Hi @Amycao,
The config_tri_res.txt for the config of nvinferserver is missing. Sorry, I forgot to add that in the above reply. Will edit the reply. Let me know if that was the only issue.
Thanks.
Issue still there, after copy config_tri_res.txt to deepstream-resnet101_/
root@57dac851f719:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ./deepstream-test1-app file:///opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.h264
(deepstream-test1-app:3491): GLib-GObject-CRITICAL **: 06:21:23.654: g_object_set: assertion ‘G_IS_OBJECT (object)’ failed
(deepstream-test1-app:3491): GStreamer-CRITICAL **: 06:21:23.654: gst_bin_add_many: assertion ‘GST_IS_ELEMENT (element_1)’ failed
creating uridecodebin for [file:///opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.h264]
(deepstream-test1-app:3491): GStreamer-CRITICAL **: 06:21:23.655: gst_element_link_many: assertion ‘GST_IS_ELEMENT (element_2)’ failed
Elements could not be linked: 2. Exiting.
root@57dac851f719:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls
Makefile config_tri_res.txt deepstream-test1-app deepstream_test_resnet101.c deepstream_test_resnet101.o resnet101_
Hi duttaneil16,
Is this still an issue to support? Any update?
Hi kaycc,
It is an issue but not able to understand why ur setup is not working. I will share the files again later by the day after checking the setup which is working on my system.
Hi,
root@3586c4a4b128:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls
Makefile config_tri_res.txt deepstream-test1-app deepstream_test_resnet101.c deepstream_test_resnet101.o resnet101_
root@3586c4a4b128:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls resnet101_/
1 config.pbtxt
root@3586c4a4b128:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls resnet101_/1/
model.graphdef
Here are the files
config_tri_res.txt (809 Bytes) deepstream_test_resnet101.c (14.6 KB) Makefile (2.1 KB) config.pbtxt (1.2 KB)
root@3586c4a4b128:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# make
root@3586c4a4b128:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ./deepstream-test1-app /opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.h264
The hardware I run it on is RTX2080. Please send the error details if it does not recreate the same issue or throws different error.
Hi, I have same output for the directory as you, see below, but i still got same failure.
Can you send the log you run successfully?
root@b5c3d768b2bc:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls
Makefile config_tri_res.txt deepstream-test1-app deepstream_test_resnet101.c deepstream_test_resnet101.o resnet101_
root@b5c3d768b2bc:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls resnet101_/
1 config.pbtxt
root@b5c3d768b2bc:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ls resnet101_/1
model.graphdef
root@b5c3d768b2bc:/opt/nvidia/deepstream/deepstream-5.0/sources/apps/sample_apps/deepstream-resnet101_# ./deepstream-test1-app /opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.h264
(deepstream-test1-app:13): GLib-GObject-CRITICAL **: 03:03:58.064: g_object_set: assertion ‘G_IS_OBJECT (object)’ failed
(deepstream-test1-app:13): GStreamer-CRITICAL **: 03:03:58.065: gst_bin_add_many: assertion ‘GST_IS_ELEMENT (element_1)’ failed
creating uridecodebin for [/opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.h264]
(deepstream-test1-app:13): GStreamer-CRITICAL **: 03:03:58.066: gst_element_link_many: assertion ‘GST_IS_ELEMENT (element_2)’ failed
Elements could not be linked: 2. Exiting.
Hi @Amycao,
Command will be:
./deepstream-test1-app file:///opt/nvidia/deepstream/deepstream-5.0/samples/streams/sample_720p.h264
sorry for that typo.
Although that is not the issue. I am adding a typescript which will tell you the exact output I get in terminal.
typescript (11.1 KB)
You can check this for the result and error I am facing.
You use nvinferserver in the pipeline, right? if i change to nvinfer, the error before met will gone, although another error raise, will ignore this, and focus on nvinferserver, if you using inferserver.
I am using tf graph so cant use nvinfer. nvinfer uses trt and I have some unsupported layers cant be changed to trt. Thats why I am using nvinferserver. so want to resolve in nvinferserver
You issue is below, right?
ERROR: infer_trtis_server.cpp:276 TRTIS: failed to get response status, trtis_err_str:INTERNAL, err_msg:2 root error(s) found.
(0) Failed precondition: Attempting to use uninitialized value stage4_unit1_bn1_var
[[{{node stage4_unit1_bn1_var/read}}]]
[[softmax/_5]]
(1) Failed precondition: Attempting to use uninitialized value stage4_unit1_bn1_var
[[{{node stage4_unit1_bn1_var/read}}]]
0 successful operations.
0 derived errors ignored.
Hi amycao,
Yes, this is the issue I face while running the model. Seems to be some initialization issue, as I can run the same model in python. Not sure if any changes to be done from my end or not.
Hi
Sorry for the delay.
Is your model LSTM? if yes, please refer to deepstream manual gst-nvinferserver to enable it. currently lstm only support batch-1. if not lstm, run tritonserver and perf_client, to see if tritonserver server this model. also you can try with convert to savedmodel.
Hi
Resolved the issue. the issue was converting the mxnet to TensorFlow graphdef.
The converted model was running fine using python TensorFlow API with an extra set of lines but was an issue here.
using the functions → tf.compat.v1.graph_util.remove_training_nodes, tf.graph_util.convert_variables_to_constants did the trick. We can now close the topic.
Thanks.