[WIP] Update Jetpack to 4.6.1

[renzodgc]

Update the jetpack version of the dockerfile to 4.6.1 and thus the TensorRT version to 8.

This is a work in progress, right now a model can be generated (using trt_exporter.py) but fails when using it on the Pose Detection environment (See galliot-us/adaptive-pose-estimation#8)

The raised error has something to do with the batch size (even though it's correctly set), I tried to review the TensorRT documentation and try out some stuff but to no avail.

python3 inference/inference.py --device jetson --input_video softbio_vid.mp4 --out_dir inference/ --detector_model_path adaptive_object_detection/detectors/data/frozen_inference_graph.bin --label_map train_outputs/label_map.pbtxt --batch_size 2 --pose_model_path models/data/fast_pose_fp16_b2.trt 
2021-10-05 10:00:38.348486: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.2
WARNING:tensorflow:Deprecation warnings have been disabled. Set TF_ENABLE_DEPRECATION_WARNINGS=1 to re-enable them.
opened video  softbio_vid.mp4
WARNING:adaptive_object_detection.detectors.jetson_detector:
WARNING:adaptive_object_detection.detectors.jetson_detector:adaptive_object_detection/detectors/data/frozen_inference_graph.bin
WARNING:adaptive_object_detection.detectors.jetson_detector:
WARNING:root: 
WARNING:root:adaptive_object_detection/detectors/data/frozen_inference_graph.bin
WARNING:root: 
[TensorRT] WARNING: The logger passed into createInferRuntime differs from one already provided for an existing builder, runtime, or refitter. TensorRT maintains only a single logger pointer at any given time, so the existing value, which can be retrieved with getLogger(), will be used instead. In order to use a new logger, first destroy all existing builder, runner or refitter objects.

WARNING:root:started deserialize
WARNING:root:<tensorrt.tensorrt.ICudaEngine object at 0x7f73e99340>
[TensorRT] ERROR: 3: [executionContext.cpp::execute::473] Error Code 3: Internal Error (Parameter check failed at: runtime/api/executionContext.cpp::execute::473, condition: batchSize > 0 && batchSize <= mEngine.getMaxBatchSize()
)
[TensorRT] ERROR: 3: [executionContext.cpp::execute::473] Error Code 3: Internal Error (Parameter check failed at: runtime/api/executionContext.cpp::execute::473, condition: batchSize > 0 && batchSize <= mEngine.getMaxBatchSize()
)

[mahsa-shams]

Hi @renzodgc, I solved the issue and sent a this PR to you.

[renzodgc]

Hi @renzodgc, I solved the issue and sent a this PR to you.

Hi @mahsa-shams, Awesome! I'll merge to this pr and add you as a reviewer for my two drafted ones :)

EDIT: It seems I can't add you as a reviewer, but I marked the PRs to be ready for review