Export model straight from .hdf5 without training

How to export a model (like yolo_v4) straight from the pretrained weights without doing train/ transfer learning and/or pruning?

It is not supported for exporting the hdf5 pretrained weight files.
End users can export the tlt files. For example, there are some purpose-built model files. See Overview — TAO Toolkit 3.22.05 documentation

Yes, I want the TLT only(yolo_v4), without having to do training or pruning. I want the default YoloV4 TLT or eTLT. Is the “starting point” Yolo_V4 not usable without 1st doing some sort of training?

As mentioned above, there are some purpose-built model files (tlt format). End users can export the unpruned tlt files or pruned tlt files to etlt files.

The pretrained models in ngc (hdf5 format), see NVIDIA TAO Documentation and NVIDIA TAO Documentation , they are unpruned models with just the feature extractor weights, and may not be used without retraining to deploy in a classification application.

1 Like

This was the answer I was looking for, thank you.

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.