthapa120@gmail.com> * style(pre-commit): autofix Signed-off-by: Samrat Thapa <samratthapa120@gmail.com> * style fix Signed-off-by: Samrat Thapa <samratthapa120@gmail.com> * style(pre-commit): autofix --------- Signed-off-by: Samrat Thapa <samratthapa120@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
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.
Description
When making minor architectural changes to onnx checkpoints, we found that the tensorrt IO tensor order changed, which results in the current code failing because the order of tensors is fixed in current code.
autoware.universe/perception/autoware_lidar_transfusion/include/autoware/lidar_transfusion/utils.hpp
Line 39 in c2438fd
It seems that the tensor output order of the optimized tensorrt engine can be different from onnx IO order, but the tensor names match the onnx IO names. So, it is better to use the names from trt engine API to find the IO tensor positions.
We also found out that the current deployed models have faulty output names i.e the names of dir_cls_pred0 and bbox_pred0 are swapped, so they do not work with this PR. Therefore, we had to update the deployed onnx models with this PR.
Related links
Merge with autowarefoundation/autoware#5318
Parent Issue:
How was this PR tested?
Tested using TIER IV's test rosbags locally within docker environment.
Tested using webauto evaluator. Only once test fails, because the lidar does not include intensity field in it, and transfusion requires intensity value to function properly
Notes for reviewers
None.
Interface changes
None.