Graph Composer Registry Incompatible gxf core version

• Hardware Platform (Jetson / GPU) RTX 3060 Laptop
• DeepStream Version 6.3
• TensorRT Version 8.5.3.1
• NVIDIA GPU Driver Version (valid for GPU only) nvidia-driver-535

So I want to create a pipeline for emotion detection with Graph Composer. I’m using Ubuntu 20.04 x86_64. I followed all the steps to download and install Graph Composer 3.0.0 and GXF 3.0.0 as available in Ubuntu 20.04 LTS (x86_64) — DeepStream 6.3 Release documentation. Also followed the step by step to install DeepStream 6.3 (requirement for the Graph Composer) available in Quickstart Guide — DeepStream 6.3 Release documentation.

Everything seemed to work, I am able to execute the Graph Composer app and run the gxf_server on the terminal. The issue I am having is that when i sync the Graph Composer registry with the “ngc-public” registry, in the gxf_server terminal I can see the following:

DEBUG:Registry:Skipping extension NvDsInferenceExt version 1.1.0
DEBUG:Registry:Incompatible gxf core version
DEBUG:Registry:Registry gxf core version: 3.0.0 Extension gxf core version: 2.5.0
DEBUG:Registry:Skipping extension NvDsBaseExt version 1.2.0
DEBUG:Registry:Incompatible gxf core version
DEBUG:Registry:Registry gxf core version: 3.0.0 Extension gxf core version: 2.5.0

Those lines are just an example of some of the extensions that couldn’t be synched due to incompatible gxf version, there are a lot more with the same issue. To be clear not all extensions failed to sync, but a lot have and I need them for my AI application.

Please help me find a way fix this. The extensions have different gxf core versions, like 2.5.0 and 2.4.2 and mine is 3.0.0.

The registry sync will sync all extensions including historical extensions for previous DeepStream versions. It is OK to ignore such error.

My goal is to be able to run the emotion graph available at https://github.com/NVIDIA-AI-IOT/deepstream_tao_apps/tree/master/graphs/tao_others/emotion. I followed all the steps and I am able to run the faciallandmarks graph without any trouble. I understand your comment, but for what I see in the terminal, all the versions for the NvDsEmotionExt were for previous versions of DeepStream (<6.3) and are being skipped. The extension for the Emotion Net does not show on the Graph Composer interface and when I try to run the graph I get errors from the part that should use it in my understanding.

/opt/nvidia/graph-composer/execute_graph.sh App-Emotion.yaml -d /opt/nvidia/deepstream/deepstream/reference_graphs/common/target_x86_64.yaml

Graphs: App-Emotion.yaml
Target: /opt/nvidia/deepstream/deepstream/reference_graphs/common/target_x86_64.yaml

Running App-Emotion.yaml

[INFO] Writing manifest to /tmp/ds.App-Emotion/manifest.yaml
2023-10-23 23:25:39.842 INFO gxf/gxe/gxe.cpp@182: Creating context
2023-10-23 23:25:39.986 INFO gxf/gxe/gxe.cpp@107: Loading app: ‘/opt/nvidia/deepstream/deepstream-6.3/deepstream_tao_apps/graphs/tao_others/emotion/App-Emotion.yaml’
2023-10-23 23:25:39.986 INFO gxf/std/yaml_file_loader.cpp@170: Loading GXF entities from YAML file ‘/opt/nvidia/deepstream/deepstream-6.3/deepstream_tao_apps/graphs/tao_others/emotion/App-Emotion.yaml’…
2023-10-23 23:25:39.990 INFO gxf/gxe/gxe.cpp@259: Initializing…
2023-10-23 23:25:39.998 INFO extensions/nvdsbase/nvds_scheduler.cpp@270: This program is linked against GStreamer 1.16.3

2023-10-23 23:25:39.999 INFO extensions/nvdsmuxdemux/nvstreammux.hpp@27: initialize: nvstreammux nv_ds_stream_mux18…9cf8

2023-10-23 23:25:40.000 INFO extensions/nvdsinference/nvinferbin.hpp@24: initialize: nvinferbin nv_ds_infer22…9780

2023-10-23 23:25:40.000 INFO extensions/nvdsvisualization/nvosdbin.hpp@24: initialize: nvosdbin nv_ds_osd27…8f60

2023-10-23 23:25:40.000 INFO extensions/nvdsinference/nvinferbin.hpp@24: initialize: nvinferbin nv_ds_infer_video2…fe48

2023-10-23 23:25:40.000 INFO extensions/nvdstemplate/nvdsvideotemplate.hpp@23: initialize: nvdsvideotemplate nv_ds_video_template0…8550

2023-10-23 23:25:40.000 INFO extensions/nvdsvisualization/nvtilerbin.hpp@37: initialize: nvtilerbin nv_ds_tiler4…7630

2023-10-23 23:25:40.001 INFO extensions/nvdsoutputsink/nvvideorenderersinkbin.hpp@24: initialize: nvvideorenderersinkbin NVidia Video Renderer10

2023-10-23 23:25:40.001 INFO extensions/nvdssource/nvcamerasrcbin.hpp@23: initialize: nvcamerasrcbin Camera Source Input13

2023-10-23 23:25:40.001 INFO gxf/gxe/gxe.cpp@266: Running…
2023-10-23 23:25:40.001 INFO extensions/nvdsbase/nvds_scheduler.cpp@121: Scheduling 8 elements and 3 components
2023-10-23 23:25:40.001 INFO extensions/nvdsmuxdemux/nvstreammux.hpp@37: create_element: nvstreammux nv_ds_stream_mux18…9cf8

2023-10-23 23:25:40.003 INFO extensions/nvdsmuxdemux/nvstreammux.hpp@61: bin_add: nvstreammux nv_ds_stream_mux18…9cf8

2023-10-23 23:25:40.003 INFO extensions/nvdsinference/nvinferbin.hpp@32: create_element: nvinferbin nv_ds_infer22…9780

2023-10-23 23:25:40.010 INFO extensions/nvdsinference/nvinferbin.hpp@56: bin_add: nvinferbin nv_ds_infer22…9780

2023-10-23 23:25:40.010 INFO extensions/nvdsvisualization/nvosdbin.hpp@32: create_element: nvosdbin nv_ds_osd27…8f60

2023-10-23 23:25:40.010 INFO extensions/nvdsvisualization/nvosdbin.hpp@56: bin_add: nvosdbin nv_ds_osd27…8f60

2023-10-23 23:25:40.010 INFO extensions/nvdsinference/nvinferbin.hpp@32: create_element: nvinferbin nv_ds_infer_video2…fe48

2023-10-23 23:25:40.011 INFO extensions/nvdsinference/nvinferbin.hpp@56: bin_add: nvinferbin nv_ds_infer_video2…fe48

2023-10-23 23:25:40.011 INFO extensions/nvdstemplate/nvdsvideotemplate.hpp@31: create_element: nvdsvideotemplate nv_ds_video_template0…8550

2023-10-23 23:25:40.011 INFO extensions/nvdstemplate/nvdsvideotemplate.hpp@55: bin_add: nvdsvideotemplate nv_ds_video_template0…8550

2023-10-23 23:25:40.011 INFO extensions/nvdsvisualization/nvtilerbin.hpp@45: create_element: nvtilerbin nv_ds_tiler4…7630

2023-10-23 23:25:40.011 INFO extensions/nvdsvisualization/nvtilerbin.hpp@69: bin_add: nvtilerbin nv_ds_tiler4…7630

2023-10-23 23:25:40.011 INFO extensions/nvdsoutputsink/nvvideorenderersinkbin.hpp@32: create_element: nvvideorenderersinkbin NVidia Video Renderer10

2023-10-23 23:25:40.011 INFO extensions/nvdsoutputsink/nvvideorenderersinkbin.hpp@54: bin_add: nvvideorenderersinkbin NVidia Video Renderer10

2023-10-23 23:25:40.011 INFO extensions/nvdssource/nvcamerasrcbin.hpp@31: create_element: nvcamerasrcbin Camera Source Input13

2023-10-23 23:25:40.011 INFO extensions/nvdssource/nvcamerasrcbin.hpp@49: bin_add: nvcamerasrcbin Camera Source Input13

2023-10-23 23:25:40.248 ERROR extensions/nvdsbase/nvds_scheduler.cpp@312: Error from /GstPipeline:NvDsScheduler-Pipeline/GstNvDsVideoTemplate:NvDsVideoTemplate…75f8/nv_ds_video_template0…8550: Exception occurred
2023-10-23 23:25:40.248 ERROR extensions/nvdsbase/nvds_scheduler.cpp@314: Debug info:
Exception occurred
2023-10-23 23:25:40.249 ERROR extensions/nvdsbase/nvds_scheduler.cpp@184: Failed to set GStreamer pipeline to PLAYING
Returned, stopping playback
Deleting pipeline
2023-10-23 23:25:40.256 INFO gxf/gxe/gxe.cpp@288: Deinitializing…
2023-10-23 23:25:40.256 INFO gxf/gxe/gxe.cpp@296: Destroying context
2023-10-23 23:25:40.257 INFO gxf/gxe/gxe.cpp@302: Context destroyed.


End App-Emotion.yaml


As described in https://github.com/NVIDIA-AI-IOT/deepstream_tao_apps/blob/master/graphs/tao_others/emotion/App-Emotion.yaml is the nv_ds_video_template0…8550 that is configured with NvDsEmotionTemplateLib that is a NvDsEmotionExt component according to NvDsEmotionExt — DeepStream 6.3 Release documentation. As seen in the debug log, this is the component that is getting the error.

Worth mentioning that I am now doing everything inside the deepstream:6.3-gc-triton-devel docker container to discard any possible mistakes by my part in the deepstream or graph composer instalation (the issue with the extensions is still the same). Available in DeepStream | NVIDIA NGC

Today I was able to run the emotion net pipeline with the proper extensions by using the deepstream:6.0.1-devel container. With the Graph Composer 1.0.0 I was able to sync ALL the extensions. Now just a final question. I’m using this container inside a x86_64 Ubuntu 20.04. My goal is to run the graph on a Jetson Nano with 4.6.1 JetPack and 6.0.1 DeepStream. In this case am I right to build the graph using this container that have DeepStream 6.0.1 or the DeepStream version for the container that I’m using the Graph Composer to build the graph doesn’t matter?

Yes. You can build graph in x86_64 DeepStream 6.0.1 and the graph can run in JetPack 4.6.1 with DeepStream 6.0.1.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.