FloatFunctional¶
- class torch.ao.nn.quantized.FloatFunctional[source]¶
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