Error using Frame VAD Marblenet model with riva-build
#2
by
ibrahim-601
- opened
I am trying to create a custom Riva image with the Conformer-CTC ASR model and Frame VAD, but I am encountering an error with riva-build.
Traceback (most recent call last):
File "/usr/local/bin/riva-build", line 8, in <module>
sys.exit(build())
^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/servicemaker/cli/build.py", line 97, in build
pipeline_config.init_from(nm.get_config())
File "/usr/local/lib/python3.12/dist-packages/servicemaker/pipelines/asr.py", line 782, in init_from
raise Exception(f"Model class {model_class} not recognized")
Exception: Model class nemo.collections.asr.models.classification_models.EncDecFrameClassificationModel not recognized
I converted to riva using nemo2riva==2.21.0
I tried riva-build with both riva-speech:2.18.0 and riva-speech:2.19.0
Hi Ibrahim,
Which version of NeMo are you using? The current main branch in NeMo includes the EncDecFrameClassificationModel class.
https://github.com/NVIDIA/NeMo/blob/main/nemo/collections/asr/models/__init__.py
I am using latest NeMo. I am trying to deploy this model with Riva 2.19.0. Is this model supported by Riva pipeline? When I try to convert this model Riva using nemo2riva it says
[NeMo W 2025-08-22 18:59:55 schema:205] Validation schema not found for nemo.collections.asr.models.EncDecFrameClassificationModel.
That means Riva does not yet support a pipeline for this network and likely will not work with it.
I assume it is not supported by Riva