ttnn.leaky_relu_bw

ttnn.leaky_relu_bw(grad_tensor: ttnn.Tensor, input_tensor_a: ttnn.Tensor, *, negative_slope: float = 0.01, memory_config: ttnn.MemoryConfig = None) → List of ttnn.Tensor

Performs backward operations for leaky_relu on input_tensor, negative_slope, with given grad_tensor.

Parameters:

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

Keyword Arguments:

negative_slope (float, optional) – negative_slope value for the hardshrink formula . Defaults to 0.01.
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 leaky relu 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
)

# Call the leaky_relu_bw function with negative slope parameter
output = ttnn.leaky_relu_bw(grad_tensor, input_tensor, negative_slope=0.01)
logger.info(f"Leaky ReLU Backward: {output}")