Riva 1.8.0 deploy pretrained Tacotron2+Waveglow TTS

Hardware - GPU (A100/T4/V100)
Hardware - CPU
Operating System ubuntu 20.04
Riva Version 1.8.0b0
TLT Version (if relevant)

After downloading Tacotron2 and Waveglow (Speech Synthesis — NVIDIA Riva Speech Skills v1.8.0-beta documentation) and following the deploy procedure described on Speech Synthesis — NVIDIA Riva Speech Skills v1.8.0-beta documentation I receive the following error

W0125 21:40:14.122266 140647562446656 schema.py:118] Validation schema not found for nemo.collections.tts.models.Tacotron2Model.
That means Riva does not yet support a pipeline for this network and likely will not work with it.
I0125 21:40:14.896794 140647562446656 artifacts.py:54] Found model at ./model_weights.ckpt
I0125 21:40:14.897564 140647562446656 artifacts.py:120] Retrieved artifacts: dict_keys(['model_config.yaml'])
E0125 21:40:14.898230 140647562446656 cookbook.py:58] Your NeMo model class (nemo.collections.tts.models.tacotron2.Tacotron2Model) is not Exportable.

The same happens if I try to deploy a model trained with the latest NeMo:1.6.0rc0. Any plan on when the models will be supported? At least a note on the official Riva documentation stating that these models are not yet supported would be really welcome.

Please see nemo2riva Tacotron2 model · Issue #3509 · NVIDIA/NeMo · GitHub for the solution

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