ttnn.addcdiv

ttnn.addcdiv(input_tensor_a: ttnn.Tensor, input_tensor_b: ttnn.Tensor, input_tensor_c: ttnn.Tensor or Number, *, value: float | None, memory_config: ttnn.MemoryConfig = None) → ttnn.Tensor

Multiplies input_tensor_b by a scalar, divides the result element-wise by input_tensor_c, and adds it to input_tensor_a. Returns a tensor with the same layout as input_tensor_a.

\[\mathrm{{output\_tensor}}_i = \mathrm{{input\_tensor\_a}}_i + \frac{(value * \mathrm{input\_tensor\_b}_i)}{\mathrm{input\_tensor\_c}_i}\]

Parameters:

input_tensor_a (ttnn.Tensor) – the input tensor to be added.
input_tensor_b (ttnn.Tensor) – the input numerator tensor.
input_tensor_c (ttnn.Tensor or Number) – the input denominator tensor.

Keyword Arguments:

value (float, optional) – scalar value to be multiplied with input_tensor_b.
memory_config (ttnn.MemoryConfig, optional) – memory configuration for the operation. Defaults to None.

Returns:

ttnn.Tensor – the output tensor.

Note

Supported dtypes and layouts:

Dtypes	Layouts
FLOAT32, BFLOAT16, BFLOAT8_B	TILE

bfloat8_b/bfloat4_b supports only on TILE_LAYOUT

Example

# Create three tensors and a value for the operation
value = 1.0
tensor1 = ttnn.from_torch(
    torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device
)
tensor2 = ttnn.from_torch(
    torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device
)
tensor3 = ttnn.from_torch(
    torch.tensor([[1, 2], [3, 4]], dtype=torch.bfloat16), layout=ttnn.TILE_LAYOUT, device=device
)

# Perform the addcdiv operation: tensor1 + value * (tensor2 / tensor3)
output = ttnn.addcdiv(tensor1, tensor2, tensor3, value=value)
logger.info(f"Addcdiv result: {output}")