reducers
BaseReducer
¶
Bases: BaseHook
, ABC
Converts an unreduced loss tensor into a single number.
In other words, if the loss tensor has shape (N,)
,
the reducer converts it to shape (1,)
.
Source code in pytorch_adapt\hooks\reducers.py
14 15 16 17 18 19 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 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 |
|
__init__(apply_to=None, default_reducer=None, **kwargs)
¶
Parameters:
Name | Type | Description | Default |
---|---|---|---|
apply_to |
List[str]
|
list of loss names to apply reduction to |
None
|
default_reducer |
'BaseReducer'
|
a reducer to use for losses that
are not already reduced and are also not
specified in |
None
|
Source code in pytorch_adapt\hooks\reducers.py
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 |
|
EntropyReducer
¶
Bases: BaseReducer
Implementation of "entropy conditioning" from
Conditional Adversarial Domain Adaptation.
It weights loss elements using
EntropyWeights
.
The entropy weights are derived from classifier logits.
Source code in pytorch_adapt\hooks\reducers.py
94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 |
|
__init__(f_hook=None, domains=None, entropy_weights_fn=None, detach_weights=True, **kwargs)
¶
Parameters:
Name | Type | Description | Default |
---|---|---|---|
f_hook |
BaseHook
|
the hook for computing logits from which entropy weights are derived |
None
|
domains |
List[str]
|
the domains that |
None
|
entropy_weights_fn |
Callable[[torch.Tensor], torch.Tensor]
|
the function for computing the weights that will be multiplied with the unreduced losses. |
None
|
detach_weights |
bool
|
If |
True
|
Source code in pytorch_adapt\hooks\reducers.py
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 |
|
MeanReducer
¶
Bases: BaseReducer
Reduces loss elements by taking the mean.
Source code in pytorch_adapt\hooks\reducers.py
160 161 162 163 164 165 166 167 168 169 170 171 172 |
|