Hi team,
I converted the tf ssd mobilnet v2 frozen graph into onnx model on jetson xavier. It is working well but when I tried to convert the ONNX model into TensorRT Engine. I am getting the below error.
Building an engine. This would take a while...
(Use "-v" or "--verbose" to enable verbose logging.)
[TensorRT] WARNING: onnx2trt_utils.cpp:220: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[TensorRT] WARNING: onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[TensorRT] WARNING: onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
ERROR: Failed to parse the ONNX file.
In node -1 (importResize): UNSUPPORTED_NODE: Assertion failed: scales.is_weights() && "Resize scales must be an initializer!"
ERROR: failed to build the TensorRT engine!
I used this script for conversion.
Hi,
Which TensorRT/JetPack version do you use?trtexec to see if it works?
/usr/src/tensorrt/bin/trtexec --onnx=[model]
Thanks.
Hi
Package: nvidia-jetpackAutonomous Machines | NVIDIA Developer
thanks,
could you exactly tell me ? how to call the trtexec ? where the output model will be saved?
Hi ,
when i used the trtexec . I am getting the below error.
./trtexec --onnx=/home/rachel2/Desktop/TRT_object_detection-master/1.1.0/outputmodelNewDataType.onnx --saveEngine=/home/rachel2/Desktop/TRT_object_detection-master/1.1.0/MODEL_frozen.engine
&&&& RUNNING TensorRT.trtexec # ./trtexec --onnx=/home/rachel2/Desktop/TRT_object_detection-master/1.1.0/outputmodelNewDataType.onnx --saveEngine=/home/rachel2/Desktop/TRT_object_detection-master/1.1.0/MODEL_frozen.engine
[10/22/2021-17:24:09] [I] === Model Options ===
[10/22/2021-17:24:09] [I] Format: ONNX
[10/22/2021-17:24:09] [I] Model: /home/rachel2/Desktop/TRT_object_detection-master/1.1.0/outputmodelNewDataType.onnx
[10/22/2021-17:24:09] [I] Output:
[10/22/2021-17:24:09] [I] === Build Options ===
[10/22/2021-17:24:09] [I] Max batch: 1
[10/22/2021-17:24:09] [I] Workspace: 16 MB
[10/22/2021-17:24:09] [I] minTiming: 1
[10/22/2021-17:24:09] [I] avgTiming: 8
[10/22/2021-17:24:09] [I] Precision: FP32
[10/22/2021-17:24:09] [I] Calibration:
[10/22/2021-17:24:09] [I] Safe mode: Disabled
[10/22/2021-17:24:09] [I] Save engine: /home/rachel2/Desktop/TRT_object_detection-master/1.1.0/MODEL_frozen.engine
[10/22/2021-17:24:09] [I] Load engine:
[10/22/2021-17:24:09] [I] Builder Cache: Enabled
[10/22/2021-17:24:09] [I] NVTX verbosity: 0
[10/22/2021-17:24:09] [I] Inputs format: fp32:CHW
[10/22/2021-17:24:09] [I] Outputs format: fp32:CHW
[10/22/2021-17:24:09] [I] Input build shapes: model
[10/22/2021-17:24:09] [I] Input calibration shapes: model
[10/22/2021-17:24:09] [I] === System Options ===
[10/22/2021-17:24:09] [I] Device: 0
[10/22/2021-17:24:09] [I] DLACore:
[10/22/2021-17:24:09] [I] Plugins:
[10/22/2021-17:24:09] [I] === Inference Options ===
[10/22/2021-17:24:09] [I] Batch: 1
[10/22/2021-17:24:09] [I] Input inference shapes: model
[10/22/2021-17:24:09] [I] Iterations: 10
[10/22/2021-17:24:09] [I] Duration: 3s (+ 200ms warm up)
[10/22/2021-17:24:09] [I] Sleep time: 0ms
[10/22/2021-17:24:09] [I] Streams: 1
[10/22/2021-17:24:09] [I] ExposeDMA: Disabled
[10/22/2021-17:24:09] [I] Spin-wait: Disabled
[10/22/2021-17:24:09] [I] Multithreading: Disabled
[10/22/2021-17:24:09] [I] CUDA Graph: Disabled
[10/22/2021-17:24:09] [I] Skip inference: Disabled
[10/22/2021-17:24:09] [I] Inputs:
[10/22/2021-17:24:09] [I] === Reporting Options ===
[10/22/2021-17:24:09] [I] Verbose: Disabled
[10/22/2021-17:24:09] [I] Averages: 10 inferences
[10/22/2021-17:24:09] [I] Percentile: 99
[10/22/2021-17:24:09] [I] Dump output: Disabled
[10/22/2021-17:24:09] [I] Profile: Disabled
[10/22/2021-17:24:09] [I] Export timing to JSON file:
[10/22/2021-17:24:09] [I] Export output to JSON file:
[10/22/2021-17:24:09] [I] Export profile to JSON file:
[10/22/2021-17:24:09] [I]
----------------------------------------------------------------
Input filename: /home/rachel2/Desktop/TRT_object_detection-master/1.1.0/outputmodelNewDataType.onnx
ONNX IR version: 0.0.5
Opset version: 10
Producer name:
Producer version:
Domain:
Model version: 0
Doc string:
----------------------------------------------------------------
[10/22/2021-17:24:11] [W] [TRT] onnx2trt_utils.cpp:220: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[10/22/2021-17:24:11] [W] [TRT] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[10/22/2021-17:24:11] [W] [TRT] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
ERROR: builtin_op_importers.cpp:2549 In function importResize:
[8] Assertion failed: scales.is_weights() && "Resize scales must be an initializer!"
[10/22/2021-17:24:11] [E] Failed to parse onnx file
[10/22/2021-17:24:11] [E] Parsing model failed
[10/22/2021-17:24:11] [E] Engine creation failed
[10/22/2021-17:24:11] [E] Engine set up failed
&&&& FAILED TensorRT.trtexec # ./trtexec --onnx=/home/rachel2/Desktop/TRT_object_detection-master/1.1.0/outputmodelNewDataType.onnx --saveEngine=/home/rachel2/Desktop/TRT_object_detection-master/1.1.0/MODEL_frozen.engine
Even I tried converting the tf model into UFF. I am able to convert the TF frozen graph into UFF. But UFF to tensorRT conversion is not working.
I saw somewhere on the Nvidia forum, UFF to TensorRT won’t work on Jetson Xavier.
Hi,
ERROR: builtin_op_importers.cpp:2549 In function importResize:
[8] Assertion failed: scales.is_weights() && "Resize scales must be an initializer!"
The error is related to some non-supported layer.
Would you mind upgrading your device to JetPack 4.6?
Thanks.
I had upgraded the jetpack into 4.6, but it is not working. I wrongly mentioned the SSD mobilenet version as 2. I am trying to convert the SSD mobilenet v3 into TensorRT Engine.
system
December 8, 2021, 2:52am
11
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.