Fix default logdir for TensorBoard Keras callback on Windows

As cl/235232824, but this one probably hits production users who do not
specify a logdir, because the default argument value is not portable.
For root cause, see:
<#26021>

Tested:
Added a regression test. On Unix, the test passes both before and after
this change. On Windows, the test fails before this change with

    tensorflow.python.framework.errors_impl.NotFoundError: Failed to create a directory: ./logs\train; No such file or directory [Op:CreateSummaryFileWriter]

and passes after this change.

RELNOTES: n/a
PiperOrigin-RevId: 236228230

tensorflow/python/keras/callbacks.py

@@ -1149,7 +1149,7 @@ class TensorBoard(Callback):
   # pylint: enable=line-too-long
 
   def __init__(self,
-               log_dir='./logs',
+               log_dir='logs',
                histogram_freq=0,
                write_graph=True,
                write_images=False,

tensorflow/python/keras/callbacks_test.py

@@ -1048,6 +1048,31 @@ class TestTensorBoardV2(keras_parameterized.TestCase):
     model.compile('sgd', 'mse', run_eagerly=testing_utils.should_run_eagerly())
     return model
 
+  def test_TensorBoard_default_logdir(self):
+    """Regression test for cross-platform pathsep in default logdir."""
+    os.chdir(self.get_temp_dir())
+
+    model = self._get_model()
+    x, y = np.ones((10, 10, 10, 1)), np.ones((10, 1))
+    tb_cbk = keras.callbacks.TensorBoard()  # no logdir specified
+
+    model.fit(
+        x,
+        y,
+        batch_size=2,
+        epochs=2,
+        validation_data=(x, y),
+        callbacks=[tb_cbk])
+
+    summary_file = list_summaries(logdir='.')
+    train_dir = os.path.join('.', 'logs', 'train')
+    validation_dir = os.path.join('.', 'logs', 'validation')
+    self.assertEqual(
+        summary_file.scalars, {
+            _ObservedSummary(logdir=train_dir, tag='epoch_loss'),
+            _ObservedSummary(logdir=validation_dir, tag='epoch_loss'),
+        })
+
   def test_TensorBoard_basic(self):
     model = self._get_model()
     x, y = np.ones((10, 10, 10, 1)), np.ones((10, 1))

tensorflow/tools/api/golden/v2/tensorflow.keras.callbacks.-tensor-board.pbtxt

@@ -5,7 +5,7 @@ tf_class {
   is_instance: "<type \'object\'>"
   member_method {
     name: "__init__"
-    argspec: "args=[\'self\', \'log_dir\', \'histogram_freq\', \'write_graph\', \'write_images\', \'update_freq\', \'profile_batch\'], varargs=None, keywords=kwargs, defaults=[\'./logs\', \'0\', \'True\', \'False\', \'epoch\', \'2\'], "
+    argspec: "args=[\'self\', \'log_dir\', \'histogram_freq\', \'write_graph\', \'write_images\', \'update_freq\', \'profile_batch\'], varargs=None, keywords=kwargs, defaults=[\'logs\', \'0\', \'True\', \'False\', \'epoch\', \'2\'], "
   }
   member_method {
     name: "on_batch_begin"