Internal change

PiperOrigin-RevId: 481234282

official/projects/yt8m/tasks/yt8m_task.py

@@ -13,6 +13,8 @@
 # limitations under the License.
 
 """Video classification task definition."""
+from typing import Dict, List, Optional, Tuple
+
 from absl import logging
 import tensorflow as tf
 
@@ -95,31 +97,46 @@ class YT8MTask(base_task.Task):
 
     return dataset
 
-  def build_losses(self, labels, model_outputs, aux_losses=None):
+  def build_losses(self,
+                   labels,
+                   model_outputs,
+                   label_weights=None,
+                   aux_losses=None):
     """Sigmoid Cross Entropy.
 
     Args:
       labels: tensor containing truth labels.
       model_outputs: output logits of the classifier.
+      label_weights: optional tensor of label weights.
       aux_losses: tensor containing auxiliarly loss tensors, i.e. `losses` in
         keras.Model.
 
     Returns:
-      Tensors: The total loss, model loss tensors.
+      A dict of tensors contains total loss, model loss tensors.
     """
     losses_config = self.task_config.losses
     model_loss = tf.keras.losses.binary_crossentropy(
         labels,
         model_outputs,
         from_logits=losses_config.from_logits,
-        label_smoothing=losses_config.label_smoothing)
+        label_smoothing=losses_config.label_smoothing,
+        axis=None)
 
+    if label_weights is None:
       model_loss = tf_utils.safe_mean(model_loss)
+    else:
+      model_loss = model_loss * label_weights
+      # Manutally compute weighted mean loss.
+      total_loss = tf.reduce_sum(model_loss)
+      total_weight = tf.cast(
+          tf.reduce_sum(label_weights), dtype=total_loss.dtype)
+      model_loss = tf.math.divide_no_nan(total_loss, total_weight)
+
     total_loss = model_loss
     if aux_losses:
       total_loss += tf.add_n(aux_losses)
 
-    return total_loss, model_loss
+    return {'total_loss': total_loss, 'model_loss': model_loss}
 
   def build_metrics(self, training=True):
     """Gets streaming metrics for training/validation.
@@ -130,10 +147,10 @@ class YT8MTask(base_task.Task):
        top_n: A positive Integer specifying the average precision at n, or None
         to use all provided data points.
     Args:
-      training: bool value, true for training mode, false for eval/validation.
+      training: Bool value, true for training mode, false for eval/validation.
 
     Returns:
-      list of strings that indicate metrics to be used
+      A list of strings that indicate metrics to be used.
     """
     metrics = []
     metric_names = ['total_loss', 'model_loss']
@@ -149,15 +166,48 @@ class YT8MTask(base_task.Task):
 
     return metrics
 
+  def process_metrics(self,
+                      metrics: List[tf.keras.metrics.Metric],
+                      labels: tf.Tensor,
+                      outputs: tf.Tensor,
+                      model_losses: Optional[Dict[str, tf.Tensor]] = None,
+                      label_weights: Optional[tf.Tensor] = None,
+                      training: bool = True,
+                      **kwargs) -> Dict[str, Tuple[tf.Tensor, ...]]:
+    """Updates metrics.
+
+    Args:
+      metrics: Evaluation metrics to be updated.
+      labels: A tensor containing truth labels.
+      outputs: Model output logits of the classifier.
+      model_losses: An optional dict of model losses.
+      label_weights: Optional label weights, can be broadcast into shape of
+        outputs/labels.
+      training: Bool indicates if in training mode.
+      **kwargs: Additional input arguments.
+
+    Returns:
+      Updated dict of metrics log.
+    """
+    if model_losses is None:
+      model_losses = {}
+
+    logs = {}
+    if not training:
+      logs.update({self.avg_prec_metric.name: (labels, outputs)})
+
+    for m in metrics:
+      m.update_state(model_losses[m.name])
+      logs[m.name] = m.result()
+    return logs
+
   def train_step(self, inputs, model, optimizer, metrics=None):
     """Does forward and backward.
 
     Args:
-      inputs: a dictionary of input tensors. output_dict = {
-          "video_ids": batch_video_ids,
-          "video_matrix": batch_video_matrix,
-          "labels": batch_labels,
-          "num_frames": batch_frames, }
+      inputs: a dictionary of input tensors. output_dict = { "video_ids":
+        batch_video_ids, "video_matrix": batch_video_matrix, "labels":
+        batch_labels, "num_frames": batch_frames, }
       model: the model, forward pass definition.
       optimizer: the optimizer for this training step.
       metrics: a nested structure of metrics objects.
@@ -167,6 +217,7 @@ class YT8MTask(base_task.Task):
     """
     features, labels = inputs['video_matrix'], inputs['labels']
     num_frames = inputs['num_frames']
+    label_weights = inputs.get('label_weights', None)
 
     # sample random frames / random sequence
     num_frames = tf.cast(num_frames, tf.float32)
@@ -183,26 +234,28 @@ class YT8MTask(base_task.Task):
       # Casting output layer as float32 is necessary when mixed_precision is
       # mixed_float16 or mixed_bfloat16 to ensure output is casted as float32.
       outputs = tf.nest.map_structure(lambda x: tf.cast(x, tf.float32), outputs)
-
       # Computes per-replica loss
-      loss, model_loss = self.build_losses(
-          model_outputs=outputs, labels=labels, aux_losses=model.losses)
+      all_losses = self.build_losses(
+          model_outputs=outputs,
+          labels=labels,
+          label_weights=label_weights,
+          aux_losses=model.losses)
+
+      loss = all_losses['total_loss']
       # Scales loss as the default gradients allreduce performs sum inside the
       # optimizer.
       scaled_loss = loss / num_replicas
 
       # For mixed_precision policy, when LossScaleOptimizer is used, loss is
       # scaled for numerical stability.
-      if isinstance(optimizer,
-                    tf.keras.mixed_precision.LossScaleOptimizer):
+      if isinstance(optimizer, tf.keras.mixed_precision.LossScaleOptimizer):
         scaled_loss = optimizer.get_scaled_loss(scaled_loss)
 
     tvars = model.trainable_variables
     grads = tape.gradient(scaled_loss, tvars)
     # Scales back gradient before apply_gradients when LossScaleOptimizer is
     # used.
-    if isinstance(optimizer,
-                  tf.keras.mixed_precision.LossScaleOptimizer):
+    if isinstance(optimizer, tf.keras.mixed_precision.LossScaleOptimizer):
       grads = optimizer.get_unscaled_gradients(grads)
 
     # Apply gradient clipping.
@@ -213,12 +266,14 @@ class YT8MTask(base_task.Task):
 
     logs = {self.loss: loss}
 
-    all_losses = {'total_loss': loss, 'model_loss': model_loss}
-
-    if metrics:
-      for m in metrics:
-        m.update_state(all_losses[m.name])
-        logs.update({m.name: m.result()})
+    logs.update(
+        self.process_metrics(
+            metrics,
+            labels=labels,
+            outputs=outputs,
+            model_losses=all_losses,
+            label_weights=label_weights,
+            training=True))
 
     return logs
 
@@ -226,11 +281,9 @@ class YT8MTask(base_task.Task):
     """Validatation step.
 
     Args:
-      inputs: a dictionary of input tensors. output_dict = {
-        "video_ids": batch_video_ids,
-        "video_matrix": batch_video_matrix,
-        "labels": batch_labels,
-        "num_frames": batch_frames, }
+      inputs: a dictionary of input tensors. output_dict = { "video_ids":
+        batch_video_ids, "video_matrix": batch_video_matrix, "labels":
+        batch_labels, "num_frames": batch_frames, }
       model: the model, forward definition
       metrics: a nested structure of metrics objects.
 
@@ -239,6 +292,7 @@ class YT8MTask(base_task.Task):
     """
     features, labels = inputs['video_matrix'], inputs['labels']
     num_frames = inputs['num_frames']
+    label_weights = inputs.get('label_weights', None)
 
     # sample random frames (None, 5, 1152) -> (None, 30, 1152)
     sample_frames = self.task_config.validation_data.num_frames
@@ -252,23 +306,28 @@ class YT8MTask(base_task.Task):
     outputs = tf.nest.map_structure(lambda x: tf.cast(x, tf.float32), outputs)
     if self.task_config.validation_data.segment_labels:
       # workaround to ignore the unrated labels.
-      outputs *= inputs['label_weights']
+      outputs *= label_weights
       # remove padding
       outputs = outputs[~tf.reduce_all(labels == -1, axis=1)]
       labels = labels[~tf.reduce_all(labels == -1, axis=1)]
-    loss, model_loss = self.build_losses(
-        model_outputs=outputs, labels=labels, aux_losses=model.losses)
 
-    logs = {self.loss: loss}
+    all_losses = self.build_losses(
+        labels=labels,
+        model_outputs=outputs,
+        label_weights=label_weights,
+        aux_losses=model.losses)
 
-    all_losses = {'total_loss': loss, 'model_loss': model_loss}
+    logs = {self.loss: all_losses['total_loss']}
 
-    logs.update({self.avg_prec_metric.name: (labels, outputs)})
+    logs.update(
+        self.process_metrics(
+            metrics,
+            labels=labels,
+            outputs=outputs,
+            model_losses=all_losses,
+            label_weights=inputs.get('label_weights', None),
+            training=False))
 
-    if metrics:
-      for m in metrics:
-        m.update_state(all_losses[m.name])
-        logs.update({m.name: m.result()})
     return logs
 
   def inference_step(self, inputs, model):