Serialization assertion stdVersionRead == kSERIALIZATION_VERSION failed.Version tag does not match

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

• Hardware: Xavier
• Network Type: Yolo_v4
• Deepstream 7

I generated an engine file from a Yolo_v4 model using Deepstream, but when I want to use that engine file using TensorRT it shows me the following error:

Serialization assertion stdVersionRead == kSERIALIZATION_VERSION failed.Version tag does not match. Note: Current Version: 239, Serialized Engine Version: 236

I also generated the engine file with tao-convert but I got the same error.
That error only occurs with JetPack 6 but not when I use older JetPacks.

3

Are you testing on the same device? TRT engine file is hardware dependent. TRT engine file created on one device can’t be used on a new device.

4

Yes, I generated the engine on the same device and tried to run it in the same environment. Under Deepstream works but under the TensorRT script, it shows me the error message mentioned above.

5

Please make sure you are using the same TensorRT version between deepsteram and your script. Please double check the script.

6

I am using a Jetson Device. I thought only one version of TensorRT was available, So Deepstream and the TensorRT script should use the same version. Is it possible that my device has two versions of TensorRT?

7

Please check the TensorRT version when you generate engine using deepstream.
Then, please use the same TensorRT version when you run any inference script.

8

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