ertk.pytorch.utils.frame_tensor

ertk.pytorch.utils.frame_tensor(x: Tensor, frame_size: int, frame_shift: int, pad: bool = False, axis: int = 0) → Tensor

Frames an array over a given axis with optional padding and copying.

Parameters:

x: torch.Tensor: Array to frame.
frame_size: int: Width of frame.
frame_shift: int: Amount each window is moved for subsequent frames.
pad: bool: If the frames don’t neatly fit the axis length, this indicated whether to pad or cut the remaining length. Default is False which means any remaining length is cut. If pad=True then any remaining length is padded with zeros and an additional frame is produced. If frames neatly fit into the axis length then this option has no effect.
axis: int: Axis to frame over. Default is 0.

Returns: