TAO 5.0 Classification (PyTorch) deploy error

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Please provide the following information when requesting support.

• Hardware (RTX A6000)
• Network Type (Classification)

I followed the tutorial Image Classification PyT to generate the ONNX model file, and then deployed to deepstream according to the tutorial Deploying to DeepStream for Classification TF1/TF2/PyTorch, and I got the following error when running the program, how to solve it?

I’ve tried onnx-simplifier but it doesn’t work.

cpp:680:gst_nvinfer_logger:<primary_gie> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::buildModel() <nvdsinfer_context_impl.cpp:1923> [UID = 1]: Trying to create engine from model files
WARNING: [TRT]: onnx2trt_utils.cpp:377: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
SSA validation FAIL: tensor __mye97 has multiple definitions
peephole.cpp:657: DCHECK(g_->ssa_validation()) failed. 
Aborted (core dumped)
3

Refer to Unable to run TAO toolkit in Jetson device - #5 by Morganh

4

I’ve read the relevant answers you mentioned, but haven’t found the answer I need, can you help me? Thank you

5

Please follow Deploying to DeepStream for Classification TF1/TF2/PyTorch - NVIDIA Docs to use deepstream-app to run inference. The similar example for config file can be found in Issue with image classification tutorial and testing with deepstream-app - #21 by Morganh

6

I have adopted the configuration file you provided, but still report the same error. I think it may not be a problem with the configuration file. Here are all the files I used, please help to see if there is a problem.

onnx-file:
classification_model_export.onnx (28.2 MB)
configuration file:
ds_classification_as_primary_gie.txt (3.5 KB)
config_as_primary_gie.txt (788 Bytes)
label file:
labels.txt (10 Bytes)

Thanks!

7

Could you please share full command and full log? Thanks.

8

yeah

lab@lab:/opt/nvidia/deepstream/deepstream/sources/apps/sample_apps/deepstream-class$ deepstream-app -c ds_classification_as_primary_gie.txt 
Warning: 'input-dims' parameter has been deprecated. Use 'infer-dims' instead.
gstnvtracker: Loading low-level lib at /opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so
gstnvtracker: Batch processing is ON
gstnvtracker: Past frame output is OFF
[NvTrackerParams::getConfigRoot()] !!![WARNING] Empty config file path is provided. Will go ahead with default values
[NvMultiObjectTracker] Initialized
0:00:00.108288222 28822 0x561b18c25720 INFO                 nvinfer gstnvinfer.cpp:680:gst_nvinfer_logger:<primary_gie> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::buildModel() <nvdsinfer_context_impl.cpp:1923> [UID = 1]: Trying to create engine from model files
WARNING: [TRT]: onnx2trt_utils.cpp:377: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
SSA validation FAIL: tensor __mye53 has multiple definitions
peephole.cpp:657: DCHECK(g_->ssa_validation()) failed. 
Aborted (core dumped)
9

Please try to pull deepstream docker nvcr.io/nvidia/deepstream:6.2-triton and then run again inside it.
$ docker run --runtime=nvidia -it nvcr.io/nvidia/deepstream:6.2-triton /bin/bash

10

Sorry, I was using TAO for the first time and I didn’t know how to use Trition. How do I put model files and configuration files into an image?

11

You can use -v argument.
$ docker run --runtime=nvidia -it -v your/local/path:/docker/path nvcr.io/nvidia/deepstream:6.2-triton /bin/bash

12

I tried and got the same error, is there a problem with my model file?

root@9a392fcbd0cb:/opt/nvidia/deepstream/deepstream-6.2# cd /model/
root@9a392fcbd0cb:/model# ls
classification_model_export.onnx  ds_classification_as_primary_gie.txt
config_as_primary_gie.txt         ds_config.txt
config_infer_primary.txt          labels.txt
root@9a392fcbd0cb:/model# deepstream-app -c ds_classification_as_primary_gie.txt 

(gst-plugin-scanner:119): GStreamer-WARNING **: 01:54:35.411: Failed to load plugin '/usr/lib/x86_64-linux-gnu/gstreamer-1.0/deepstream/libnvdsgst_udp.so': librivermax.so.0: cannot open shared object file: No such file or directory

(gst-plugin-scanner:119): GStreamer-WARNING **: 01:54:35.503: Failed to load plugin '/usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstopenmpt.so': libmpg123.so.0: cannot open shared object file: No such file or directory

(gst-plugin-scanner:119): GStreamer-WARNING **: 01:54:35.508: Failed to load plugin '/usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstmpeg2dec.so': libmpeg2.so.0: cannot open shared object file: No such file or directory

(gst-plugin-scanner:119): GStreamer-WARNING **: 01:54:35.512: Failed to load plugin '/usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstmpeg2enc.so': libmpeg2encpp-2.1.so.0: cannot open shared object file: No such file or directory

(gst-plugin-scanner:119): GStreamer-WARNING **: 01:54:35.535: Failed to load plugin '/usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstmpg123.so': libmpg123.so.0: cannot open shared object file: No such file or directory

(gst-plugin-scanner:119): GStreamer-WARNING **: 01:54:35.539: Failed to load plugin '/usr/lib/x86_64-linux-gnu/gstreamer-1.0/libgstchromaprint.so': libavcodec.so.58: cannot open shared object file: No such file or directory
Warning: 'input-dims' parameter has been deprecated. Use 'infer-dims' instead.
gstnvtracker: Loading low-level lib at /opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so
gstnvtracker: Batch processing is ON
gstnvtracker: Past frame output is OFF
[NvTrackerParams::getConfigRoot()] !!![WARNING] Empty config file path is provided. Will go ahead with default values
[NvMultiObjectTracker] Initialized
0:00:00.701462433   118 0x5618dfa85ed0 INFO                 nvinfer gstnvinfer.cpp:680:gst_nvinfer_logger:<primary_gie> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::buildModel() <nvdsinfer_context_impl.cpp:1923> [UID = 1]: Trying to create engine from model files
WARNING: [TRT]: onnx2trt_utils.cpp:377: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
WARNING: [TRT]: onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
SSA validation FAIL: tensor __mye53 has multiple definitions
peephole.cpp:657: DCHECK(g_->ssa_validation()) failed. 
Aborted (core dumped)
13

I was prompted with the following information when exporting the model with tao, does this matter?

2023-09-04 05:34:25,669 - mmcls - INFO - ********************** Start logging for Export **********************
load checkpoint from local path: /tmp/tmph7dwtz6c.pth
/usr/local/lib/python3.8/dist-packages/mmcv/onnx/symbolic.py:481: UserWarning: DeprecationWarning: This function will be deprecated in future. Welcome to use the unified model deployment toolbox MMDeploy: https://github.com/open-mmlab/mmdeploy
  warnings.warn(msg)
/usr/local/lib/python3.8/dist-packages/nvidia_tao_pytorch/cv/classification/models/fan.py:231: UserWarning: Replacing default adatpive_avg_pool2d to custom implementation for ONNX export
  warnings.warn("Replacing default adatpive_avg_pool2d to custom implementation for ONNX export")
/usr/local/lib/python3.8/dist-packages/nvidia_tao_pytorch/cv/backbone/fan.py:603: TracerWarning: Converting a tensor to a Python float might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  kernel_width = math.ceil(inp_size[2] / size[0])
/usr/local/lib/python3.8/dist-packages/nvidia_tao_pytorch/cv/backbone/fan.py:604: TracerWarning: Converting a tensor to a Python float might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  kernel_height = math.ceil(inp_size[3] / size[1])
/usr/local/lib/python3.8/dist-packages/nvidia_tao_pytorch/cv/classification/models/fan.py:255: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  attn = attn.repeat_interleave(repeat_time, dim=-1) if attn.shape[-1] > 1 else attn
========== Diagnostic Run torch.onnx.export version 1.14.0a0+44dac51 ===========
verbose: False, log level: Level.ERROR
======================= 0 NONE 0 NOTE 0 WARNING 0 ERROR ========================

2023-09-04 05:34:29,824 - mmcls - INFO - Successfully exported ONNX model: /results/classification_experiment/export/classification_model_export.onnx
Execution status: PASS
2023-09-04 13:34:44,732 [TAO Toolkit] [INFO] nvidia_tao_cli.components.docker_handler.docker_handler 337: Stopping container.
14

Did you ever run the notebook? i.e., https://github.com/NVIDIA/tao_tutorials/blob/main/notebooks/tao_launcher_starter_kit/classification_pyt/classification.ipynb, please double check if you can run inference correctly with pth file and onnx file.

15

yeah, I can confirm that inference is possible on the notebook, here is my inference log and results.

2023-08-31 15:49:51,432 [TAO Toolkit] [INFO] root 160: Registry: ['nvcr.io']
2023-08-31 15:49:51,470 [TAO Toolkit] [INFO] nvidia_tao_cli.components.instance_handler.local_instance 360: Running command in container: nvcr.io/nvidia/tao/tao-toolkit:5.0.0-pyt
2023-08-31 15:49:51,485 [TAO Toolkit] [WARNING] nvidia_tao_cli.components.docker_handler.docker_handler 262: 
Docker will run the commands as root. If you would like to retain your
local host permissions, please add the "user":"UID:GID" in the
DockerOptions portion of the "/home/lab/.tao_mounts.json" file. You can obtain your
users UID and GID by using the "id -u" and "id -g" commands on the
terminal.
2023-08-31 15:49:51,485 [TAO Toolkit] [INFO] nvidia_tao_cli.components.docker_handler.docker_handler 275: Printing tty value True
/usr/local/lib/python3.8/dist-packages/mmcv/__init__.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details.
  warnings.warn(
INFO: Created a temporary directory at /tmp/tmp8l1wbqmb
INFO: Writing /tmp/tmp8l1wbqmb/_remote_module_non_scriptable.py
/usr/local/lib/python3.8/dist-packages/mmcv/__init__.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details.
  warnings.warn(
INFO: Created a temporary directory at /tmp/tmpwn0rgiwk
INFO: Writing /tmp/tmpwn0rgiwk/_remote_module_non_scriptable.py
sys:1: UserWarning: 
'test_cats_dogs.yaml' is validated against ConfigStore schema with the same name.
This behavior is deprecated in Hydra 1.1 and will be removed in Hydra 1.2.
See https://hydra.cc/docs/next/upgrades/1.0_to_1.1/automatic_schema_matching for migration instructions.
<frozen core.hydra.hydra_runner>:107: UserWarning: 
'test_cats_dogs.yaml' is validated against ConfigStore schema with the same name.
This behavior is deprecated in Hydra 1.1 and will be removed in Hydra 1.2.
See https://hydra.cc/docs/next/upgrades/1.0_to_1.1/automatic_schema_matching for migration instructions.
/usr/local/lib/python3.8/dist-packages/hydra/_internal/hydra.py:119: UserWarning: Future Hydra versions will no longer change working directory at job runtime by default.
See https://hydra.cc/docs/next/upgrades/1.1_to_1.2/changes_to_job_working_dir/ for more information.
  ret = run_job(
<frozen core.loggers.api_logging>:245: UserWarning: Log file already exists at /results/classification_experiment/inference/status.json
Starting Classification inference
2023-08-31 07:49:59,867 - mmcls - INFO - **********************Start logging for Evaluation**********************
fatal: not a git repository: /opt/nvidia/tools/../.git/modules/tao-converter
load checkpoint from local path: /tmp/tmp9q2unozw.pth
[                                                  ] 0/1601, elapsed: 0s, ETA:/usr/local/lib/python3.8/dist-packages/mmcv/__init__.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details.
  warnings.warn(
INFO: Created a temporary directory at /tmp/tmpsm57p1sj
INFO: Writing /tmp/tmpsm57p1sj/_remote_module_non_scriptable.py
/usr/local/lib/python3.8/dist-packages/mmcv/__init__.py:20: UserWarning: On January 1, 2023, MMCV will release v2.0.0, in which it will remove components related to the training process and add a data transformation module. In addition, it will rename the package names mmcv to mmcv-lite and mmcv-full to mmcv. See https://github.com/open-mmlab/mmcv/blob/master/docs/en/compatibility.md for more details.
  warnings.warn(
INFO: Created a temporary directory at /tmp/tmp8r9f6lx4
INFO: Writing /tmp/tmp8r9f6lx4/_remote_module_non_scriptable.py
[>>>>>>>>>>>>>>>>>>>>>>>>>>>] 1601/1601, 59.9 task/s, elapsed: 27s, ETA:     0s2023-08-31 07:50:28,077 - mmcls - INFO - The inference result is saved at: /results/classification_experiment/inference/result.csv
Execution status: PASS
2023-08-31 15:52:56,811 [TAO Toolkit] [INFO] nvidia_tao_cli.components.docker_handler.docker_handler 337: Stopping container.

Screenshot from 2023-09-05 10-09-07
Screenshot from 2023-09-05 10-09-071247×612 183 KB

16

Please go through the last section “Deploy” to check if the onnx->tensorrt engine and the inference with this engine works.

17

It can be converted to an engine file and used for normal inference.

18

OK, to narrow down, please go back to
nvcr.io/nvidia/deepstream:6.2-triton , then follow TRTEXEC with Classification TF1/TF2/PyT - NVIDIA Docs to generate a tensorrt engine.

/usr/src/tensorrt/bin/trtexec --onnx=/path/to/model.onnx \
       --maxShapes=input_1:1x3x224x224 \
       --minShapes=input_1:1x3x224x224 \
       --optShapes=input_1:1x3x224x224 \
       --fp32 \
       --saveEngine=/path/to/save/trt/model.engine

Then config this model.engine correctly in the config config_as_primary_gie.txt.
model-engine-file=/path/to/model.engine

Comment out onnx file, and let deepstream run the engine directly.

19

Sorry, I got this error when I converted

root@9a392fcbd0cb:/model# /usr/src/tensorrt/bin/trtexec --onnx=/model/classification_model_export.onnx \
>        --maxShapes=input_1:1x3x224x224 \
>        --minShapes=input_1:1x3x224x224 \
>        --optShapes=input_1:1x3x224x224 \
>        --fp32 \
>        --saveEngine=/model/classification_model_export.engine

&&&& RUNNING TensorRT.trtexec [TensorRT v8502] # /usr/src/tensorrt/bin/trtexec --onnx=/model/classification_model_export.onnx --maxShapes=input_1:1x3x224x224 --minShapes=input_1:1x3x224x224 --optShapes=input_1:1x3x224x224 --fp32 --saveEngine=/model/classification_model_export.engine

[09/05/2023-02:41:09] [E] Unknown option: --fp32

20

Please use

/usr/src/tensorrt/bin/trtexec --onnx=/path/to/model.onnx \
       --maxShapes=input_1:1x3x224x224 \
       --minShapes=input_1:1x3x224x224 \
       --optShapes=input_1:1x3x224x224 \
       --saveEngine=/path/to/save/trt/model.engine

The engine will be fp32 mode.
Also, please set
network-mode=0

21

This strange question arises again

root@9a392fcbd0cb:/model# /usr/src/tensorrt/bin/trtexec --onnx=/model/classification_model_export.onnx        --maxShapes=input_1:1x3x224x224        --minShapes=input_1:1x3x224x224        --optShapes=input_1:1x3x224x224              --saveEngine=/model/classification_model_export.engine
&&&& RUNNING TensorRT.trtexec [TensorRT v8502] # /usr/src/tensorrt/bin/trtexec --onnx=/model/classification_model_export.onnx --maxShapes=input_1:1x3x224x224 --minShapes=input_1:1x3x224x224 --optShapes=input_1:1x3x224x224 --saveEngine=/model/classification_model_export.engine
[09/05/2023-02:50:23] [I] === Model Options ===
[09/05/2023-02:50:23] [I] Format: ONNX
[09/05/2023-02:50:23] [I] Model: /model/classification_model_export.onnx
[09/05/2023-02:50:23] [I] Output:
[09/05/2023-02:50:23] [I] === Build Options ===
[09/05/2023-02:50:23] [I] Max batch: explicit batch
[09/05/2023-02:50:23] [I] Memory Pools: workspace: default, dlaSRAM: default, dlaLocalDRAM: default, dlaGlobalDRAM: default
[09/05/2023-02:50:23] [I] minTiming: 1
[09/05/2023-02:50:23] [I] avgTiming: 8
[09/05/2023-02:50:23] [I] Precision: FP32
[09/05/2023-02:50:23] [I] LayerPrecisions: 
[09/05/2023-02:50:23] [I] Calibration: 
[09/05/2023-02:50:23] [I] Refit: Disabled
[09/05/2023-02:50:23] [I] Sparsity: Disabled
[09/05/2023-02:50:23] [I] Safe mode: Disabled
[09/05/2023-02:50:23] [I] DirectIO mode: Disabled
[09/05/2023-02:50:23] [I] Restricted mode: Disabled
[09/05/2023-02:50:23] [I] Build only: Disabled
[09/05/2023-02:50:23] [I] Save engine: /model/classification_model_export.engine
[09/05/2023-02:50:23] [I] Load engine: 
[09/05/2023-02:50:23] [I] Profiling verbosity: 0
[09/05/2023-02:50:23] [I] Tactic sources: Using default tactic sources
[09/05/2023-02:50:23] [I] timingCacheMode: local
[09/05/2023-02:50:23] [I] timingCacheFile: 
[09/05/2023-02:50:23] [I] Heuristic: Disabled
[09/05/2023-02:50:23] [I] Preview Features: Use default preview flags.
[09/05/2023-02:50:23] [I] Input(s)s format: fp32:CHW
[09/05/2023-02:50:23] [I] Output(s)s format: fp32:CHW
[09/05/2023-02:50:23] [I] Input build shape: input_1=1x3x224x224+1x3x224x224+1x3x224x224
[09/05/2023-02:50:23] [I] Input calibration shapes: model
[09/05/2023-02:50:23] [I] === System Options ===
[09/05/2023-02:50:23] [I] Device: 0
[09/05/2023-02:50:23] [I] DLACore: 
[09/05/2023-02:50:23] [I] Plugins:
[09/05/2023-02:50:23] [I] === Inference Options ===
[09/05/2023-02:50:23] [I] Batch: Explicit
[09/05/2023-02:50:23] [I] Input inference shape: input_1=1x3x224x224
[09/05/2023-02:50:23] [I] Iterations: 10
[09/05/2023-02:50:23] [I] Duration: 3s (+ 200ms warm up)
[09/05/2023-02:50:23] [I] Sleep time: 0ms
[09/05/2023-02:50:23] [I] Idle time: 0ms
[09/05/2023-02:50:23] [I] Streams: 1
[09/05/2023-02:50:23] [I] ExposeDMA: Disabled
[09/05/2023-02:50:23] [I] Data transfers: Enabled
[09/05/2023-02:50:23] [I] Spin-wait: Disabled
[09/05/2023-02:50:23] [I] Multithreading: Disabled
[09/05/2023-02:50:23] [I] CUDA Graph: Disabled
[09/05/2023-02:50:23] [I] Separate profiling: Disabled
[09/05/2023-02:50:23] [I] Time Deserialize: Disabled
[09/05/2023-02:50:23] [I] Time Refit: Disabled
[09/05/2023-02:50:23] [I] NVTX verbosity: 0
[09/05/2023-02:50:23] [I] Persistent Cache Ratio: 0
[09/05/2023-02:50:23] [I] Inputs:
[09/05/2023-02:50:23] [I] === Reporting Options ===
[09/05/2023-02:50:23] [I] Verbose: Disabled
[09/05/2023-02:50:23] [I] Averages: 10 inferences
[09/05/2023-02:50:23] [I] Percentiles: 90,95,99
[09/05/2023-02:50:23] [I] Dump refittable layers:Disabled
[09/05/2023-02:50:23] [I] Dump output: Disabled
[09/05/2023-02:50:23] [I] Profile: Disabled
[09/05/2023-02:50:23] [I] Export timing to JSON file: 
[09/05/2023-02:50:23] [I] Export output to JSON file: 
[09/05/2023-02:50:23] [I] Export profile to JSON file: 
[09/05/2023-02:50:23] [I] 
[09/05/2023-02:50:23] [I] === Device Information ===
[09/05/2023-02:50:23] [I] Selected Device: NVIDIA RTX A6000
[09/05/2023-02:50:23] [I] Compute Capability: 8.6
[09/05/2023-02:50:23] [I] SMs: 84
[09/05/2023-02:50:23] [I] Compute Clock Rate: 1.8 GHz
[09/05/2023-02:50:23] [I] Device Global Memory: 48675 MiB
[09/05/2023-02:50:23] [I] Shared Memory per SM: 100 KiB
[09/05/2023-02:50:23] [I] Memory Bus Width: 384 bits (ECC disabled)
[09/05/2023-02:50:23] [I] Memory Clock Rate: 8.001 GHz
[09/05/2023-02:50:23] [I] 
[09/05/2023-02:50:23] [I] TensorRT version: 8.5.2
[09/05/2023-02:50:23] [I] [TRT] [MemUsageChange] Init CUDA: CPU +14, GPU +0, now: CPU 27, GPU 1105 (MiB)
[09/05/2023-02:50:25] [I] [TRT] [MemUsageChange] Init builder kernel library: CPU +547, GPU +118, now: CPU 628, GPU 1223 (MiB)
[09/05/2023-02:50:25] [I] Start parsing network model
[09/05/2023-02:50:25] [I] [TRT] ----------------------------------------------------------------
[09/05/2023-02:50:25] [I] [TRT] Input filename:   /model/classification_model_export.onnx
[09/05/2023-02:50:25] [I] [TRT] ONNX IR version:  0.0.7
[09/05/2023-02:50:25] [I] [TRT] Opset version:    12
[09/05/2023-02:50:25] [I] [TRT] Producer name:    pytorch
[09/05/2023-02:50:25] [I] [TRT] Producer version: 1.14.0
[09/05/2023-02:50:25] [I] [TRT] Domain:           
[09/05/2023-02:50:25] [I] [TRT] Model version:    0
[09/05/2023-02:50:25] [I] [TRT] Doc string:       
[09/05/2023-02:50:25] [I] [TRT] ----------------------------------------------------------------
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:377: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:25] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:30] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:31] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:31] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:31] [W] [TRT] onnx2trt_utils.cpp:403: One or more weights outside the range of INT32 was clamped
[09/05/2023-02:50:32] [I] Finish parsing network model
[09/05/2023-02:50:33] [I] [TRT] [MemUsageChange] Init cuBLAS/cuBLASLt: CPU +6, GPU +10, now: CPU 677, GPU 1233 (MiB)
[09/05/2023-02:50:34] [I] [TRT] [MemUsageChange] Init cuDNN: CPU +2, GPU +10, now: CPU 679, GPU 1243 (MiB)
[09/05/2023-02:50:34] [I] [TRT] Local timing cache in use. Profiling results in this builder pass will not be stored.
SSA validation FAIL: tensor __mye53 has multiple definitions
peephole.cpp:657: DCHECK(g_->ssa_validation()) failed. 
Aborted (core dumped)