Skip to content

loss

LossOp

Bases: TensorOp

Abstract base LossOp class.

A base class for loss operations. It can be used directly to perform value pass-through (see the adversarial training showcase for an example of when this is useful).

Parameters:

Name Type Description Default
inputs Union[None, str, Iterable[str]]

A tuple or list like: [, ].

 None
outputs Union[None, str, Iterable[str]]

String key under which to store the computed loss.

 None
mode Union[None, str, Iterable[str]]

What mode(s) to execute this Op in. For example, "train", "eval", "test", or "infer". To execute regardless of mode, pass None. To execute in all modes except for a particular one, you can pass an argument like "!infer" or "!train".

 '!infer'
ds_id Union[None, str, Iterable[str]]

What dataset id(s) to execute this Op in. To execute regardless of ds_id, pass None. To execute in all ds_ids except for a particular one, you can pass an argument like "!ds1".

 None
average_loss bool

Whether to average the element-wise loss after the Loss Op.

 True
Source code in fastestimator/fastestimator/op/tensorop/loss/loss.py 
20
21
22
23
24
25
26
27
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
class LossOp(TensorOp):
    """Abstract base LossOp class.

    A base class for loss operations. It can be used directly to perform value pass-through (see the adversarial
    training showcase for an example of when this is useful).

    Args:
        inputs: A tuple or list like: [<y_pred>, <y_true>].
        outputs: String key under which to store the computed loss.
        mode: What mode(s) to execute this Op in. For example, "train", "eval", "test", or "infer". To execute
            regardless of mode, pass None. To execute in all modes except for a particular one, you can pass an argument
            like "!infer" or "!train".
        ds_id: What dataset id(s) to execute this Op in. To execute regardless of ds_id, pass None. To execute in all
            ds_ids except for a particular one, you can pass an argument like "!ds1".
        average_loss: Whether to average the element-wise loss after the Loss Op.
    """
    def __init__(self,
                 inputs: Union[None, str, Iterable[str]] = None,
                 outputs: Union[None, str, Iterable[str]] = None,
                 mode: Union[None, str, Iterable[str]] = "!infer",
                 ds_id: Union[None, str, Iterable[str]] = None,
                 average_loss: bool = True):
        super().__init__(inputs=inputs, outputs=outputs, mode=mode, ds_id=ds_id)
        self.average_loss = average_loss

    @property
    def true_key(self) -> str:
        return self.inputs[self.true_key_idx]

    @property
    def pred_key(self) -> str:
        return self.inputs[self.pred_key_idx]

    @property
    def true_key_idx(self) -> int:
        return 1

    @property
    def pred_key_idx(self) -> int:
        return 0