ttnn.relu_max

ttnn.relu_max(input_tensor: ttnn.Tensor, upper_limit: float or int, *, memory_config: ttnn.MemoryConfig = None, output_tensor: ttnn.Tensor = None, sub_core_grids: ttnn.CoreRangeSet = None) → ttnn.Tensor

Applies relu_max to input_tensor element-wise with upper_limit.

This function caps off the input to a max value and a min value of 0.

\[\mathrm{output\_tensor}_i = \min(\max(\mathrm{input\_tensor}_i, 0), \verb|upper_limit|)\]

Parameters:

input_tensor (ttnn.Tensor) – the input tensor.
upper_limit (float or int) – The max value for ReLU function..

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.
sub_core_grids (ttnn.CoreRangeSet, optional) – sub core grids for the operation. Defaults to None.

Returns:

ttnn.Tensor – the output tensor.

Note

Supported dtypes and layouts:

Dtypes	Layouts
FLOAT32, BFLOAT16, BFLOAT8_B, INT32, UINT32, UINT16, UINT8	TILE, ROW_MAJOR

System memory is not supported.

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,
)
upper_limit = 3

# Apply ReLU with upper limit
output = ttnn.relu_max(tensor, upper_limit)
logger.info(f"ReLU max: {output}")