How could I generate a engine file from etlt or tlt in dgpu?

[huihui308]

Please provide the following information when requesting support.

• Hardware (T4/)
• Network Type (Detectnet_v2)
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here)
• Training spec file(If have, please share here)
• How to reproduce the issue ? (This is for errors. Please share the command line and the detailed log here.)
I train a modle in tao, how could I generate a engine file from etlt or tlt in dgpu?

thank you very much.

[Morganh]

Usually there are 5 ways.

1. Use " tao detectnet_v2 export xxx ". There is an option “--engine_file” . It is the expected tensorrt engine.
2. Use " tao converter xxx ". The converter is inside the docker, it can generate tensorrt engine based on .etlt file.
3. Actually it is the same as method 2. But it is running outside the docker. See DetectNet_v2 — TAO Toolkit 3.22.05 documentation and TensorRT — TAO Toolkit 3.22.05 documentation
4. Use deepstream. Config the .etlt file in deepstream config file. Run deepstream and it will generate tensort engine. See DetectNet_v2 — TAO Toolkit 3.22.05 documentation
5. Use triton-app. Config the .etlt file in GitHub - NVIDIA-AI-IOT/tao-toolkit-triton-apps: Sample app code for deploying TAO Toolkit trained models to Triton . Let it generate model.plan(i.e., tensorrt engine)

[huihui308]

That is a very detailed answer, thank you very much.