Tlt-converter on jetson nano is stuck

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi,
I’am trying to use tlt3.0.
My aim is to use it on my own dataset using the notebook yolov4.
To test it before I followed the notebook provide by nvidia yolov4 on the kitti dataset which recognized pedestrian, car and bicycle. On my machine all steps have been passed successfully. So I try to export my model on jetson (XavierNX and Nano). I installed the corresponding tlt-converter on each according to the jetpack I used. For the Xavier I obtained the .trt file (after 15-20 min ). Unfortunately for the Jetson Nano (4Go) the tlt converter start but even after 1 day I do not retrieve my .trt file.

Is there an another way to convert my etlt file outside the Nano but still using it on Nano ???

Sincerely,

3

Is there any log?

4

here are the log when I am running the tlt command.
thx

[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.
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[WARNING] onnx2trt_utils.cpp:246: One or more weights outside the range of INT32 was clamped
[INFO] ModelImporter.cpp:135: No importer registered for op: BatchedNMSDynamic_TRT. Attempting to import as plugin.
[INFO] builtin_op_importers.cpp:3659: Searching for plugin: BatchedNMSDynamic_TRT, plugin_version: 1, plugin_namespace: 
[INFO] builtin_op_importers.cpp:3676: Successfully created plugin: BatchedNMSDynamic_TRT
[INFO] Detected input dimensions from the model: (-1, 3, 384, 1248)
[INFO] Model has dynamic shape. Setting up optimization profiles.
[INFO] Using optimization profile min shape: (1, 3, 384, 1248) for input: Input
[INFO] Using optimization profile opt shape: (8, 3, 384, 1248) for input: Input
[INFO] Using optimization profile max shape: (16, 3, 384, 1248) for input: Input
[INFO] Some tactics do not have sufficient workspace memory to run. Increasing workspace size may increase performance, please check verbose output.
5

Can you share the command?

6

On the Jetson Xavier it works but not on my Jetson Nano

./tlt-converter my/path/for_tlt_converter/cuda10.2_trt7.1_jp4.4/yolov4_resnet18_epoch_080.etlt -k &KEY  -p Input,1x3x384x1248,8x3x384x1248,16x3x384x1248 -e /my/path/cuda10.2_trt7.1_jp4.4/trt.engine.trt -t fp16
7

Please try
./tlt-converter my/path/for_tlt_converter/cuda10.2_trt7.1_jp4.4/yolov4_resnet18_epoch_080.etlt -k &KEY -p Input,1x3x384x1248,1x3x384x1248,1x3x384x1248 -e /my/path/cuda10.2_trt7.1_jp4.4/trt.engine.trt -t fp16 -m 1 -w 100000000

8

thx it works now !!! Just for my information. Can you explain what was the issue ?

9

See ./tlt-converter -h for more info.
Just try to increase the workspace size and decrease the batch size. Also set lower optimization profile shapes.

10

Thanks a lot

12

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.