[TF FE] Support Div operation for TensorFlow (#21730)
* Support Div operation for TensorFlow * Update test_tf_Div.py * Update div.cpp * Update op_table and common_op_table * update translate_div_op * print inputs * update div.cpp * set m_pythondiv to false * update div.cpp * update div.cpp * Update tests/layer_tests/tensorflow_tests/test_tf_Div.py * Update tests/layer_tests/tensorflow_tests/test_tf_Div.py --------- Co-authored-by: Roman Kazantsev <roman.kazantsev@intel.com>
This commit is contained in:
sami0i
GitHub
parent
9c373d3335
commit
e2c9f24400
|
|
@ -106,6 +106,7 @@ const std::map<std::string, CreatorFunction> get_supported_ops() {
|
|||
{"BitwiseAnd", CreatorFunction(translate_binary_op<opset13::BitwiseAnd>)},
|
||||
{"BitwiseOr", CreatorFunction(translate_binary_op<opset13::BitwiseOr>)},
|
||||
{"BitwiseXor", CreatorFunction(translate_binary_op<opset13::BitwiseXor>)},
|
||||
{"Div", CreatorFunction(translate_div_op)},
|
||||
{"Equal", CreatorFunction(translate_binary_op<opset8::Equal>)},
|
||||
{"FloorMod", CreatorFunction(translate_binary_op<opset8::FloorMod>)},
|
||||
{"Greater", CreatorFunction(translate_binary_op<opset8::Greater>)},
|
||||
|
|
|
|||
|
|
@ -63,6 +63,7 @@ OP_CONVERTER(translate_crop_and_resize_op);
|
|||
OP_CONVERTER(translate_depth_to_space_op);
|
||||
OP_CONVERTER(translate_depthwise_conv_2d_native_op);
|
||||
OP_CONVERTER(translate_div_no_nan_op);
|
||||
OP_CONVERTER(translate_div_op);
|
||||
OP_CONVERTER(translate_mul_op);
|
||||
OP_CONVERTER(translate_dynamic_partition_op);
|
||||
OP_CONVERTER(translate_einsum_op);
|
||||
|
|
|
|||
|
|
@ -0,0 +1,59 @@
|
|||
// Copyright (C) 2018-2023 Intel Corporation
|
||||
// SPDX-License-Identifier: Apache-2.0
|
||||
//
|
||||
|
||||
#include "common_op_table.hpp"
|
||||
#include "openvino/op/divide.hpp"
|
||||
|
||||
using namespace std;
|
||||
using namespace ov::op;
|
||||
|
||||
namespace ov {
|
||||
namespace frontend {
|
||||
namespace tensorflow {
|
||||
namespace op {
|
||||
OutputVector translate_div_op(const NodeContext& node) {
|
||||
default_op_checks(node, 2, {"Div"});
|
||||
auto x = node.get_input(0);
|
||||
auto y = node.get_input(1);
|
||||
// Check if the element type is a signed integer
|
||||
if (x.get_element_type().is_integral_number() && x.get_element_type().is_signed()) {
|
||||
// prepare auxiliary zero constants of the same type as the inputs
|
||||
auto const_zero = create_same_type_const_scalar<int32_t>(x, 0);
|
||||
|
||||
// compute the modulus of x and y
|
||||
auto mod_result = make_shared<v1::Mod>(x, y);
|
||||
// compute a mask to get positions of non-zero values of mod result
|
||||
auto mod_non_zero = make_shared<v1::NotEqual>(mod_result, const_zero);
|
||||
|
||||
// compute the division of x and y
|
||||
auto divide = make_shared<v1::Divide>(x, y);
|
||||
// compute a mask to get positions of negative values of division result
|
||||
auto div_is_neg = make_shared<v1::Less>(divide, const_zero);
|
||||
|
||||
// compute a boolean mask of elements for non-zero values of Mod result and negative values of Divide result
|
||||
auto mask = make_shared<v1::LogicalAnd>(mod_non_zero, div_is_neg);
|
||||
|
||||
// prepare auxiliary one constants of the same type as the inputs
|
||||
auto const_one = create_same_type_const_scalar<int32_t>(x, 1);
|
||||
// add 1 to the divide result
|
||||
auto add_result = make_shared<v1::Add>(divide, const_one);
|
||||
|
||||
// select division results based on the mask
|
||||
// - perform floor division for non-negative values.
|
||||
// - round negative values to the nearest zero.
|
||||
auto div = make_shared<v1::Select>(mask, add_result, divide);
|
||||
set_node_name(node.get_name(), div);
|
||||
return div->outputs();
|
||||
} else {
|
||||
// for other cases (non-signed-integer types)
|
||||
// compute regular division of x and y
|
||||
auto div = make_shared<v1::Divide>(x, y);
|
||||
set_node_name(node.get_name(), div);
|
||||
return div->outputs();
|
||||
}
|
||||
}
|
||||
} // namespace op
|
||||
} // namespace tensorflow
|
||||
} // namespace frontend
|
||||
} // namespace ov
|
||||
|
|
@ -0,0 +1,50 @@
|
|||
# Copyright (C) 2018-2023 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
|
||||
|
||||
|
||||
class TestDiv(CommonTFLayerTest):
|
||||
def _prepare_input(self, inputs_info):
|
||||
assert 'x' in inputs_info
|
||||
assert 'y' in inputs_info
|
||||
x_shape = inputs_info['x']
|
||||
y_shape = inputs_info['y']
|
||||
inputs_data = {}
|
||||
# generate x and y
|
||||
inputs_data['x'] = np.random.randint(-10, 10, x_shape).astype(self.input_type)
|
||||
inputs_data['y'] = np.random.randint(1, 10, y_shape)*np.random.choice([-1,1], y_shape)
|
||||
return inputs_data
|
||||
|
||||
def create_div_net(self, input_shape, input_type):
|
||||
self.input_type = input_type
|
||||
tf.compat.v1.reset_default_graph()
|
||||
# Create the graph and model
|
||||
with tf.compat.v1.Session() as sess:
|
||||
x = tf.compat.v1.placeholder(input_type, input_shape, 'x')
|
||||
y = tf.compat.v1.placeholder(input_type, input_shape, 'y')
|
||||
tf.raw_ops.Div(x=x, y=y)
|
||||
tf.compat.v1.global_variables_initializer()
|
||||
tf_net = sess.graph_def
|
||||
|
||||
return tf_net, None
|
||||
|
||||
test_data_basic = [
|
||||
dict(input_shape=[10, 20], input_type=np.float32),
|
||||
dict(input_shape=[2, 3, 4], input_type=np.float32),
|
||||
pytest.param(dict(input_shape=[8, 5], input_type=np.int32),
|
||||
marks=pytest.mark.xfail(reason='Ticket TBD - Divide inconsistent behavior on different systems')),
|
||||
dict(input_shape=[], input_type=np.float32),
|
||||
]
|
||||
|
||||
@pytest.mark.parametrize("params", test_data_basic)
|
||||
@pytest.mark.precommit_tf_fe
|
||||
@pytest.mark.nightly
|
||||
def test_div_basic(self, params, ie_device, precision, ir_version, temp_dir,
|
||||
use_new_frontend, use_old_api):
|
||||
self._test(*self.create_div_net(**params),
|
||||
ie_device, precision, ir_version, temp_dir=temp_dir,
|
||||
use_new_frontend=use_new_frontend, use_old_api=use_old_api)
|
||||
Loading…
Reference in New Issue