[TF FE] Refactor Split and SplitV translators and add layer tests (#15662)

Signed-off-by: Kazantsev, Roman <roman.kazantsev@intel.com>

src/frontends/tensorflow_common/src/op/split.cpp

@@ -14,25 +14,25 @@ namespace tensorflow {
 namespace op {
 
 OutputVector translate_split_op(const NodeContext& node) {
-    TENSORFLOW_OP_VALIDATION(node, node.get_input_size() > 1, "Split must have at least two inputs.");
+    default_op_checks(node, 2, {"Split", "SPLIT"});
     auto axis = node.get_input(0);
-    auto input = node.get_input(1);
+    auto value = node.get_input(1);
     auto num_split = node.get_attribute<int64_t>("num_split");
 
-    auto res = make_shared<Split>(input, axis, num_split);
-    set_node_name(node.get_name(), res);
-    return res->outputs();
+    auto split = make_shared<Split>(value, axis, num_split);
+    set_node_name(node.get_name(), split);
+    return split->outputs();
 }
 
 OutputVector translate_split_v_op(const NodeContext& node) {
-    TENSORFLOW_OP_VALIDATION(node, node.get_input_size() > 2, "Split must have at least three inputs.");
-    auto input = node.get_input(0);
-    auto split_lengths = node.get_input(1);
+    default_op_checks(node, 3, {"SplitV", "SPLIT_V"});
+    auto value = node.get_input(0);
+    auto size_splits = node.get_input(1);
     auto axis = node.get_input(2);
 
-    auto res = make_shared<VariadicSplit>(input, axis, split_lengths);
-    set_node_name(node.get_name(), res);
-    return res->outputs();
+    auto splitv = make_shared<VariadicSplit>(value, axis, size_splits);
+    set_node_name(node.get_name(), splitv);
+    return splitv->outputs();
 }
 }  // namespace op
 }  // namespace tensorflow

tests/layer_tests/tensorflow_tests/test_tf_Split.py

@@ -0,0 +1,37 @@
+# Copyright (C) 2018-2023 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+import tensorflow as tf
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestSplit(CommonTFLayerTest):
+    def create_split_net(self, value_shape, axis_value, num_split):
+        tf.compat.v1.reset_default_graph()
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            axis = tf.constant(axis_value, dtype=tf.int32)
+            value = tf.compat.v1.placeholder(tf.float32, value_shape, 'value')
+            split = tf.raw_ops.Split(axis=axis, value=value, num_split=num_split)
+            for output_ind in range(num_split):
+                tf.identity(split[output_ind], name="split_" + str(output_ind))
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        return tf_net, None
+
+    test_data_basic = [
+        dict(value_shape=[6], axis_value=0, num_split=2),
+        dict(value_shape=[2, 1, 6], axis_value=2, num_split=3),
+        dict(value_shape=[4, 3, 2, 7], axis_value=-4, num_split=4),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_basic)
+    @pytest.mark.precommit_tf_fe
+    @pytest.mark.nightly
+    def test_split_basic(self, params, ie_device, precision, ir_version, temp_dir,
+                         use_new_frontend, use_old_api):
+        self._test(*self.create_split_net(**params),
+                   ie_device, precision, ir_version, temp_dir=temp_dir,
+                   use_new_frontend=use_new_frontend, use_old_api=use_old_api)

tests/layer_tests/tensorflow_tests/test_tf_SplitV.py

@@ -0,0 +1,40 @@
+# Copyright (C) 2018-2023 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+import tensorflow as tf
+from common.tf_layer_test_class import CommonTFLayerTest
+
+
+class TestSplitV(CommonTFLayerTest):
+    def create_splitv_net(self, value_shape, size_splits_values, axis_value):
+        tf.compat.v1.reset_default_graph()
+        # Create the graph and model
+        with tf.compat.v1.Session() as sess:
+            axis = tf.constant(axis_value, dtype=tf.int32)
+            size_splits = tf.constant(size_splits_values, dtype=tf.int32)
+            value = tf.compat.v1.placeholder(tf.float32, value_shape, 'value')
+            num_split = len(size_splits_values)
+            splitv = tf.raw_ops.SplitV(value=value, size_splits=size_splits, axis=axis, num_split=num_split)
+            for output_ind in range(num_split):
+                if size_splits_values[output_ind] != 0:
+                    tf.identity(splitv[output_ind], name="split_" + str(output_ind))
+            tf.compat.v1.global_variables_initializer()
+            tf_net = sess.graph_def
+
+        return tf_net, None
+
+    test_data_basic = [
+        dict(value_shape=[3], size_splits_values=[0, 2, 0, 1], axis_value=0),
+        dict(value_shape=[2, 3, 9], size_splits_values=[1, 2, 3, -1, 1], axis_value=2),
+        dict(value_shape=[3, 9, 5, 4], size_splits_values=[1, 2, 0, -1, 2, 0], axis_value=-3),
+    ]
+
+    @pytest.mark.parametrize("params", test_data_basic)
+    @pytest.mark.precommit_tf_fe
+    @pytest.mark.nightly
+    def test_split_basic(self, params, ie_device, precision, ir_version, temp_dir,
+                         use_new_frontend, use_old_api):
+        self._test(*self.create_splitv_net(**params),
+                   ie_device, precision, ir_version, temp_dir=temp_dir,
+                   use_new_frontend=use_new_frontend, use_old_api=use_old_api)