ttnn.rpow_bw

ttnn.rpow_bw(grad_tensor: ttnn.Tensor, input_tensor: ttnn.Tensor, exponent: float, *, memory_config: ttnn.MemoryConfig = None) → List of ttnn.Tensor

Performs backward operations for rpow on input_tensor, exponent with given grad_tensor.

\[\mathrm{{output\_tensor}}_i = \verb|rpow_bw|(\mathrm{{grad\_tensor}}_i, \mathrm{{input\_tensor}}_i, \verb|exponent|)\]

Parameters:

• grad_tensor (ttnn.Tensor) – the input gradient tensor.

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

• exponent (float) – Exponent value.

Keyword Arguments:

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

Returns:

List of ttnn.Tensor – the output tensor.

Note

Supported dtypes, layouts, and ranks:

Dtypes	Layouts	Ranks
BFLOAT16, BFLOAT8_B	TILE	2, 3, 4

Example

# Create sample tensors for backward reverse power operation
grad_tensor = ttnn.from_torch(
    torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device
)
input_tensor = ttnn.from_torch(
    torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16, requires_grad=True), layout=ttnn.TILE_LAYOUT, device=device
)
# Define exponent for reverse power operation
exponent = 2.0

# Call the rpow_bw function
output = ttnn.rpow_bw(grad_tensor, input_tensor, exponent)
logger.info(f"Reverse Power Backward: {output}")