IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[TF FE] Refactor ZerosLike and add layer test (#15648)

Browse Source 
* [TF FE] Refactor ZerosLike and add layer test

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

* Fix test for Wide and Deep model

---------

Signed-off-by: Kazantsev, Roman <roman.kazantsev@intel.com>
This commit is contained in:
Roman Kazantsev
2023-02-12 00:50:01 +04:00
committed by GitHub
parent 1adbdd84f0
commit b300df1be6
3 changed files with 61 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/frontends/tensorflow_common/src/helper_transforms/embedding_segments_feature_fusing.cpp
View File

@@ -91,8 +91,14 @@ ov::frontend::tensorflow::pass::EmbeddingSegmentSingleFeatureFusion::EmbeddingSe
false);
auto tile = make_shared<Tile>(reshape, pack);
auto zeros_like = make_shared<Broadcast>(make_shared<Constant>(ov::element::f32, Shape{1}, std::vector<int64_t>{0}),
make_shared<ShapeOf>(sparse_segment_op));
auto zero_int_const = make_shared<Constant>(element::i32, Shape{1}, 0);
auto one_int_const = make_shared<Constant>(element::i32, Shape{1}, 1);
Output<Node> shape_of = make_shared<ShapeOf>(sparse_segment_op, element::i32);
shape_of = make_shared<Concat>(OutputVector{one_int_const, shape_of}, 0);
Output<Node> zeros_like =
make_shared<Broadcast>(make_shared<Constant>(ov::element::f32, Shape{1}, std::vector<int64_t>{0}), shape_of);
zeros_like = make_shared<Squeeze>(zeros_like, zero_int_const);
// compute number of dimensions to unsqueeze the condition
auto cond_rank = compute_subgraph_scalar_rank(tile, element::i32);

23
src/frontends/tensorflow_common/src/op/zeros_like.cpp
View File

@@ -14,12 +14,25 @@ namespace tensorflow {
namespace op {
OutputVector translate_zeros_like_op(const NodeContext& node) {
default_op_checks(node, 1, {"ZerosLike", "ZEROS_LIKE"});
auto x = node.get_input(0);
auto shape_of = make_shared<ShapeOf>(x);
auto zero = make_shared<Constant>(x.get_element_type(), Shape{1}, 0);
auto res = make_shared<Broadcast>(zero, shape_of);
set_node_name(node.get_name(), res);
return res->outputs();
Output<Node> shape_of = make_shared<ShapeOf>(x, element::i32);
auto zero_const = make_shared<Constant>(x.get_element_type(), Shape{1}, 0);
// in case of x to be scalar, we need handle it more specifically
// since Broadcast supports only broadcasting to rank greater 0
// we have to introduce extra dimension for input scalar case
auto zero_int_const = make_shared<Constant>(element::i32, Shape{1}, 0);
auto one_int_const = make_shared<Constant>(element::i32, Shape{1}, 1);
shape_of = make_shared<Concat>(OutputVector{one_int_const, shape_of}, 0);
// create a tensor of zeros of shape with extra dimension
Output<Node> zeros_like = make_shared<Broadcast>(zero_const, shape_of);
// remove extra dimension by squeezing
zeros_like = make_shared<Squeeze>(zeros_like, zero_int_const);
set_node_name(node.get_name(), zeros_like.get_node_shared_ptr());
return {zeros_like};
}
} // namespace op

35
tests/layer_tests/tensorflow_tests/test_tf_ZerosLike.py Normal file
View File

@@ -0,0 +1,35 @@
# 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 TestZerosLike(CommonTFLayerTest):
def create_zeros_like_net(self, x_shape):
tf.compat.v1.reset_default_graph()
# Create the graph and model
with tf.compat.v1.Session() as sess:
x = tf.compat.v1.placeholder(tf.float32, x_shape, 'x')
tf.raw_ops.ZerosLike(x=x)
tf.compat.v1.global_variables_initializer()
tf_net = sess.graph_def
return tf_net, None
test_data_basic = [
dict(x_shape=[]),
dict(x_shape=[3]),
dict(x_shape=[2, 1, 4]),
dict(x_shape=[2, 4, 3, 1]),
]
@pytest.mark.parametrize("params", test_data_basic)
@pytest.mark.precommit_tf_fe
@pytest.mark.nightly
def test_zeros_like_basic(self, params, ie_device, precision, ir_version, temp_dir,
use_new_frontend, use_old_api):
self._test(*self.create_zeros_like_net(**params),
ie_device, precision, ir_version, temp_dir=temp_dir,
use_new_frontend=use_new_frontend, use_old_api=use_old_api)