[eager]: Correctly handle operation arguments of mixed types in the slow path.

Consider the following:

import tensorflow as tf
tf.enable_eager_execution()
x = tf.Variable(1.0)
tf.Print(x, ["foo", x])

Prior to this commit, this snippet would fail with an error:

ValueError: exceptions.TypeError: object of type 'ResourceVariable' has no len()

raised from the call to ops.EagerTensor in convert_to_mixed_eager_tensors.

With this commit, the tf.Print call works correctly.
Note that convert_to_mixed_eager_tensors is only called in the slow path of operation execution (i.e., when TFE_Py_FastPathExecute fails). Which happens rarely (e.g., when mixing primitive string and EagerTensor/ResourceVariable arguments).

PiperOrigin-RevId: 205408407

2 files changed, 9 insertions, 5 deletions

diff --git a/tensorflow/python/eager/core_test.py b/tensorflow/python/eager/core_test.py
index 3fabe7060e..cc765725a4 100644
--- a/tensorflow/python/eager/core_test.py
+++ b/tensorflow/python/eager/core_test.py
@@ -610,6 +610,14 @@ class TFETest(test_util.TensorFlowTestCase):
       self.assertEquals(typ, dtypes.float32)
       self.assertIsInstance(t, ops.EagerTensor)
 
+  def testConvertMixedEagerTensorsWithVariables(self):
+    var = resource_variable_ops.ResourceVariable(1.0)
+    types, tensors = execute_lib.convert_to_mixed_eager_tensors(
+        ['foo', var], context.context())
+    self.assertAllEqual([dtypes.string, dtypes.float32], types)
+    for t in tensors:
+      self.assertIsInstance(t, ops.EagerTensor)
+
 
 class SendRecvTest(test_util.TensorFlowTestCase):
 
diff --git a/tensorflow/python/eager/execute.py b/tensorflow/python/eager/execute.py
index 2ff5b8d8f4..f9b8d2cb5d 100644
--- a/tensorflow/python/eager/execute.py
+++ b/tensorflow/python/eager/execute.py
@@ -198,11 +198,7 @@ def args_to_matching_eager(l, ctx, default_dtype=None):
 
 
 def convert_to_mixed_eager_tensors(values, ctx):
-  v = [
-      t if isinstance(t, ops.EagerTensor) else ops.EagerTensor(
-          t, context=ctx._handle, device=ctx.device_name)  # pylint: disable=protected-access
-      for t in values
-  ]
+  v = [ops.internal_convert_to_tensor(t, ctx=ctx) for t in values]
   types = [t._datatype_enum() for t in v]  # pylint: disable=protected-access
   return types, v