WARNING: Logging before InitGoogleLogging() is written to STDERR
I20230922 23:42:07.597575 10 video_controller_client.cc:434] gPRC connection uri : 127.0.0.1:50060
I20230922 23:42:07.598493 10 client_carrier.cpp:39] InjectSpan
I20230922 23:42:08.281419 10 video_controller_streaming_client.cc:309] Creating sender pipe video_send_aux: filesrc name=filesrc_aux ! pngdec name=imagedec_aux ! nvvideoconvert name=nvvconv_aux ! video/x-raw(memory:NVMM),format=NV12,colorimetry=bt709 ! nvv4l2h264enc profile=Main bitrate=5242800 name=nvenc_aux ! queue ! h264parse name=h264parse_aux config-interval=-1 ! fakesink sync=true name=fakesink_grpc_aux async=false
E20230922 23:42:08.355175 10 nv_gst_pipeline.cc:144] Pipeline:video_send_aux unable to set to state PLAYING
E20230922 23:42:08.355237 10 video_controller_streaming_client.cc:412] Pipeline launch failed
(video_controller_client:10): GLib-CRITICAL **: 23:42:08.355: g_main_loop_unref: assertion 'loop != NULL' failed
I20230922 23:42:08.355574 10 nv_gst_pipeline.cc:558] Pipeline:video_send_aux deleted.
E20230922 23:42:08.355620 10 video_controller_client.cc:512] Failed to create video-send pipeline
I20230922 23:42:08.355698 10 video_controller_client.cc:290] Stopping client...
terminate called without an active exception
Aborted
My Ubuntu 22.04 can “see” my RTX 2080 Ti card when I type nvidia-smi on the commandline. Is the problem caused by running these scripts on Ubuntu under Windows Subsystem for Linux 2?
The tutorial itself advises that it can only be run on sm_86 and sm_75 architectures, which my original machine did not have.
An AWS EC2 g4dn.2xlarge instance running a Tesla T4 GPU solved the GLib-CRITICAL error problem. The free NVIDIA Deep Learning Base AMI on the AWS marketplace helped immensely in starting a Linux instance that could run this Live Portrait tutorial.
I’m running into a new error now, however, when I run this command in the client’s interactive shell: