ttnn.tanhshrink

ttnn.tanhshrink(input_tensor: ttnn.Tensor, *, memory_config: ttnn.MemoryConfig = None, output_tensor: ttnn.Tensor = None, fast_and_approximate_mode: Boolean = False) → ttnn.Tensor

Applies tanhshrink to input_tensor element-wise.

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

Parameters:

input_tensor (ttnn.Tensor) – 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.
fast_and_approximate_mode (Boolean, optional) – Enables a performance-optimized approximation method. When True, the operation runs faster but may produce results with minor precision differences. Defaults to False.

Returns:

ttnn.Tensor – the output tensor.

Note

Supported dtypes, layouts, and ranks:

Dtypes	Layouts	Ranks
BFLOAT16, BFLOAT8_B, FLOAT32	TILE	2, 3, 4

BFLOAT8_B/BFLOAT4_B is supported only for approx=True mode.

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 tanh shrink function
output = ttnn.tanhshrink(tensor, fast_and_approximate_mode=False)
logger.info(f"Tanh shrink: {output}")