add_weights_regularizer. Add a weights regularizer to the provided Tensor Arguments: variable: `Variable`. Tensor to add regularization. loss: `str`. Regularization mode. weight_decay: `float`. Decay to use for regularization. add_to_collection: `str`. Add
(variable, loss="L2", weight_decay=0.001,
add_to_collection=None)
| 11 | |
| 12 | |
| 13 | def add_weights_regularizer(variable, loss="L2", weight_decay=0.001, |
| 14 | add_to_collection=None): |
| 15 | """ add_weights_regularizer. |
| 16 | |
| 17 | Add a weights regularizer to the provided Tensor |
| 18 | |
| 19 | Arguments: |
| 20 | variable: `Variable`. Tensor to add regularization. |
| 21 | loss: `str`. Regularization mode. |
| 22 | weight_decay: `float`. Decay to use for regularization. |
| 23 | add_to_collection: `str`. Add the regularization loss to the |
| 24 | specified collection. Default: tf.GraphKeys.REGULARIZATION_LOSSES. |
| 25 | |
| 26 | Returns: |
| 27 | `tf.Tensor`. The weight regularizer. |
| 28 | |
| 29 | """ |
| 30 | if not add_to_collection: |
| 31 | add_to_collection = tf.GraphKeys.REGULARIZATION_LOSSES |
| 32 | if isinstance(loss, str): |
| 33 | regul = regularizers.get(loss) |
| 34 | weights_regularizer = regul(variable, weight_decay) |
| 35 | elif loss and callable(loss): |
| 36 | weights_regularizer = loss(variable) |
| 37 | else: |
| 38 | weights_regularizer = loss |
| 39 | if add_to_collection: |
| 40 | tf.add_to_collection(add_to_collection, weights_regularizer) |
| 41 | return weights_regularizer |
| 42 | |
| 43 | |
| 44 | def add_activation_regularizer(op, loss="L2", activ_decay=0.001, |
nothing calls this directly
no test coverage detected
searching dependent graphs…