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>
Shorter plugin attribute test definitions
Fix incorrect names of some deserialize methods
Memory issues in PriorBoxPlugin
Signed-off-by: Shuyue Lan <shuyuel@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Limit output_dims == W[1] for plugin test
update embLayernorm plugin doc
Remove else after return as they are not required.
Fix FlattenConcat plugin readme
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>
This graceful failure is needed to enable tests under `trt_mod_test`.
`PLUGIN_ASSERT` is used in many places inside plugin methods, most of
which are `noexcept` and do not contain try-catch blocks. So changing
`PLUGIN_ASSERT` would require a major refactor which is beyond
rel-8.4. Instead, this MR introduces `PLUGIN_VALIDATE` as a temporary
replacement for `PLUGIN_ASSERT` in places where it is required to enable
valid/invalid attribute tests (mostly `createPlugin` and plugin constructors).
i.e. `PLUGIN_VALIDATE` implements the expected behavior of `PLUGIN_ASSERT`
(throws) while `PLUGIN_ASSERT` still aborts.
Signed-off-by: samurdhi karunaratne <skarunaratne@nvidia.com>
Signed-off-by: Rajeev Rao <rajeevrao@nvidia.com>
Documentation states:
```cpp
// Shape of boxes input should be
// [batch_size, num_boxes, 4] or [batch_size, num_boxes, 1, 4] or [batch_size, num_boxes, num_classes, 4]
```
however, the second possibility `[batch_size, num_boxes, 1, 4]` triggers a false positive in an
assertion. The third dimension must be either equal to 1 or the number
of classes.
Signed-off-by: Stephan Seitz <sseitz@nvidia.com>
Asfiya Baig