ttnn.clone

ttnn.clone() → None

Clones the input, creating a copy with the specified memory_config and converting its data type to dtype. This operation does not alter the tensor’s layout. - ROW_MAJOR_LAYOUT: Returns the tensor unpadded in the last two dimensions. - TILE_LAYOUT: Pads the tensor to ensure its width and height are multiples of 32. If the input’s current layout matches the specified layout, padding adjustments are applied to the last two dimensions as necessary.

:param * input: The tensor to be cloned. :param * dtype: The target data type of the cloned tensor. :param * memory_config: The memory configuration for the clone, options include DRAM_MEMORY_CONFIG or L1_MEMORY_CONFIG. :param * compute_kernel_config: The configuration for the compute kernel.