TLT to onnx at TAO5.0

Continuing the discussion from Tlt model to onnx:

Hi,

As of TAO5.0 is supporting onnx export, can I now convert tlt to onnx format?

or any plans?

Thanks.

For TAO 5.0, when training finished, there is a hdf5 file now instead of tlt model.
You can export it to onnx format.

I see. Instead of converting tlt to hdf5 directly, I can train or retrain tlt model then export it to onnx format. right?

Yes, it is expected to do so.
The .tlt file is actually the encryption of hdf5 file. A trick is to leverage https://github.com/NVIDIA/tao_tensorflow2_backend/blob/main/internal/decode_eff.py. Run the python file directly inside the docker.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.