ttnn.downsample

ttnn.downsample(input_tensor: ttnn.Tensor, downsample_params: List, *, dtype: ttnn.DataType | None = None) → ttnn.Tensor

Downsamples a given multi-channel 2D (spatial) data. The input data is assumed to be of the form [N, H, W, C].

Parameters:

• input_tensor (ttnn.Tensor) – the input tensor.

• downsample_params (List) – Params list: batch size, conv input H, conv input W, conv stride H, conv stride W.

Keyword Arguments:

dtype (ttnn.DataType, optional) – Defaults to None.

Returns:

ttnn.Tensor – the output tensor.