How to set log severity in DeepStream 5

Accelerated Computing Intelligent Video Analytics DeepStream SDK
#1

Right now when I run the pipeline there’s a lot of INFO messages like:

Now playing:
gstnvtracker: Loading low-level lib at ../lib/libnvds_nvdcf.so
gstnvtracker: Optional NvMOT_RemoveStreams not implemented
gstnvtracker: Batch processing is ON
[NvDCF] Initialized
0:00:01.055599330 28854 0x55ba0f683780 INFO                 nvinfer gstnvinfer.cpp:602:gst_nvinfer_logger:<primary-nvinference-engine> NvDsInferContext[UID 1]: Info from NvDsInferContextImpl::buildModel() <nvdsinfer_context_impl.cpp:1591> [UID = 1]: Trying to create engine from model files

INFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm1_162: broadcasting input0 to make tensors conform, dims(input0)=[1,38,19][NONE] dims(input1)=[256,19,19][NONE].
INFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm2_162: broadcasting input1 to make tensors conform, dims(input0)=[256,38,19][NONE] dims(input1)=[1,19,38][NONE].

NFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm1_162: broadcasting input0 to make tensors conform, dims(input0)=[1,38,19][NONE] dims(input1)=[256,19,19][NONE].
INFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm2_162: broadcasting input1 to make tensors conform, dims(input0)=[256,38,19][NONE] dims(input1)=[1,19,38][NONE].
INFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm1_170: broadcasting input0 to make tensors conform, dims(input0)=[1,76,38][NONE] dims(input1)=[128,38,38][NONE].
INFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm2_170: broadcasting input1 to make tensors conform, dims(input0)=[128,76,38][NONE] dims(input1)=[1,38,76][NONE].
INFO: ../nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: Some tactics do not have sufficient workspace memory to run. Increasing workspace size may increase performance, please check verbose output.

And so on and so forth. Is there a way to filter this out? I’m currently using a custom .so file taking a reference from libnvdsinfer_custom_Yolo.so so in the NvDsInferYoloCudaEngineGet is there a way to set the ILogger severity level (since it’s not static I can’t set it outside the scope)? Furthermore, there’s also GStreamer INFO messages that I’d like to get rid. Is there a way to do that?

• Hardware Platform: NVIDIA T4
• DeepStream Version: 5.0
• TensorRT Version: 7
• NVIDIA GPU Driver Version 440

#3

set GST_DEBUG to an integer of your choice (1-9). See details of each level at that link.
edit: nevermind. didn’t read closely. Googling I don’t see a way to set that via an environment variable so you may have to modify nvinfer unless somebody from nvidia has another solution.

#4

And so on and so forth. Is there a way to filter this out? I’m currently using a custom .so file taking a reference from libnvdsinfer_custom_Yolo.so so in the NvDsInferYoloCudaEngineGet is there a way to set the ILogger severity level (since it’s not static I can’t set it outside the scope)?

-> you can not set ILogger severity level via environments, it’s from TensorRT Logging API.

Furthermore, there’s also GStreamer INFO messages that I’d like to get rid. Is there a way to do that?

-> NFO: …/nvdsinfer/nvdsinfer_func_utils.cpp:37 [TRT]: mm1_162: broadcasting input0 to make tensors conform, dims(input0)=[1,38,19][NONE] dims(input1)=[256,19,19][NONE].
you mean this INFO? or you missed pasted?

#5

So is there no way of suppressing the TensorRT log messages? I’d ideally want to get rid of all the info messages.

There are also a couple of nvinfer log messages. I’m unsure if those are generated by GStreamer through the gst-nvinfer plugin, or TensorRT through nvinfer.

1 Like
#6

You can not disable the TRT log.

#8