ttnn.fold

ttnn.fold(input: ttnn.Tensor, stride_h: int, stride_w: int) → None

Fold TT Tensor. Input tensor must be on TT accelerator device, in ROW_MAJOR. Output tensor will be on TT accelerator device, in ROW_MAJOR.

Parameters:

• input (ttnn.Tensor) – Input tensor to be folded. Tensor of shape [N, H, W, C].

• stride_h (int) – Stride along the H-dimension.

• stride_w (int) – Stride along the W-dimension.