# Copyright 2018 The TensorFlow Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== """Utilities related to loss functions.""" import tensorflow.compat.v2 as tf from keras import backend from keras.engine import keras_tensor from keras.utils import tf_utils # isort: off from tensorflow.python.util.tf_export import keras_export @keras_export("keras.losses.Reduction", v1=[]) class ReductionV2: """Types of loss reduction. Contains the following values: * `AUTO`: Indicates that the reduction option will be determined by the usage context. For almost all cases this defaults to `SUM_OVER_BATCH_SIZE`. When used with `tf.distribute.Strategy`, outside of built-in training loops such as `tf.keras` `compile` and `fit`, we expect reduction value to be `SUM` or `NONE`. Using `AUTO` in that case will raise an error. * `NONE`: No **additional** reduction is applied to the output of the wrapped loss function. When non-scalar losses are returned to Keras functions like `fit`/`evaluate`, the unreduced vector loss is passed to the optimizer but the reported loss will be a scalar value. Caution: **Verify the shape of the outputs when using** `Reduction.NONE`. The builtin loss functions wrapped by the loss classes reduce one dimension (`axis=-1`, or `axis` if specified by loss function). `Reduction.NONE` just means that no **additional** reduction is applied by the class wrapper. For categorical losses with an example input shape of `[batch, W, H, n_classes]` the `n_classes` dimension is reduced. For pointwise losses you must include a dummy axis so that `[batch, W, H, 1]` is reduced to `[batch, W, H]`. Without the dummy axis `[batch, W, H]` will be incorrectly reduced to `[batch, W]`. * `SUM`: Scalar sum of weighted losses. * `SUM_OVER_BATCH_SIZE`: Scalar `SUM` divided by number of elements in losses. This reduction type is not supported when used with `tf.distribute.Strategy` outside of built-in training loops like `tf.keras` `compile`/`fit`. You can implement 'SUM_OVER_BATCH_SIZE' using global batch size like: ``` with strategy.scope(): loss_obj = tf.keras.losses.CategoricalCrossentropy( reduction=tf.keras.losses.Reduction.NONE) .... loss = tf.reduce_sum(loss_obj(labels, predictions)) * (1. / global_batch_size) ``` Please see the [custom training guide]( https://www.tensorflow.org/tutorials/distribute/custom_training) for more details on this. """ AUTO = "auto" NONE = "none" SUM = "sum" SUM_OVER_BATCH_SIZE = "sum_over_batch_size" @classmethod def all(cls): return (cls.AUTO, cls.NONE, cls.SUM, cls.SUM_OVER_BATCH_SIZE) @classmethod def validate(cls, key): if key not in cls.all(): raise ValueError( f'Invalid Reduction Key: {key}. Expected keys are "{cls.all()}"' ) def remove_squeezable_dimensions( labels, predictions, expected_rank_diff=0, name=None ): """Squeeze last dim if ranks differ from expected by exactly 1. In the common case where we expect shapes to match, `expected_rank_diff` defaults to 0, and we squeeze the last dimension of the larger rank if they differ by 1. But, for example, if `labels` contains class IDs and `predictions` contains 1 probability per class, we expect `predictions` to have 1 more dimension than `labels`, so `expected_rank_diff` would be 1. In this case, we'd squeeze `labels` if `rank(predictions) - rank(labels) == 0`, and `predictions` if `rank(predictions) - rank(labels) == 2`. This will use static shape if available. Otherwise, it will add graph operations, which could result in a performance hit. Args: labels: Label values, a `Tensor` whose dimensions match `predictions`. predictions: Predicted values, a `Tensor` of arbitrary dimensions. expected_rank_diff: Expected result of `rank(predictions) - rank(labels)`. name: Name of the op. Returns: Tuple of `labels` and `predictions`, possibly with last dim squeezed. """ with backend.name_scope(name or "remove_squeezable_dimensions"): if not tf_utils.is_tensor_or_extension_type(predictions): predictions = tf.convert_to_tensor(predictions) if not tf_utils.is_tensor_or_extension_type(labels): labels = tf.convert_to_tensor(labels) predictions_shape = predictions.shape predictions_rank = predictions_shape.ndims labels_shape = labels.shape labels_rank = labels_shape.ndims if (labels_rank is not None) and (predictions_rank is not None): # Use static rank. rank_diff = predictions_rank - labels_rank if rank_diff == expected_rank_diff + 1 and predictions_shape.dims[ -1 ].is_compatible_with(1): predictions = tf.squeeze(predictions, [-1]) elif rank_diff == expected_rank_diff - 1 and labels_shape.dims[ -1 ].is_compatible_with(1): labels = tf.squeeze(labels, [-1]) return labels, predictions # Use dynamic rank. rank_diff = tf.rank(predictions) - tf.rank(labels) if (predictions_rank is None) or ( predictions_shape.dims[-1].is_compatible_with(1) ): predictions = tf.cond( tf.equal(expected_rank_diff + 1, rank_diff), lambda: tf.squeeze(predictions, [-1]), lambda: predictions, ) if (labels_rank is None) or ( labels_shape.dims[-1].is_compatible_with(1) ): labels = tf.cond( tf.equal(expected_rank_diff - 1, rank_diff), lambda: tf.squeeze(labels, [-1]), lambda: labels, ) return labels, predictions def squeeze_or_expand_dimensions(y_pred, y_true=None, sample_weight=None): """Squeeze or expand last dimension if needed. 1. Squeezes last dim of `y_pred` or `y_true` if their rank differs by 1 (using `remove_squeezable_dimensions`). 2. Squeezes or expands last dim of `sample_weight` if its rank differs by 1 from the new rank of `y_pred`. If `sample_weight` is scalar, it is kept scalar. This will use static shape if available. Otherwise, it will add graph operations, which could result in a performance hit. Args: y_pred: Predicted values, a `Tensor` of arbitrary dimensions. y_true: Optional label `Tensor` whose dimensions match `y_pred`. sample_weight: Optional weight scalar or `Tensor` whose dimensions match `y_pred`. Returns: Tuple of `y_pred`, `y_true` and `sample_weight`. Each of them possibly has the last dimension squeezed, `sample_weight` could be extended by one dimension. If `sample_weight` is None, (y_pred, y_true) is returned. """ y_pred_shape = y_pred.shape y_pred_rank = y_pred_shape.ndims if y_true is not None: # If sparse matrix is provided as `y_true`, the last dimension in # `y_pred` may be > 1. Eg: y_true = [0, 1, 2] (shape=(3,)), y_pred = # [[.9, .05, .05], [.5, .89, .6], [.05, .01, .94]] (shape=(3, 3)) In # this case, we should not try to remove squeezable dimension. y_true_shape = y_true.shape y_true_rank = y_true_shape.ndims if (y_true_rank is not None) and (y_pred_rank is not None): # Use static rank for `y_true` and `y_pred`. if (y_pred_rank - y_true_rank != 1) or y_pred_shape[-1] == 1: y_true, y_pred = remove_squeezable_dimensions(y_true, y_pred) else: # Use dynamic rank. rank_diff = tf.rank(y_pred) - tf.rank(y_true) squeeze_dims = lambda: remove_squeezable_dimensions(y_true, y_pred) is_last_dim_1 = tf.equal(1, tf.shape(y_pred)[-1]) maybe_squeeze_dims = lambda: tf.cond( is_last_dim_1, squeeze_dims, lambda: (y_true, y_pred) ) y_true, y_pred = tf.cond( tf.equal(1, rank_diff), maybe_squeeze_dims, squeeze_dims ) if sample_weight is None: return y_pred, y_true weights_shape = sample_weight.shape weights_rank = weights_shape.ndims if weights_rank == 0: # If weights is scalar, do nothing. return y_pred, y_true, sample_weight if (y_pred_rank is not None) and (weights_rank is not None): # Use static rank. if weights_rank - y_pred_rank == 1: sample_weight = tf.squeeze(sample_weight, [-1]) elif y_pred_rank - weights_rank == 1: sample_weight = tf.expand_dims(sample_weight, [-1]) return y_pred, y_true, sample_weight # Use dynamic rank. weights_rank_tensor = tf.rank(sample_weight) rank_diff = weights_rank_tensor - tf.rank(y_pred) maybe_squeeze_weights = lambda: tf.squeeze(sample_weight, [-1]) def _maybe_expand_weights(): expand_weights = lambda: tf.expand_dims(sample_weight, [-1]) return tf.cond( tf.equal(rank_diff, -1), expand_weights, lambda: sample_weight ) def _maybe_adjust_weights(): return tf.cond( tf.equal(rank_diff, 1), maybe_squeeze_weights, _maybe_expand_weights ) # squeeze or expand last dim of `sample_weight` if its rank differs by 1 # from the new rank of `y_pred`. sample_weight = tf.cond( tf.equal(weights_rank_tensor, 0), lambda: sample_weight, _maybe_adjust_weights, ) return y_pred, y_true, sample_weight def _safe_mean(losses, num_present): """Computes a safe mean of the losses. Args: losses: `Tensor` whose elements contain individual loss measurements. num_present: The number of measurable elements in `losses`. Returns: A scalar representing the mean of `losses`. If `num_present` is zero, then zero is returned. """ total_loss = tf.reduce_sum(losses) return tf.math.divide_no_nan(total_loss, num_present, name="value") def _num_elements(losses): """Computes the number of elements in `losses` tensor.""" with backend.name_scope("num_elements") as scope: return tf.cast(tf.size(losses, name=scope), dtype=losses.dtype) def reduce_weighted_loss( weighted_losses, reduction=ReductionV2.SUM_OVER_BATCH_SIZE ): """Reduces the individual weighted loss measurements.""" if reduction == ReductionV2.NONE: loss = weighted_losses else: loss = tf.reduce_sum(weighted_losses) if reduction == ReductionV2.SUM_OVER_BATCH_SIZE: loss = _safe_mean(loss, _num_elements(weighted_losses)) return loss @keras_export("keras.__internal__.losses.compute_weighted_loss", v1=[]) def compute_weighted_loss( losses, sample_weight=None, reduction=ReductionV2.SUM_OVER_BATCH_SIZE, name=None, ): """Computes the weighted loss. Args: losses: `Tensor` of shape `[batch_size, d1, ... dN]`. sample_weight: Optional `Tensor` whose rank is either 0, or the same rank as `losses`, or be broadcastable to `losses`. reduction: (Optional) Type of `tf.keras.losses.Reduction` to apply to loss. Default value is `SUM_OVER_BATCH_SIZE`. name: Optional name for the op. Raises: ValueError: If the shape of `sample_weight` is not compatible with `losses`. Returns: Weighted loss `Tensor` of the same type as `losses`. If `reduction` is `NONE`, this has the same shape as `losses`; otherwise, it is scalar. """ ReductionV2.validate(reduction) # If this function is called directly, then we just default 'AUTO' to # 'SUM_OVER_BATCH_SIZE'. Eg. Canned estimator use cases. if reduction == ReductionV2.AUTO: reduction = ReductionV2.SUM_OVER_BATCH_SIZE if sample_weight is None: sample_weight = 1.0 with backend.name_scope(name or "weighted_loss"): # Save the `reduction` argument for loss normalization when distributing # to multiple replicas. Used only for estimator + v1 optimizer flow. tf.compat.v1.get_default_graph()._last_loss_reduction = reduction if not isinstance(losses, (keras_tensor.KerasTensor, tf.RaggedTensor)): losses = tf.convert_to_tensor(losses) if not isinstance( sample_weight, (keras_tensor.KerasTensor, tf.RaggedTensor) ): sample_weight = tf.convert_to_tensor(sample_weight) # Convert any non float dtypes to floats, to avoid it loss any precision # for dtype like int or bool. if not losses.dtype.is_floating: input_dtype = losses.dtype losses = tf.cast(losses, "float32") input_casted = True else: input_casted = False sample_weight = tf.cast(sample_weight, losses.dtype) # Update dimensions of `sample_weight` to match with `losses` if # possible. ( losses, _, sample_weight, ) = squeeze_or_expand_dimensions(losses, None, sample_weight) weighted_losses = tf.multiply(losses, sample_weight) # Apply reduction function to the individual weighted losses. loss = reduce_weighted_loss(weighted_losses, reduction) if input_casted: # Convert the result back to the input type. loss = tf.cast(loss, input_dtype) return loss def scale_loss_for_distribution(loss_value): """Scales and returns the given loss value by the number of replicas.""" num_replicas = tf.distribute.get_strategy().num_replicas_in_sync if num_replicas > 1: loss_value *= 1.0 / num_replicas return loss_value def cast_losses_to_common_dtype(losses): """Cast a list of losses to a common dtype. If any loss is floating-point, they will all be casted to the most-precise floating-point loss. Otherwise the losses are not casted. We also skip casting losses if there are any complex losses. Args: losses: A list of losses. Returns: `losses`, but they have been casted to a common dtype. """ highest_float = None for loss in losses: if loss.dtype.is_floating: if highest_float is None or loss.dtype.size > highest_float.size: highest_float = loss.dtype elif {loss.dtype, highest_float} == {"bfloat16", "float16"}: highest_float = "float32" if loss.dtype.is_complex: return ( losses # If we find any complex losses, do not cast any losses ) if highest_float: losses = [tf.cast(loss, highest_float) for loss in losses] return losses def get_mask(y_p): """Returns Keras mask from tensor.""" return getattr(y_p, "_keras_mask", None) def apply_mask(y_p, sw, mask): """Applies any mask on predictions to sample weights.""" if mask is not None: mask = tf.cast(mask, y_p.dtype) if sw is not None: sw = tf.cast(sw, mask.dtype) mask, _, sw = squeeze_or_expand_dimensions(mask, sample_weight=sw) sw *= mask else: sw = mask return sw def apply_valid_mask(losses, sw, mask, reduction): """Redistribute sample weights considering only valid entries.""" if mask is not None: mask = tf.cast(mask, losses.dtype) if reduction in (ReductionV2.AUTO, ReductionV2.SUM_OVER_BATCH_SIZE): # Valid entries have weight `total/valid`, while invalid ones # have 0. When summed over batch, they will be reduced to: # # mean(loss * sample_weight * total / valid) # = sum(loss * sample_weight * total / valid) / total # = sum(loss * sample_weight) / total * total / valid # = sum(loss * sample_weight) / valid total = tf.cast(tf.size(mask), losses.dtype) valid = tf.reduce_sum(mask) mask *= total / valid return apply_mask(losses, sw, mask)