Excuse me, in the new deepstream, I added DLA support in the configuration file of Infer, but it hasn't been implemented yet

[940655439]

Please provide complete information as applicable to your setup.

**• Hardware Platform (Jetson / GPU)6.1
**• DeepStream Version 7.1
**• JetPack Version (valid for Jetson only)6.1
**• TensorRT Version 10.3
**• NVIDIA GPU Driver Version (valid for GPU only)540.4.0
• Issue Type( questions, new requirements, bugs)
The Type of this mechine is Orin AGX 64G
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

Excuse me, in the new deepstream, I added DLA support in the configuration file of Infer, but it hasn’t been implemented yet

How to use models in the DLA?
file.zip (852.8 KB)

I feel that the model is not compatible with DLA

[940655439]

This model is GPU based,
I don’t know how to convert models based on Deepstream
The model that used trtexec to convert DLA also failed

[junshengy]

It looks like there is nothing wrong with your configuration file.
If you modify dstest1_pgie_config.txt as follows, does it work properly?

./deepstream-test1-app /opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.h264 

diff --git a/sources/apps/sample_apps/deepstream-test1/dstest1_pgie_config.txt b/sources/apps/sample_apps/deepstream-test1/dstest1_pgie_config.txt
index 146c732..f7c9d7f 100644
--- a/sources/apps/sample_apps/deepstream-test1/dstest1_pgie_config.txt
+++ b/sources/apps/sample_apps/deepstream-test1/dstest1_pgie_config.txt
@@ -51,7 +51,7 @@
 gpu-id=0
 net-scale-factor=0.00392156862745098
 onnx-file=../../../../samples/models/Primary_Detector/resnet18_trafficcamnet_pruned.onnx
-model-engine-file=../../../../samples/models/Primary_Detector/resnet18_trafficcamnet_pruned.onnx_b1_gpu0_int8.engine
+model-engine-file=../../../../samples/models/Primary_Detector/resnet18_trafficcamnet_pruned.onnx_b1_dla0_int8.engine
 labelfile-path=../../../../samples/models/Primary_Detector/labels.txt
 int8-calib-file=../../../../samples/models/Primary_Detector/cal_trt.bin
 batch-size=1
@@ -62,6 +62,8 @@ gie-unique-id=1
 #scaling-filter=0
 #scaling-compute-hw=0
 cluster-mode=2
+enable-dla=1
+use-dla-core=0
 
 [class-attrs-all]
 pre-cluster-threshold=0.2

DLA core 0 is already working properly

[940655439]

This model was converted to GPU, not DLA, thank you.
The configuration file above has also been added
+enable-dla=1
+use-dla-core=0

[940655439]

The first question is with command trtexec ,can we set batch size
The second question can we transform dla engine in the deepstream?

[940655439]

The first problem has been solved, there is no need to look again

[940655439]

Now we only need to look at the second question

[junshengy]

The following command line is used to generate a int8 batch-size=8 engine file for the DLA0 .

You need to specify the int8 calibration file

DLA requests all profiles have same min, max, and opt value. Otherwise the generated engine file will only run on the GPU.

/usr/src/tensorrt/bin/trtexec --minShapes="input_1:0":8x3x544x960 --maxShapes="input_1:0":8x3x544x960 --optShapes="input_1:0":8x3x544x960 --onnx=/opt/nvidia/deepstream/deepstream/samples/models/Primary_Detector/resnet18_trafficcamnet_pruned.onnx --useDLACore=0 --int8 --calib=/opt/nvidia/deepstream/deepstream/samples/models/Primary_Detector/cal_trt.bin --allowGPUFallback --saveEngine=resnet18_trafficcamnet_pruned.onnx_b8_dla0_int8.engine

@940655439

The second question can we transform dla engine in the deepstream?

Please use the above configuration file to run test1, and DLA0 can be used normally

[940655439]

ok I try it

[Fiona.Chen]

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