Please provide complete information as applicable to your setup.
• Hardware Platform (Jetson Orin 64GB Developer)
**• DeepStream Version 6.4 ** • JetPack Version 6.0 • TensorRT Version 8.6.2.3 • NVIDIA GPU Driver Version (valid for GPU only) How do I find? • Issue Type( questions, bugs) • How to reproduce the issue ? (I ran $sudo deepstream-app -version-all)
Completed a Jetpack 6 Flash with Deepstream 6.4. Visually all appears ok. I conducted an install of
Graph Composer & Deepstream reference graphs. All directories seem present :
steven@ubuntu:/opt/nvidia$ ls -al
drwxr-xr-x 3 root root 4096 Jan 15 16:34 deepstream
drwxr-xr-x 5 root root 4096 Jan 19 16:52 graph-composer
steven@ubuntu:/opt/nvidia/deepstream$ ls
deepstream deepstream-6.4
steven@ubuntu:/opt/nvidia/deepstream/deepstream-6.4$ ls -al | grep ref*
drwxr-xr-x 21 root root 4096 Jan 19 17:01 reference_graphs
steven@ubuntu:~$ sudo deepstream-app -version-all
[sudo] password for steven: > deepstream-app: error while loading shared libraries: libgstrtspserver-1.0.so.0: cannot open shared object file: No such file or directory
#1 Don’t understand why I’m getting this error. #2 I don’t see the way to run the Composer GUI application.
Update 2024-01-21:
I was assuming the ORIN Jetson flash of Deepstream would include all needed libraries. Apparently this isn’t the case or I missed the instructions for this somewhere. I followed part of the instruction in the Quickstart Guide used for manually installing Deep Stream.
I followed
Install Dependencies section
Install librdkafka (to enable Kafka protocol adaptor for message broker)
Except, I did not create the “lib” directory in Deepstream-6.4. It was already
created during the flash.
After this the --version-all option worked correctly.
I still do not know how to run the Graph Composer. Anyone?
Sync GXF extensions from the server registry repo sync -n ngc-public
Run gxf server in one terminal: gxf_server
Run the sample graph in another terminal cd /opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test1 /opt/nvidia/graph-composer/execute_graph.sh deepstream-test1.yaml parameters.yaml -d ../common/target_aarch64.yaml
steven@development-orin:~$ gxf_server
INFO:root:Environment variable ‘GXF_SERVER_PORT’ is not set. Using default port: 50051.
INFO:root:SERVER started: Server is available at [::]:50051…
Step 6 Terminal 2:
steven@development-orin:~$ cd /opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test1
steven@development-orin:/opt/nvidia/deepstream/deepstream/reference_graphs/deepstream-test1$ /opt/nvidia/graph-composer/execute_graph.sh deepstream-test1.yaml parameters.yaml -d ../common/target_aarch64.yaml
Graphs: deepstream-test1.yaml,parameters.yaml
Target: ../common/target_aarch64.yaml
===================================================================
Running deepstream-test1.yaml
===================================================================
...
2024-01-22 15:13:57.016 ERROR extensions/nvdssource/nvurisrcbin.hpp@36: Could not create GStreamer element 'nvurisrcbin'
2024-01-22 15:13:57.016 ERROR gxf/std/program.cpp@519: Couldn't run async. Deactivating...
2024-01-22 15:13:57.016 ERROR gxf/core/runtime.cpp@1392: Graph run failed with error: GXF_FAILURE
2024-01-22 15:13:57.016 ERROR gxf/gxe/gxe.cpp@269: GxfGraphRunAsync Error: GXF_FAILURE
*******************************************************************
End deepstream-test1.yaml
*******************************************************************
[INFO] Graph installation directory /tmp/ds.deepstream-test1 and manifest /tmp/ds.deepstream-test1/manifest.yaml retained
On my Dell laptop x86_64 with Ubuntu 22.04
~$ sudo dpkg -i graph_composer-3.1.0_x86_64.deb
…
dpkg: dependency problems prevent configuration of graph-composer:
graph-composer depends on libyaml-cpp-dev
so I
~$ sudo apt install libyaml-cpp-dev
~$ sudo apt --fix-broken install
~$ sudo dpkg -i graph_composer-3.1.0_x86_64.deb
Completed successfully
~$ composer
Message asked me to disable IOMMU
~$ cd /etc/default
~$ vim grub
GRUB_CMDLINE_LINUX=“intel_iommu=off”
~$ update-grub
I ran composer again, the issues above were resolved, However, now I have several new errors. I don’t know if they are important. The Composer did open in a window.
~$ compose
…
[Info] [carb] Logging to file: /home/steven/.nvidia-omniverse/logs/Kit/composer-standalone/3.1/kit_20240126_151544.log
2024-01-26 23:15:44 [1ms] [Warning] [omni.ext.plugin] [ext: omni_kit] Extensions config 'extension.toml' doesn't exist '/tmp/.tmpNvXgRaphComposer/run/composer/omni_kit' or '/tmp/.tmpNvXgRaphComposer/run/composer/omni_kit/config'
steven@DEVELOPMENT-DELL:~$ [0.380s] [ext: omni.kit.async_engine-0.0.0] startup
2024-01-26 23:15:45 [435ms] [Warning] [gpu.foundation.plugin] nvidia-smi not found in '/usr/bin/'
|---------------------------------------------------------------------------------------------|
2024-01-26 23:15:45 [520ms] [Warning] [carb.cudainterop.plugin] NVML could not be loaded. Error: libnvidia-ml.so.1: cannot open shared object file: No such file or directory
2024-01-26 23:15:45 [520ms] [Warning] [gpu.foundation.plugin] ECC check skipped: Can't query ECC mode for device 0.
[0.550s] [ext: omni.kit.pipapi-0.0.0] startup
[1.101s] [ext: omni.usd_resolver-1.0.0] startup
2024-01-26 23:15:45 [1,084ms] [Warning] [omni.ext.plugin] [ext: omni.usd_resolver-1.0.0] Native plugin search dir contains several files: /tmp/.tmpNvXgRaphComposer/run/kit/exts/omni.usd_resolver/bin . It makes dir walk slow, replace with the exact path to the plugin instead of a wildcard.
[1.105s] [ext: omni.kit.widget.nucleus_connector-1.1.0] startup
[1.602s] [ext: omni.exp.graph.composer-3.1.0] startup
2024-01-26 23:15:46 [1,587ms] [Warning] [omni.kit.app._impl] carb.imgui is deprecated. Use omni.kit.imgui instead.
[1.776s] [ext: composer-standalone-3.1.0] startup
[1.819s] app ready
2024-01-26 23:15:46 [2,073ms] [Error] [omni.exp.graph.composer.registry_window] Failed to get repo list: <AioRpcError of RPC that terminated with:
status = StatusCode.UNAVAILABLE
details = "failed to connect to all addresses; last error: UNKNOWN: ipv4:127.0.0.1:50051: Failed to connect to remote host: Connection refused"
debug_error_string = "UNKNOWN:failed to connect to all addresses; last error: UNKNOWN: ipv4:127.0.0.1:50051: Failed to connect to remote host: Connection refused {created_time:"2024-01-26T15:15:46.772138-08:00", grpc_status:14}"
>
2024-01-26 23:15:46 [2,073ms] [Error] [omni.exp.graph.composer.node_registry_model] Registry Model update failed
steven@DEVELOPMENT-DELL:~$ 2024-01-27 00:11:48 [3,363,981ms] [Warning] [omni.kit.window.filepicker.api] No item exists with url '/opt/nvidia/deepstream/deepstream/reference_graphs'.
2024-01-27 00:11:48 [3,363,998ms] [Warning] [omni.kit.window.filepicker.api] No item exists with url '/opt/nvidia/deepstream/deepstream'.
2024-01-27 00:11:48 [3,364,083ms] [Warning] [omni.kit.window.filepicker.api] No item exists with url '/opt/nvidia/deepstream/deepstream'.
2024-01-27 00:11:48 [3,364,100ms] [Warning] [omni.kit.window.filepicker.api] No item exists with url '/opt/nvidia/deepstream'.
steven@DEVELOPMENT-DELL:~$
One last question:
I reviewed the 7 requirements for the Graph Composer as you pointed out. I can do all except I can’t
see how to install DeepStream SDK 6.4 for DeepStream extensions. All of the installation procedures are for dGPU or Jetson. I have a Dell XPS with an integrated iGPU (Intel(R) Iris(R) Xe Graphics). Can you point me to the install procedures in this case, or is this not needed?
There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks
iGPU device can only be used to run composer GUI to construct graphs. No DeepStream is needed for the case.