nemo_curator.models.audio.sed

Sound-event-detection model architectures and lazy PANNs registry.

Resolution is lazy so importing the SED stage does not require torchlibrosa until a model is built.

Submodules

nemo_curator.models.audio.sed.base
nemo_curator.models.audio.sed.build_sed_tensorrt_engine
nemo_curator.models.audio.sed.cnn14
nemo_curator.models.audio.sed.panns
nemo_curator.models.audio.sed.tensorrt

Package Contents

Functions

Name	Description
`get_model_class`	Resolve a checkpoint name to its vendored architecture class.

Data

SUPPORTED_MODEL_TYPES

_MODEL_CLASS_NAMES

API

nemo_curator.models.audio.sed.get_model_class(
    model_type: str
) -> type[torch.nn.Module]

Resolve a checkpoint name to its vendored architecture class.

nemo_curator.models.audio.sed.SUPPORTED_MODEL_TYPES: tuple[str, ...] = tuple(_MODEL_CLASS_NAMES)

nemo_curator.models.audio.sed._MODEL_CLASS_NAMES: dict[str, str] = {'Cnn14_DecisionLevelMax': 'Cnn14DecisionLevelMax', 'Cnn14_DecisionLevelAvg': 'C...