Utility helpers for the optional PointNeXt VIN encoder.
aria_nbv.vin.encoders.pointnext.PointNeXtSEncoder wraps an OpenPoints PointNeXt-S model for semidense point-cloud context. This sidecar owns the unstructured OpenPoints/checkpoint glue so the torch.nn.Module implementation can stay focused on module lifecycle and tensor contracts.
Tensor, mapping, tuple, or list returned by an external point encoder. OpenPoints variants commonly use keys such as feat or features but checkpoint/model wrappers are not fully uniform.
# vin.encoders.pointnext_utils { #aria_nbv.vin.encoders.pointnext_utils }`vin.encoders.pointnext_utils`Utility helpers for the optional PointNeXt VIN encoder.`aria_nbv.vin.encoders.pointnext.PointNeXtSEncoder` wraps an OpenPointsPointNeXt-S model for semidense point-cloud context. This sidecar owns theunstructured OpenPoints/checkpoint glue so the `torch.nn.Module` implementationcan stay focused on module lifecycle and tensor contracts.## Functions| Name | Description || ---| ---||[extract_point_encoder_tensor](#aria_nbv.vin.encoders.pointnext_utils.extract_point_encoder_tensor)| Return the first tensor-like feature payload from an OpenPoints output. ||[load_checkpoint_strict](#aria_nbv.vin.encoders.pointnext_utils.load_checkpoint_strict)| Load an OpenPoints-style checkpoint with strict PyTorch key matching. ||[load_pointnext_cfg](#aria_nbv.vin.encoders.pointnext_utils.load_pointnext_cfg)| Load a PointNeXt/OpenPoints YAML config with `openpoints.utils.EasyConfig`. |### extract_point_encoder_tensor { #aria_nbv.vin.encoders.pointnext_utils.extract_point_encoder_tensor }```pythonvin.encoders.pointnext_utils.extract_point_encoder_tensor(output)```Return the first tensor-like feature payload from an OpenPoints output.#### Parameters {.doc-section .doc-section-parameters}| Name | Type | Description | Default ||--------|---------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------|| output |[Any](`typing.Any`)| Tensor, mapping, tuple, or list returned by an external point encoder. OpenPoints variants commonly use keys such as ``feat`` or ``features`` but checkpoint/model wrappers are not fully uniform. | _required_ |#### Returns {.doc-section .doc-section-returns}| Name | Type | Description ||--------|--------------------------|-------------------------------------------------|||[Tensor](`torch.Tensor`)| The first `torch.Tensor` feature payload found. |#### Raises {.doc-section .doc-section-raises}| Name | Type | Description ||--------|--------------------------|--------------------------------------|||[TypeError](`TypeError`)| If no tensor payload can be located. |### load_checkpoint_strict { #aria_nbv.vin.encoders.pointnext_utils.load_checkpoint_strict }```pythonvin.encoders.pointnext_utils.load_checkpoint_strict(model, checkpoint_path)```Load an OpenPoints-style checkpoint with strict PyTorch key matching.#### Parameters {.doc-section .doc-section-parameters}| Name | Type | Description | Default ||-----------------|----------------------------------------------|-------------------------------------------------------------------------------------------------------------------------|------------|| model |[nn](`torch.nn`).[Module](`torch.nn.Module`)| PointNeXt module built by OpenPoints. | _required_ || checkpoint_path |[Path](`pathlib.Path`)| Path to a checkpoint containing a state dict directly or under common wrapper keys such as ``model`` or ``state_dict``. | _required_ |#### Raises {.doc-section .doc-section-raises}| Name | Type | Description ||--------|--------------------------------|--------------------------------------------------------------------------------------------------------|||[RuntimeError](`RuntimeError`)| If the checkpoint does not contain a dictionary state payload after unwrapping common OpenPoints keys. |### load_pointnext_cfg { #aria_nbv.vin.encoders.pointnext_utils.load_pointnext_cfg }```pythonvin.encoders.pointnext_utils.load_pointnext_cfg(cfg_path)```Load a PointNeXt/OpenPoints YAML config with `openpoints.utils.EasyConfig`.#### Parameters {.doc-section .doc-section-parameters}| Name | Type | Description | Default ||----------|------------------------|----------------------------------------------------------|------------|| cfg_path |[Path](`pathlib.Path`)| Absolute path to the OpenPoints YAML configuration file. | _required_ |#### Returns {.doc-section .doc-section-returns}| Name | Type | Description ||--------|---------------------|----------------------------------|||[Any](`typing.Any`)| Parsed OpenPoints config object. |