[TF FE] Refactor Where translator and test it (#13880)
Signed-off-by: Kazantsev, Roman <roman.kazantsev@intel.com> Signed-off-by: Kazantsev, Roman <roman.kazantsev@intel.com>
This commit is contained in:
Roman Kazantsev authored and 
GitHub committed
1 parent
2e9a8cd332
commit
995bcdbe47
2 files changed
+27
-101
No files matched your search
@@ -6,6 +6,7 @@
|
||||
#include "openvino/opsets/opset8.hpp"
|
||||
|
||||
using namespace std;
|
||||
using namespace ov;
|
||||
using namespace ov::opset8;
|
||||
|
||||
namespace ov {
|
||||
@@ -14,9 +15,10 @@ namespace tensorflow {
|
||||
namespace op {
|
||||
|
||||
OutputVector translate_where_op(const NodeContext& node) {
|
||||
auto x = node.get_input(0);
|
||||
auto non_zero = make_shared<NonZero>(x);
|
||||
auto transpose_order = make_shared<Constant>(element::i64, Shape{2}, vector<int64_t>{1, 0});
|
||||
default_op_checks(node, 1, {"Where"});
|
||||
auto condition = node.get_input(0);
|
||||
auto non_zero = make_shared<NonZero>(condition, element::i64);
|
||||
auto transpose_order = make_shared<Constant>(element::i32, Shape{2}, vector<int32_t>{1, 0});
|
||||
auto res = make_shared<opset8::Transpose>(non_zero, transpose_order);
|
||||
set_node_name(node.get_name(), res);
|
||||
return res->outputs();
|
||||
|
||||
@@ -1,119 +1,43 @@
|
||||
# Copyright (C) 2018-2022 Intel Corporation
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
|
||||
import numpy as np
|
||||
import pytest
|
||||
|
||||
import tensorflow as tf
|
||||
from common.tf_layer_test_class import CommonTFLayerTest
|
||||
from common.utils.tf_utils import permute_nchw_to_nhwc
|
||||
|
||||
|
||||
# TODO: refactor and correct this test in the scope of 92881 ticket
|
||||
class TestWhere(CommonTFLayerTest):
|
||||
def create_where_net(self, shape_condition, shape_input, ir_version, use_new_frontend):
|
||||
"""
|
||||
Tensorflow net IR net
|
||||
|
||||
Condition --| Condition --|
|
||||
v v
|
||||
Input_1-> Where Input_1-> Select
|
||||
^ ^
|
||||
Input_2-----| Input_2-----|
|
||||
"""
|
||||
|
||||
import tensorflow as tf
|
||||
def _prepare_input(self, inputs_info):
|
||||
assert 'condition' in inputs_info, "Test error: inputs_info must contain `condition`"
|
||||
condition_shape = inputs_info['condition']
|
||||
inputs_data = {}
|
||||
inputs_data['condition'] = np.random.randint(-2, 2, condition_shape)
|
||||
return inputs_data
|
||||
|
||||
def create_where_net(self, condition_shape, condition_type):
|
||||
tf.compat.v1.reset_default_graph()
|
||||
|
||||
# Create the graph and model
|
||||
with tf.compat.v1.Session() as sess:
|
||||
# Permute shapes NCHW -> NHWC for TF network creation
|
||||
shape_condition_net = permute_nchw_to_nhwc(shape_condition)
|
||||
shape_input_net = permute_nchw_to_nhwc(shape_input)
|
||||
|
||||
condition = tf.compat.v1.placeholder(tf.bool, shape_condition_net, 'Input_condition')
|
||||
input_1 = tf.compat.v1.placeholder(tf.float32, shape_input_net, 'Input_1')
|
||||
input_2 = tf.compat.v1.placeholder(tf.float32, shape_input_net, 'Input_2')
|
||||
|
||||
tf.compat.v1.where(condition, input_1, input_2, name='Operation')
|
||||
|
||||
condition = tf.compat.v1.placeholder(condition_type, condition_shape, 'condition')
|
||||
tf.raw_ops.Where(condition=condition, name='where')
|
||||
tf.compat.v1.global_variables_initializer()
|
||||
|
||||
tf_net = sess.graph_def
|
||||
|
||||
#
|
||||
# Create reference IR net
|
||||
# Please, specify 'type': 'Input' for input node
|
||||
# Moreover, do not forget to validate ALL layer attributes!!!
|
||||
#
|
||||
return tf_net, None
|
||||
|
||||
ref_net = None
|
||||
|
||||
return tf_net, ref_net
|
||||
|
||||
test_data_1D = [dict(shape_condition=[2], shape_input=[2])]
|
||||
|
||||
@pytest.mark.parametrize("params", test_data_1D)
|
||||
@pytest.mark.nightly
|
||||
def test_where_1D(self, params, ie_device, precision, ir_version, temp_dir, use_new_frontend,
|
||||
use_old_api):
|
||||
self._test(*self.create_where_net(**params, ir_version=ir_version,
|
||||
use_new_frontend=use_new_frontend),
|
||||
ie_device, precision, ir_version, temp_dir=temp_dir,
|
||||
use_new_frontend=use_new_frontend, use_old_api=use_old_api)
|
||||
|
||||
test_data_2D = [
|
||||
pytest.param(dict(shape_condition=[2], shape_input=[2, 3]), marks=pytest.mark.precommit_tf_fe),
|
||||
dict(shape_condition=[3, 5], shape_input=[3, 5]),
|
||||
test_data_basic = [
|
||||
dict(condition_shape=[2], condition_type=tf.float32),
|
||||
dict(condition_shape=[3, 4], condition_type=tf.bool),
|
||||
dict(condition_shape=[2, 4, 5], condition_type=tf.int32),
|
||||
]
|
||||
|
||||
@pytest.mark.parametrize("params", test_data_2D)
|
||||
@pytest.mark.parametrize("params", test_data_basic)
|
||||
@pytest.mark.precommit_tf_fe
|
||||
@pytest.mark.nightly
|
||||
def test_where_2D(self, params, ie_device, precision, ir_version, temp_dir, use_new_frontend,
|
||||
use_old_api):
|
||||
self._test(*self.create_where_net(**params, ir_version=ir_version,
|
||||
use_new_frontend=use_new_frontend),
|
||||
ie_device, precision, ir_version, temp_dir=temp_dir,
|
||||
use_new_frontend=use_new_frontend, use_old_api=use_old_api)
|
||||
|
||||
test_data_3D = [
|
||||
dict(shape_condition=[3], shape_input=[3, 4, 5]),
|
||||
dict(shape_condition=[3, 4, 5], shape_input=[3, 4, 5]),
|
||||
]
|
||||
|
||||
@pytest.mark.parametrize("params", test_data_3D)
|
||||
@pytest.mark.nightly
|
||||
def test_where_3D(self, params, ie_device, precision, ir_version, temp_dir, use_new_frontend,
|
||||
use_old_api):
|
||||
self._test(*self.create_where_net(**params, ir_version=ir_version,
|
||||
use_new_frontend=use_new_frontend),
|
||||
ie_device, precision, ir_version, temp_dir=temp_dir,
|
||||
use_new_frontend=use_new_frontend, use_old_api=use_old_api)
|
||||
|
||||
test_data_4D = [
|
||||
dict(shape_condition=[3], shape_input=[3, 4, 5, 6]),
|
||||
dict(shape_condition=[3, 4, 5, 6], shape_input=[3, 4, 5, 6]),
|
||||
]
|
||||
|
||||
@pytest.mark.parametrize("params", test_data_4D)
|
||||
@pytest.mark.nightly
|
||||
@pytest.mark.precommit
|
||||
def test_where_4D(self, params, ie_device, precision, ir_version, temp_dir, use_new_frontend,
|
||||
use_old_api):
|
||||
self._test(*self.create_where_net(**params, ir_version=ir_version,
|
||||
use_new_frontend=use_new_frontend),
|
||||
ie_device, precision, ir_version, temp_dir=temp_dir,
|
||||
use_new_frontend=use_new_frontend, use_old_api=use_old_api)
|
||||
|
||||
test_data_5D = [
|
||||
dict(shape_condition=[3], shape_input=[3, 4, 5, 6, 7]),
|
||||
dict(shape_condition=[3, 4, 5, 6, 7], shape_input=[3, 4, 5, 6, 7]),
|
||||
]
|
||||
|
||||
# TODO mark as precommit (after successfully passing in nightly)
|
||||
@pytest.mark.parametrize("params", test_data_5D)
|
||||
@pytest.mark.nightly
|
||||
def test_where_5D(self, params, ie_device, precision, ir_version, temp_dir, use_new_frontend,
|
||||
use_old_api):
|
||||
self._test(*self.create_where_net(**params, ir_version=ir_version,
|
||||
use_new_frontend=use_new_frontend),
|
||||
def test_where_basic(self, params, ie_device, precision, ir_version, temp_dir,
|
||||
use_new_frontend, use_old_api):
|
||||
self._test(*self.create_where_net(**params),
|
||||
ie_device, precision, ir_version, temp_dir=temp_dir,
|
||||
use_new_frontend=use_new_frontend, use_old_api=use_old_api)
|
||||
Reference in new issue
Block a user