* Updated submodule onnx to 10.5
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* Updated files to 10.5
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* Restore the previous .gitmodules
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* .gitmodules new line
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* Update public docker files
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* Update 10.5 oss deleted files
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* Update deprecated APIs
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
---------
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
* Replaced non existing call setINit8Mode
INT8 ANN quantization is started by setting the builder flag kINT8 over the setFlags of the builder option
Signed-off-by: Francesco Spadafora <sf-fs@windowslive.com>
* Fix wring naming in example
Replaced name with tensor_name in example
Signed-off-by: Francesco Spadafora <sf-fs@windowslive.com>
* setStrictTypeConstraints updated to BuilderFlag::kINT8
The documentation regarding debugging of the INT8 mode is updated to match the builder flags
Signed-off-by: Francesco Spadafora <sf-fs@windowslive.com>
* typo fix
Signed-off-by: Francesco Spadafora <sf-fs@windowslive.com>
---------
Signed-off-by: Francesco Spadafora <sf-fs@windowslive.com>
Co-authored-by: Francesco Spadafora <francesco.spadafora@advertima.com>
Also adds goldenIO testcases: one for class agnostic NMS and another for class aware NMS.
With this, golden_io_snippet.py is also modified to support checking for multiple
plugin outputs.
Adds `class_agnostic` parameter to `EfficientNMS_TRT` plugin:
- Some object detection networks/architectures like YOLO series need to
use class-independent NMS operations.
- If `class_agnostic` is enabled, class-independent NMS is performed;
otherwise, different classes would do NMS separately.
Signed-off-by: Samurdhi Karunaratne <skarunaratne@nvidia.com>
Signed-off-by: Shuyue Lan <shuyuel@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Update tools including onnx-graphsurgeon(v0.3.25), Polygraphy(v0.43.1), tensorflow-quantization(v0.2.0)
Update onnx-tensorrt(8.5GA)
Mass merge sample, plugin, demo, python changes for TRT-8.5 GA release
Update demoBERT benchmark data for TensorRT 8.5.1
Signed-off-by: Simeng Liu <simengl@nvidia.com>
Signed-off-by: Simeng Liu <simengl@nvidia.com>
- trtexec should not call reportToProfiler if getEnqueueEmitsProfile is true
- Fix protobuf version to 3.15.7
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Add logic for supporting `'inputName':PATH` input format for trtexec's commandline options.
In trtexec documentation we state that input names can be wrapped in single quotes '', but we don't actually support that.
On Windows, --loadInputs flag will generally take a path that starts with `c:\..`, so splitting on last seen colon does not work in this case.
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Kevin Chen