ttnn.pow_bw

ttnn.pow_bw(grad_tensor: ttnn.Tensor, input_tensor_a: ttnn.Tensor, exponent: Number, *, memory_config: ttnn.MemoryConfig | None = None, output_tensor: ttnn.Tensor | None = None, queue_id: int | None = 0) → List of ttnn.Tensor

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

Parameters:

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

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

  • exponent (Number) – Exponent value [must be non-negative].

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.

  • queue_id (int, optional) – command queue id. Defaults to 0.

Returns:

List of ttnn.Tensor – the output tensor.

Note

Supported dtypes, layouts, and ranks:

Dtypes

Layouts

Ranks

BFLOAT16, BFLOAT8_B

TILE

2, 3, 4

For more details about BFLOAT8_B, refer to the BFLOAT8_B limitations.

Example

>>> grad_tensor = ttnn.from_torch(torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device)
>>> input = ttnn.from_torch(torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16, requires_grad=True), layout=ttnn.TILE_LAYOUT, device=device)
>>> output = ttnn.pow_bw(grad_tensor, tensor, exponent)