onnxruntime/tools/ci_build/github/linux
Yifan Li 5c7e6b2e2a
[EP Perf] Add CI option to enable TRT-OSS parser (#19448)
### Description
<!-- Describe your changes. -->
* Introducing CI option to enable TRT-OSS parser, during ep perf
testing:

![image](https://github.com/microsoft/onnxruntime/assets/109183385/a9ba6393-6b94-4b8f-8ca4-ba7bc7954504)

By default, open-sourced onnx-tensorrt parser listed under
[cmake/deps.txt](https://github.com/microsoft/onnxruntime/blob/main/cmake/deps.txt#L39-L40)
will be used if enabling this option.


### To verify this option and check the difference during ORT image
build:
If this option is enabled:
<img width="649" alt="image"
src="https://github.com/microsoft/onnxruntime/assets/109183385/3b778583-451e-4617-ba8c-c064442e60fd">

If this option is not enabled (by default):
<img width="683" alt="image"
src="https://github.com/microsoft/onnxruntime/assets/109183385/cd8383ba-eff4-4536-94ab-a1424bb858ab">

* update default usage of cmake/trt version to the latest

### Motivation and Context
Make it easier to test oss parser and find potential gap between
tensorrt builtin/oss parser.

Schedule runs with oss parser will be set after this PR gets merged
2024-02-12 23:04:08 -08:00
..
docker
ort_minimal
tvm
build_cuda_c_api_package.sh
build_linux_python_package.sh
build_rocm_c_api_package.sh
build_tensorrt_c_api_package.sh
build_yocto.sh
copy_strip_binary.sh
create_package.sh
extract_and_bundle_gpu_package.sh
java_copy_strip_binary.sh
java_linux_final_test.sh
run_build.sh
run_dockerbuild.sh
run_python_dockerbuild.sh
run_python_dockertest.sh
run_python_tests.sh
test_custom_ops_pytorch_export.sh
upload_code_coverage_data.sh
upload_ortsrv_binaries.sh
yocto_build_toolchain.sh