ttnn.reciprocal

ttnn.reciprocal(input_tensor: ttnn.Tensor or ComplexTensor, *, memory_config: ttnn.MemoryConfig = None, output_tensor: ttnn.Tensor = None) → ttnn.Tensor

Applies reciprocal to input_tensor element-wise.

\[\mathrm{output\_tensor}_i = \verb|reciprocal|(\mathrm{input\_tensor}_i)\]
Parameters:

input_tensor (ttnn.Tensor or ComplexTensor) – the input tensor.

Keyword Arguments:

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

  • output_tensor (ttnn.Tensor, optional) – preallocated output tensor. Defaults to None.

Returns:

ttnn.Tensor – the output tensor.

Note

Supported dtypes, layouts, and ranks:

Dtypes

Layouts

Ranks

BFLOAT16, BFLOAT8_B

TILE

2, 3, 4

BFLOAT8_B is supported only for non-zero inputs. Inputs containing zero may produce inaccurate results due to the characteristics of the block-FP format. More information about the BFLOAT8_B.

Example

# Create a tensor with specific values
tensor = ttnn.from_torch(
    torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device
)

# Compute the reciprocal (1/x)
output = ttnn.reciprocal(tensor)
logger.info(f"Reciprocal: {output}")