ttnn.untilize_with_unpadding

ttnn.untilize_with_unpadding(input_tensor: ttnn.Tensor, output_tensor_end: shape, *, memory_config: ttnn.MemoryConfig = None, use_multicore: bool = True, use_pack_untilize: bool = True) → List of ttnn.Tensor

Changes data layout of input tensor to ROW_MAJOR and unpads/removes elements from the tensor.

Input tensor must be on TT accelerator device, in TILE layout, and have BFLOAT16 data type.

Output tensor will be on TT accelerator device, in ROW_MAJOR layout, and have BFLOAT16 data type.

Parameters:

  • input_tensor (ttnn.Tensor) – the input tensor

  • output_tensor_end (shape) – End indices of input tensor in output tensor.

Keyword Arguments:

  • memory_config (ttnn.MemoryConfig, optional) – Memory configuration for the operation. Defaults to None.

  • use_multicore (bool, optional) – Whether to use multicore. Defaults to True.

  • use_pack_untilize (bool, optional) – Whether to use pack untilize. Defaults to True.

Returns:

List of ttnn.Tensor – the output tensor.

Example

# Create a tilized tensor with padding
tilized_tensor = ttnn.rand((1, 1, 64, 64), dtype=ttnn.bfloat16, layout=ttnn.Layout.TILE, device=device)

# Untilize the tensor with unpadding
untilized_tensor = ttnn.untilize_with_unpadding(tilized_tensor, output_tensor_end=(1, 1, 64, 32))