Merge branch 'linting2' into linting

official/vision/beta/projects/yolo/modeling/decoders/yolo_decoder.py

@@ -331,13 +331,13 @@ class YoloPAN(tf.keras.layers.Layer):
 
     Args:
       minimum_depth: `int` depth of the smallest branch of the FPN.
-      inputs: `dict[str, tf.InputSpec]` of the shape of input args as a dictionary of
-        lists.
+      inputs: `dict[str, tf.InputSpec]` of the shape of input args as a
+        dictionary of lists.
 
     Returns:
       The unscaled depths of the FPN branches.
     """
-    
+
     depths = []
     if len(inputs.keys()) > 3 or self._fpn_filter_scale > 1:
       for i in range(self._min_level, self._max_level + 1):
@@ -386,8 +386,8 @@ class YoloDecoder(tf.keras.Model):
                kernel_regularizer=None,
                bias_regularizer=None,
                **kwargs):
-    """Yolo Decoder initialization function. A unified model that ties all decoder
-    components into a conditionally build YOLO decder.
+    """Yolo Decoder initialization function. A unified model that ties all
+    decoder components into a conditionally build YOLO decoder.
 
     Args:
       input_specs: `dict[str, tf.InputSpec]`: input specs of each of the inputs
@@ -409,7 +409,7 @@ class YoloDecoder(tf.keras.Model):
         zero.
       kernel_initializer: kernel_initializer for convolutional layers.
       kernel_regularizer: tf.keras.regularizers.Regularizer object for Conv2D.
-      bias_regularizer: tf.keras.regularizers.Regularizer object for Conv2d.
+      bias_regularizer: tf.keras.regularizers.Regularizer object for Conv2D.
       **kwargs: keyword arguments to be passed.
     """
 

official/vision/beta/projects/yolo/modeling/layers/nn_blocks.py

@@ -1152,8 +1152,8 @@ class SAM(tf.keras.layers.Layer):
 
   def call(self, inputs, training=None):
     if self._use_pooling:
-      depth_max = tf.reduce_max(inputs, axis=-1, keep_dims=True)
-      depth_avg = tf.reduce_mean(inputs, axis=-1, keep_dims=True)
+      depth_max = tf.reduce_max(inputs, axis=-1, keepdims=True)
+      depth_avg = tf.reduce_mean(inputs, axis=-1, keepdims=True)
       input_maps = tf.concat([depth_avg, depth_max], axis=-1)
     else:
       input_maps = inputs
@@ -1545,7 +1545,7 @@ class DarkRouteProcess(tf.keras.layers.Layer):
       elif layer == 'spp':
         self.layers.append(self._spp(self._filters, dark_conv_args))
       elif layer == 'sam':
-        self.layers.append(self._sam(-1, _args))
+        self.layers.append(self._sam(-1, dark_conv_args))
 
     self._lim = len(self.layers)
     super().build(input_shape)