Skip to content

sometimes

Sometimes

Bases: TensorOp

Perform a NumpyOp with a given probability.

Note that Sometimes should not be used to wrap an op whose output key(s) do not already exist in the data dictionary. This would result in a problem when future ops / traces attempt to reference the output key, but Sometimes declined to generate it. If you want to create a default value for a new key, simply use a LambdaOp before invoking the Sometimes.

Parameters:

Name Type Description Default
tensor_op TensorOp

The operator to be performed.

 required
prob float

The probability of execution, which should be in the range: [0-1).

 0.5
Source code in fastestimator/fastestimator/op/tensorop/meta/sometimes.py 
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
@traceable()
class Sometimes(TensorOp):
    """Perform a NumpyOp with a given probability.

    Note that Sometimes should not be used to wrap an op whose output key(s) do not already exist in the data
    dictionary. This would result in a problem when future ops / traces attempt to reference the output key, but
    Sometimes declined to generate it. If you want to create a default value for a new key, simply use a LambdaOp before
    invoking the Sometimes.

    Args:
        tensor_op: The operator to be performed.
        prob: The probability of execution, which should be in the range: [0-1).
    """
    def __init__(self, tensor_op: TensorOp, prob: float = 0.5) -> None:
        # We're going to try to collect any missing output keys from the data dictionary so that they don't get
        # overridden when Sometimes chooses not to execute.
        inps = set(tensor_op.inputs)
        outs = set(tensor_op.outputs)
        self.extra_inputs = list(outs - inps)  # Used by traceability
        self.inp_idx = len(tensor_op.inputs)
        super().__init__(inputs=tensor_op.inputs + self.extra_inputs,
                         outputs=tensor_op.outputs,
                         mode=tensor_op.mode,
                         ds_id=tensor_op.ds_id)
        # Note that in_list and out_list will always be true
        self.op = tensor_op
        self.prob = prob
        self.prob_fn = None

    def build(self, framework: str, device: Optional[torch.device] = None) -> None:
        self.op.build(framework, device)
        if framework == 'tf':
            self.prob_fn = tfp.distributions.Uniform()
        elif framework == 'torch':
            self.prob_fn = torch.distributions.uniform.Uniform(low=0, high=1)
        else:
            raise ValueError("unrecognized framework: {}".format(framework))

    def get_fe_loss_keys(self) -> Set[str]:
        return self.op.get_fe_loss_keys()

    def get_fe_models(self) -> Set[Model]:
        return self.op.get_fe_models()

    def fe_retain_graph(self, retain: Optional[bool] = None) -> Optional[bool]:
        return self.op.fe_retain_graph(retain)

    def __getstate__(self) -> Dict[str, Dict[Any, Any]]:
        return {'op': self.op.__getstate__() if hasattr(self.op, '__getstate__') else {}}

    def forward(self, data: List[Tensor], state: Dict[str, Any]) -> List[Tensor]:
        """Execute the wrapped operator a certain fraction of the time.

        Args:
            data: The information to be passed to the wrapped operator.
            state: Information about the current execution context, for example {"mode": "train"}.

        Returns:
            The original `data`, or the `data` after running it through the wrapped operator.
        """
        if self.prob > self.prob_fn.sample():
            data = data[:self.inp_idx]  # Cut off the unnecessary inputs
            if not self.op.in_list:
                data = data[0]
            data = self.op.forward(data, state)
            if not self.op.out_list:
                data = [data]
        else:
            data = [data[self.inputs.index(out)] for out in self.outputs]
        return data

forward

Execute the wrapped operator a certain fraction of the time.

Parameters:

Name Type Description Default
data List[Tensor]

The information to be passed to the wrapped operator.

 required
state Dict[str, Any]

Information about the current execution context, for example {"mode": "train"}.

 required

Returns:

Type Description
List[Tensor]

The original data, or the data after running it through the wrapped operator.
Source code in fastestimator/fastestimator/op/tensorop/meta/sometimes.py 
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
def forward(self, data: List[Tensor], state: Dict[str, Any]) -> List[Tensor]:
    """Execute the wrapped operator a certain fraction of the time.

    Args:
        data: The information to be passed to the wrapped operator.
        state: Information about the current execution context, for example {"mode": "train"}.

    Returns:
        The original `data`, or the `data` after running it through the wrapped operator.
    """
    if self.prob > self.prob_fn.sample():
        data = data[:self.inp_idx]  # Cut off the unnecessary inputs
        if not self.op.in_list:
            data = data[0]
        data = self.op.forward(data, state)
        if not self.op.out_list:
            data = [data]
    else:
        data = [data[self.inputs.index(out)] for out in self.outputs]
    return data