fastvideo.pipelines.preprocess.preprocess_stages#

Module Contents#

Classes#

TextTransformStage

Process text data according to the cfg rate.

VideoTransformStage

Crop a video in temporal dimension.

API#

class fastvideo.pipelines.preprocess.preprocess_stages.TextTransformStage(cfg_uncondition_drop_rate: float, seed: int)[source]#

Bases: fastvideo.pipelines.stages.base.PipelineStage

Process text data according to the cfg rate.

Initialization

forward(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) fastvideo.pipelines.pipeline_batch_info.ForwardBatch[source]#
class fastvideo.pipelines.preprocess.preprocess_stages.VideoTransformStage(train_fps: int, num_frames: int, max_height: int, max_width: int, do_temporal_sample: bool)[source]#

Bases: fastvideo.pipelines.stages.base.PipelineStage

Crop a video in temporal dimension.

Initialization

forward(batch: fastvideo.pipelines.pipeline_batch_info.ForwardBatch, fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) fastvideo.pipelines.pipeline_batch_info.ForwardBatch[source]#