This repository was archived by the owner on Feb 3, 2025. It is now read-only.
Repository navigation
This repository was archived by the owner on Feb 3, 2025. It is now read-only.
[Jetson] No OpKernel was registered to support Op 'TRTEngineOp' #326
Description
I have been getting this error while running inference on TFTRT converted models (centernet_hg104_512x512_coco17_tpu-8, yolov4-tiny TF converted model). on both models I got the same error
Platform : Jetson TX2
OS : ubuntu 20.01
python: 3.6.9
Tensorflow: 2.3.0
Can you please help to find a solution for this error
Reacted by Maksim Kukushkin
Activity
DEKHTIARJonathan commentedon Sep 28, 2022
It looks like you are using a TensorFlow version that was compiled without TensorRT and TF-TRT support.
Where did you download tensorflow from ?
DEKHTIARJonathan commentedon Sep 28, 2022
@MattConley CC
[-]tensorflow.python.framework.errors_impl.InvalidArgumentError: No OpKernel was registered to support Op 'TRTEngineOp' used by {{node PartitionedCall/TRTEngineOp_0_2}} with these attrs: [output_shapes=[], workspace_size_bytes=522292960, max_cached_engines_count=1, segment_func=__inference_TRTEngineOp_0_2_native_segment_1517[], segment_funcdef_name="", use_calibration=false, fixed_input_size=true, input_shapes=[[?,416,416,3]], OutT=[DT_FLOAT], _allow_build_at_runtime=true, precision_mode="FP16", static_engine=false, serialized_segment="", cached_engine_batches=[], InT=[DT_FLOAT], calibration_data="", _use_implicit_batch=true] Registered devices: [CPU, XLA_CPU] Registered kernels: device='GPU' [[PartitionedCall/TRTEngineOp_0_2]] [Op:__inference_signature_wrapper_10677][/-][+][TF] No OpKernel was registered to support Op 'TRTEngineOp'[/+][-][TF] No OpKernel was registered to support Op 'TRTEngineOp'[/-][+][Jetson] No OpKernel was registered to support Op 'TRTEngineOp'[/+]Pacifist-99 commentedon Sep 28, 2022
I am using TF version 2.3 which was installed using method in this link
https://developer.download.nvidia.com/compute/redist/jp/v44 tensorflow==2.3.0+nv20.09
Pacifist-99 commentedon Sep 29, 2022
if you know Can you tell where should I download tensorflow 2.3.0 compiled with TRT
Pacifist-99 commentedon Oct 13, 2022
I have now compiled TensorRT with tensorflow but the issue percists
Is there any way to check if the compilation part done is correct or not?
DEKHTIARJonathan commentedon Oct 13, 2022
@MattConley to help