stages.video.embedding.internvideo2
#
Module Contents#
Classes#
Stage for generating embeddings from InternVideo2 input frames. |
|
Stage for creating InternVideo2 input frames from video clips. |
API#
- class stages.video.embedding.internvideo2.InternVideo2EmbeddingStage#
Bases:
nemo_curator.stages.base.ProcessingStage
[nemo_curator.tasks.video.VideoTask
,nemo_curator.tasks.video.VideoTask
]Stage for generating embeddings from InternVideo2 input frames.
This class processes video clips through a series of steps including embedding generation, text verification, and memory management.
- gpu_memory_gb: float#
10.0
- inputs() tuple[list[str], list[str]] #
Define stage input requirements.
Returns (tuple[list[str], list[str]]): Tuple of (required_attributes, required_columns) where: - required_top_level_attributes: List of task attributes that must be present - required_data_attributes: List of attributes within the data that must be present
- model_dir: str#
‘InternVideo2’
- num_gpus_per_worker: float#
1.0
- outputs() tuple[list[str], list[str]] #
Define stage output specification.
Returns (tuple[list[str], list[str]]): Tuple of (output_attributes, output_columns) where: - output_top_level_attributes: List of task attributes this stage adds/modifies - output_data_attributes: List of attributes within the data that this stage adds/modifies
- process(
- task: nemo_curator.tasks.video.VideoTask,
Process a task and return the result. Args: task (X): Input task to process Returns (Y | list[Y]): - Single task: For 1-to-1 transformations - List of tasks: For 1-to-many transformations (e.g., readers) - None: If the task should be filtered out
- setup(
- worker_metadata: nemo_curator.backends.base.WorkerMetadata | None = None,
Setup method called once before processing begins. Override this method to perform any initialization that should happen once per worker. Args: worker_metadata (WorkerMetadata, optional): Information about the worker (provided by some backends)
- setup_on_node(
- node_info: nemo_curator.backends.base.NodeInfo,
- worker_metadata: nemo_curator.backends.base.WorkerMetadata,
Download the weights for the InternVideo2 model on the node.
- texts_to_verify: list[str] | None#
None
- verbose: bool#
False
- class stages.video.embedding.internvideo2.InternVideo2FrameCreationStage#
Bases:
nemo_curator.stages.base.ProcessingStage
[nemo_curator.tasks.video.VideoTask
,nemo_curator.tasks.video.VideoTask
]Stage for creating InternVideo2 input frames from video clips.
This class processes video clips through a series of steps including frame extraction, model initialization, and input frame creation.
- inputs() tuple[list[str], list[str]] #
Define stage input requirements.
Returns (tuple[list[str], list[str]]): Tuple of (required_attributes, required_columns) where: - required_top_level_attributes: List of task attributes that must be present - required_data_attributes: List of attributes within the data that must be present
- model_dir: str#
‘InternVideo2’
- outputs() tuple[list[str], list[str]] #
Define stage output specification.
Returns (tuple[list[str], list[str]]): Tuple of (output_attributes, output_columns) where: - output_top_level_attributes: List of task attributes this stage adds/modifies - output_data_attributes: List of attributes within the data that this stage adds/modifies
- process(
- task: nemo_curator.tasks.video.VideoTask,
Process a task and return the result. Args: task (X): Input task to process Returns (Y | list[Y]): - Single task: For 1-to-1 transformations - List of tasks: For 1-to-many transformations (e.g., readers) - None: If the task should be filtered out
- setup(
- worker_metadata: nemo_curator.backends.base.WorkerMetadata | None = None,
Setup method called once before processing begins. Override this method to perform any initialization that should happen once per worker. Args: worker_metadata (WorkerMetadata, optional): Information about the worker (provided by some backends)
- target_fps: float#
2.0
- verbose: bool#
False