Please provide complete information as applicable to your setup.
• DGX GPU A100
• DeepStream 8.0
I have a question, in DS version 8.0 the max-batch-size property of the multiurisrcbin must be set from the beginning when starting the pipeline, my question is if that parameter can be changed in real time when the pipeline is PLAYING? I’m working with Python and its bindings with GStreamer, DS, etc. I have that question, for example the streamdemux in this version 8.0 has dynamic source pads.
How do you use nvmultiurisrcbin? in deepstream-server or custom pipeline? are you using legacy nvsteammux or new streammux? nvmultiurisrcbin is opensource. As gst_nvmultiurisrcbincreator_init shown, the “max-batch-size” will be set for legacy nvstreammux.
What do you mean by “streamdemux in this version 8.0 has dynamic source pads.”? is there any bug?
I’m using the nvmultiurisrcbin in a custom pipeline instead of nvurisrcbin+streammux, I’m using the legacy (original/old) streammux.
My main question is:
I used the streamdemux example specifically for the dynamic nature of what I’m talking about. Previously, you had to set the src pads when starting the pipeline. Now, they’re on demand. When adding or removing cameras in real time, I can add or remove src pads from the demuxer (only works in DS 8.0).
After adding or removing cameras, you don’t need to modify “max-batch-size” property accordingly, which will be set for nvstreammux. it is just a max batch size, not the actual batch size. you may set this value when starting the pipeline. Please refer to the sample deepstream-server.
I see, the problem occurs when I start a dynamic pipeline with just 1 dummy source and a max batch size of 32 for example, the pipeline crashes
Sorry for the late reply, Is this still an DeepStream issue to support? Thanks! If the crash issue persists, can you use native deepstream-server to reproduce the issue? if so, could you provide the simplified code, cfgs and test video? Thanks!
I’m just using the multiurisrcbin inside a pipeline developed with python bindings, I can share you the image of all the pipeline, at the beginning it starts with 1 dummy source. Don’t worry about the 3 elements in the image that are not linked to the pipeline:
Could you elborate on “dynamic pipeline”? could you use gdb to get a crash stack? could you could you simplified pipeline and share a simplifid code/method to reproduce this issue? Thanks!
I can share you some configs of the plugins, starting with the dummy and a max batch size of 10 it works
Ok. To help reprodcue, Please share the related files. You can use forum private email. please click forum avatar → "personal messages’ → “view all personal messages” → “new message”.
Sorry for the late reply, Is this still an DeepStream issue to support? Thanks!
