ttnn.softplus

ttnn.softplus(input_tensor: ttnn.Tensor, *, beta: float = 1, threshold: float = 20, memory_config: ttnn.MemoryConfig = None, output_tensor: ttnn.Tensor = None) → ttnn.Tensor

Applies softplus to input_tensor element-wise.

\[\mathrm{output\_tensor}_i = softplus(\mathrm{input\_tensor}_i)\]

Parameters:

input_tensor (ttnn.Tensor) – the input tensor.

Keyword Arguments:

beta (float, optional) – Scales the input before applying the Softplus function. By modifying beta, you can adjust the steepness of the function. A higher beta value makes the function steeper, approaching a hard threshold like the ReLU function for large values of beta. Defaults to 1.
threshold (float, optional) – Used to switch to a linear function for large values to improve numerical stability. This avoids issues with floating-point representation for very large values. Defaults to 20.
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

Example

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

# Apply Softplus activation function
output = ttnn.softplus(tensor, beta=1.0, threshold=20.0)
logger.info(f"Softplus: {output}")