ttnn.prelu

ttnn.prelu = Operation(python_fully_qualified_name='ttnn.prelu', function=<ttnn._ttnn.operations.binary.prelu_t object>, preprocess_golden_function_inputs=<function default_preprocess_golden_function_inputs>, golden_function=<function _golden_function_prelu>, postprocess_golden_function_outputs=<function default_postprocess_golden_function_outputs>, is_cpp_operation=True, is_experimental=False)

Perform an eltwise-prelu operation.

\[\mathrm{output\_tensor} = \verb|prelu|(\mathrm{input\_tensor\_a,input\_tensor\_b})\]

Parameters:

input_tensor_a (ttnn.Tensor) – the input tensor.
input_tensor_b (ttnn.Tensor or List[float] of length 1 or Number) – weight.

Keyword Arguments:

memory_config (ttnn.MemoryConfig, optional) – memory configuration for the operation. 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, 5

PReLU supports the case where weight is a scalar or 1D list/array of size=1 or a 1D tensor input_tensor_b of size = the second dimension in input_tensor_a

Example

>>> tensor1 = ttnn.from_torch(torch.rand([1, 2, 32, 32], dtype=torch.bfloat16), device=device)
>>> tensor2 = ttnn.from_torch(torch.tensor([1, 2], dtype=torch.bfloat16), device=device)
>>> output = ttnn.prelu(tensor1, tensor2/scalar)