FloatFunctional

class torch.nn.quantized.FloatFunctional

State collector class for float operations.

The instance of this class can be used instead of the torch. prefix for some operations. See example usage below.

Note

This class does not provide a forward hook. Instead, you must use one of the underlying functions (e.g. add).

Examples:

>>> f_add = FloatFunctional()
>>> a = torch.tensor(3.0)
>>> b = torch.tensor(4.0)
>>> f_add.add(a, b)  # Equivalent to ``torch.add(a, b)``

Valid operation names:

• add

• cat

• mul

• add_relu

• add_scalar

• mul_scalar