Fixing a bug in linear_model where the name for the model is always set to 'linear_model'. This causes issues when we create multiple linear models in the same graph.

PiperOrigin-RevId: 201270816

2 files changed, 28 insertions, 5 deletions

diff --git a/tensorflow/python/feature_column/feature_column.py b/tensorflow/python/feature_column/feature_column.py
index 5ae60028f4..40219e4b34 100644
--- a/tensorflow/python/feature_column/feature_column.py
+++ b/tensorflow/python/feature_column/feature_column.py
@@ -452,13 +452,15 @@ def linear_model(features,
     ValueError: if an item in `feature_columns` is neither a `_DenseColumn`
       nor `_CategoricalColumn`.
   """
+  with variable_scope.variable_scope(None, 'linear_model') as vs:
+    model_name = _strip_leading_slashes(vs.name)
   linear_model_layer = _LinearModel(
       feature_columns=feature_columns,
       units=units,
       sparse_combiner=sparse_combiner,
       weight_collections=weight_collections,
       trainable=trainable,
-      name='linear_model')
+      name=model_name)
   retval = linear_model_layer(features)  # pylint: disable=not-callable
   if cols_to_vars is not None:
     cols_to_vars.update(linear_model_layer.cols_to_vars())
diff --git a/tensorflow/python/feature_column/feature_column_test.py b/tensorflow/python/feature_column/feature_column_test.py
index c80c1d1866..dc3dde6710 100644
--- a/tensorflow/python/feature_column/feature_column_test.py
+++ b/tensorflow/python/feature_column/feature_column_test.py
@@ -1257,14 +1257,14 @@ class CrossedColumnTest(test.TestCase):
         }, (crossed,))
 
 
-def get_linear_model_bias():
-  with variable_scope.variable_scope('linear_model', reuse=True):
+def get_linear_model_bias(name='linear_model'):
+  with variable_scope.variable_scope(name, reuse=True):
     return variable_scope.get_variable('bias_weights')
 
 
-def get_linear_model_column_var(column):
+def get_linear_model_column_var(column, name='linear_model'):
   return ops.get_collection(ops.GraphKeys.GLOBAL_VARIABLES,
-                            'linear_model/' + column.name)[0]
+                            name + '/' + column.name)[0]
 
 
 def get_keras_linear_model_predictions(features,
@@ -1928,6 +1928,27 @@ class LinearModelTest(test.TestCase):
       with self.assertRaisesOpError('Feature .* cannot have rank 0'):
         sess.run(net, feed_dict={features['price']: np.array(1)})
 
+  def test_multiple_linear_models(self):
+    price = fc.numeric_column('price')
+    with ops.Graph().as_default():
+      features1 = {'price': [[1.], [5.]]}
+      features2 = {'price': [[2.], [10.]]}
+      predictions1 = fc.linear_model(features1, [price])
+      predictions2 = fc.linear_model(features2, [price])
+      bias1 = get_linear_model_bias(name='linear_model')
+      bias2 = get_linear_model_bias(name='linear_model_1')
+      price_var1 = get_linear_model_column_var(price, name='linear_model')
+      price_var2 = get_linear_model_column_var(price, name='linear_model_1')
+      with _initialized_session() as sess:
+        self.assertAllClose([0.], bias1.eval())
+        sess.run(price_var1.assign([[10.]]))
+        sess.run(bias1.assign([5.]))
+        self.assertAllClose([[15.], [55.]], predictions1.eval())
+        self.assertAllClose([0.], bias2.eval())
+        sess.run(price_var2.assign([[10.]]))
+        sess.run(bias2.assign([5.]))
+        self.assertAllClose([[25.], [105.]], predictions2.eval())
+
 
 class _LinearModelTest(test.TestCase):