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

TFLite layer tests second part (#17688)

Browse Source 
* tfl - removes redundant params

* tfl - batch matmul

* tfl - expand_dims

* tfl - squeeze

* tfl - hardswish

* tfl - batch matmul

* tfl - padv2

* tfl - fixes for dynamic shapes

* tfl - where

* tfl - zeros_like

* tfl - zeros_like

* tfl - precommit fix

* tfl - shape and xfail for expand dims
This commit is contained in:
Ruslan Nugmanov 2023-05-29 13:50:34 +02:00 committed by GitHub
parent bd0117d648
commit 28d2e77a92
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
19 changed files with 402 additions and 65 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
tests/layer_tests/common/layer_utils.py
View File

@ -119,6 +119,9 @@ class InferAPI20(BaseInfer):
net = core.read_model(self.model, self.weights)
inputs_info = {}
for item in net.inputs:
inputs_info[item.get_any_name()] = list(item.shape)
if item.partial_shape.is_dynamic:
inputs_info[item.get_any_name()] = item.partial_shape
else:
inputs_info[item.get_any_name()] = item.partial_shape.to_shape()
return inputs_info

8
tests/layer_tests/common/utils/tflite_utils.py
View File

@ -1,5 +1,6 @@
import itertools
import os
import warnings
import numpy as np
import tensorflow as tf
@ -90,6 +91,13 @@ def get_tflite_results(use_new_frontend, use_old_api, inputs_dict, model_path):
for layer, data in inputs_dict.items():
tensor_index = input_name_to_id_mapping[layer]
tensor_id = next(i for i, tensor in enumerate(input_details) if tensor['index'] == tensor_index)
if list(input_details[tensor_id]['shape']) != list(data.shape):
warnings.warn(f'Model and data have different shapes:\nModel {tensor_id} '
f'input shape{input_details[tensor_id]["shape"]}\nInput data shape: {data.shape}')
interpreter.resize_tensor_input(input_details[tensor_id]['index'], data.shape)
interpreter.allocate_tensors()
interpreter.set_tensor(input_details[tensor_id]['index'], data)
interpreter.invoke()

65
tests/layer_tests/tensorflow_lite_tests/test_tfl_BatchMatmul.py Normal file
View File

@ -0,0 +1,65 @@
import pytest
import tensorflow as tf
import numpy as np
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shapes': ((None, 4, 5), (None, 5, 6), (3, 4, 5), (3, 5, 6)), 'adjoint_a': True, 'adjoint_b': True},
{'shapes': ((None, 1, 3, 4), (None, 4, 2), (2, 1, 3, 4), (5, 4, 2)), 'adjoint_a': False, 'adjoint_b': False},
{'shapes': ((None, None, None, 3, 4), (None, None, None, 4, 3),
(2, 2, 2, 3, 4), (2, 2, 2, 4, 3)), 'adjoint_a': False, 'adjoint_b': False},
]
class TestTFLiteBatchMatmulLayerTest(TFLiteLayerTest):
inputs = ["Input", "Input1"]
outputs = ["BatchMatmul"]
allowed_ops = ['BATCH_MATMUL']
def _prepare_input(self, inputs_dict, generator=None):
input0_shape = self.shapes[2]
adj_a = self.adjoint_a
adj_b = self.adjoint_b
if adj_a:
input0_shape = self._swap_last_two_dims(*input0_shape)
inputs_dict['Input'] = np.float32((1.0 - (-1.0)) * np.random.random_sample(input0_shape) + (-1.0))
input1_shape = self.shapes[3] if not adj_b else self._swap_last_two_dims(*self.shapes[3])
inputs_dict['Input1'] = np.float32((1.0 - (-1.0)) * np.random.random_sample(input1_shape) + (-1.0))
return inputs_dict
def _swap_last_two_dims(self, *args):
"""Return a tuple with the last two dimensions swapped."""
return args[:-2] + (args[-1],) + (args[-2],)
def make_model(self, params):
assert len(set(params.keys()).intersection({'shapes', 'adjoint_a', 'adjoint_b'})) == 3, \
'Unexpected parameters for test: ' + ','.join(params.keys())
self.shapes = params['shapes']
self.adjoint_a = params['adjoint_a']
self.adjoint_b = params['adjoint_b']
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
placeholder0_shape = self.shapes[0]
adj_a = params["adjoint_a"]
adj_b = params["adjoint_b"]
if adj_a:
placeholder0_shape = self._swap_last_two_dims(*placeholder0_shape)
input0_tensor = tf.compat.v1.placeholder(dtype=tf.float32, shape=placeholder0_shape, name=self.inputs[0])
if adj_b:
placeholder1_shape = self._swap_last_two_dims(*self.shapes[1])
else:
placeholder1_shape = self.shapes[1]
input1_tensor = tf.compat.v1.placeholder(dtype=tf.float32, shape=placeholder1_shape, name=self.inputs[1])
tf.matmul(input0_tensor, input1_tensor, adjoint_a=adj_a, adjoint_b=adj_b, name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_batch_matmul(self, params, ie_device, precision, temp_dir):
self._test(ie_device, precision, temp_dir, params)

16
tests/layer_tests/tensorflow_lite_tests/test_tfl_Conv2D.py
View File

@ -7,14 +7,10 @@ from common.tflite_layer_test_class import TFLiteLayerTest
np.random.seed(42)
test_params = [
{'shape': [1, 22, 22, 8], 'ksize': [32, 3, 4, 4], 'strides': 2, 'padding': 'SAME', 'data_format': 'NHWC',
'dilations': [1, 1, 1, 1]},
{'shape': [1, 22, 22, 9], 'ksize': [32, 3, 3, 3], 'strides': (2, 1), 'padding': 'SAME', 'data_format': 'NHWC',
'dilations': [1, 2, 2, 1]},
{'shape': [1, 22, 22, 8], 'ksize': [1, 3, 4, 4], 'strides': 2, 'padding': 'VALID', 'data_format': 'NHWC',
'dilations': [1, 1, 1, 1]},
{'shape': [1, 22, 22, 3], 'ksize': [1, 3, 3, 3], 'strides': (3, 4), 'padding': 'VALID', 'data_format': 'NHWC',
'dilations': [1, 2, 2, 1]},
{'shape': [1, 22, 22, 8], 'ksize': [32, 3, 4, 4], 'strides': 2, 'padding': 'SAME', 'dilations': [1, 1, 1, 1]},
{'shape': [1, 22, 22, 9], 'ksize': [32, 3, 3, 3], 'strides': (2, 1), 'padding': 'SAME', 'dilations': [1, 2, 2, 1]},
{'shape': [1, 22, 22, 8], 'ksize': [1, 3, 4, 4], 'strides': 2, 'padding': 'VALID', 'dilations': [1, 1, 1, 1]},
{'shape': [1, 22, 22, 3], 'ksize': [1, 3, 3, 3], 'strides': (3, 4), 'padding': 'VALID', 'dilations': [1, 2, 2, 1]},
]
@ -25,14 +21,14 @@ class TestTFLiteConv2DLayerTest(TFLiteLayerTest):
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'ksize', 'strides',
'padding', 'data_format', 'dilations'})) == 6, \
'padding', 'dilations'})) == 5, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
weights = tf.constant(np.random.randint(-1, 1, params['ksize']), dtype=tf.float32)
place_holder = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape'],
name=self.inputs[0])
tf.nn.conv2d(place_holder, weights, params['strides'], params['padding'], params['data_format'],
tf.nn.conv2d(place_holder, weights, params['strides'], params['padding'], 'NHWC',
params['dilations'], name=self.outputs[0])
net = sess.graph_def
return net

8
tests/layer_tests/tensorflow_lite_tests/test_tfl_DepthToSpace.py
View File

@ -4,8 +4,8 @@ import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': [8, 10, 10, 16], 'block_size': 2, 'data_format': 'NHWC'},
{'shape': [24, 10, 10, 50], 'block_size': 5, 'data_format': 'NHWC'},
{'shape': [8, 10, 10, 16], 'block_size': 2},
{'shape': [24, 10, 10, 50], 'block_size': 5},
]
@ -15,13 +15,13 @@ class TestTFLiteDepthToSpaceLayerTest(TFLiteLayerTest):
allowed_ops = ['DEPTH_TO_SPACE']
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'block_size', 'data_format'})) == 3, \
assert len(set(params.keys()).intersection({'shape', 'block_size'})) == 2, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
place_holder = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape'],
name=self.inputs[0])
tf.nn.depth_to_space(place_holder, params['block_size'], params['data_format'], name=self.outputs[0])
tf.nn.depth_to_space(place_holder, params['block_size'], 'NHWC', name=self.outputs[0])
net = sess.graph_def
return net

8
tests/layer_tests/tensorflow_lite_tests/test_tfl_DepthwiseConv2D.py
View File

@ -7,9 +7,9 @@ from common.tflite_layer_test_class import TFLiteLayerTest
np.random.seed(42)
test_params = [
{'shape': [1, 22, 22, 8], 'ksize': [3, 3, 8, 2], 'strides': [1, 2, 2, 1], 'padding': 'SAME', 'data_format': 'NHWC',
{'shape': [1, 22, 22, 8], 'ksize': [3, 3, 8, 2], 'strides': [1, 2, 2, 1], 'padding': 'SAME',
'dilations': [1, 1]},
{'shape': [1, 22, 22, 9], 'ksize': [3, 3, 9, 1], 'strides': [1, 1, 1, 1], 'padding': 'SAME', 'data_format': 'NHWC',
{'shape': [1, 22, 22, 9], 'ksize': [3, 3, 9, 1], 'strides': [1, 1, 1, 1], 'padding': 'SAME',
'dilations': [1, 1]},
]
@ -21,14 +21,14 @@ class TestTFLiteDepthwiseConv2DLayerTest(TFLiteLayerTest):
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'ksize', 'strides',
'padding', 'data_format', 'dilations'})) == 6, \
'padding', 'dilations'})) == 5, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
weights = tf.constant(np.random.randint(-1, 1, params['ksize']), dtype=tf.float32)
place_holder = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape'],
name=self.inputs[0])
tf.nn.depthwise_conv2d(place_holder, weights, params['strides'], params['padding'], params['data_format'],
tf.nn.depthwise_conv2d(place_holder, weights, params['strides'], params['padding'], 'NHWC',
params['dilations'], name=self.outputs[0])
net = sess.graph_def
return net

46
tests/layer_tests/tensorflow_lite_tests/test_tfl_ExpandDims.py Normal file
View File

@ -0,0 +1,46 @@
import pytest
import tensorflow as tf
import numpy as np
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': [2], 'axis': [0]},
{'shape': [2, 2], 'axis': [2]},
{'shape': [2, 2], 'axis': [0, 2]},
{'shape': [5, 2, 2, 2], 'axis': [1]},
{'shape': [5, 2, 2, 2], 'axis': [1, 1]},
{'shape': [5, 2, 2, 2], 'axis': [-1, -2, -3]},
]
class TestTFLiteExpandDimsLayerTest(TFLiteLayerTest):
inputs = ["Input", "Input1"]
outputs = ["ExpandDims"]
allowed_ops = ['EXPAND_DIMS']
def _prepare_input(self, inputs_dict, generator=None):
inputs_dict['Input'] = (1.0 - (-1.0)) * np.random.random_sample(inputs_dict['Input']) + (-1.0)
inputs_dict['Input1'] = self.axis
return inputs_dict
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'axis'})) == 2, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
self.axis = params['axis']
with tf.compat.v1.Session() as sess:
input_value1 = tf.compat.v1.placeholder(tf.float32, params['shape'], name=self.inputs[0])
axis = tf.compat.v1.placeholder(tf.int32, [len(params['axis'])], name=self.inputs[1])
tf.expand_dims(input_value1, axis, name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_expand_dims(self, params, ie_device, precision, temp_dir):
pytest.xfail("CVS-111983")
self._test(ie_device, precision, temp_dir, params)

13
tests/layer_tests/tensorflow_lite_tests/test_tfl_FullyConnected.py
View File

@ -4,9 +4,9 @@ import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape_x': [40, 37], 'shape_y': [37, 37], 'transpose_a': False, 'transpose_b': True},
{'shape_x': [5, 5], 'shape_y': [4, 5], 'transpose_a': False, 'transpose_b': True},
{'shape_x': [1, 5, 5], 'shape_y': [4, 5], 'transpose_a': False, 'transpose_b': True},
{'shape_x': [40, 37], 'shape_y': [37, 37]},
{'shape_x': [5, 5], 'shape_y': [4, 5]},
{'shape_x': [1, 5, 5], 'shape_y': [4, 5]},
]
@ -16,16 +16,15 @@ class TestTFLiteFullyConnectedLayerTest(TFLiteLayerTest):
allowed_ops = ['FULLY_CONNECTED']
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape_x', 'shape_y',
'transpose_a', 'transpose_b'})) == 4, \
assert len(set(params.keys()).intersection({'shape_x', 'shape_y'})) == 2, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
x = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape_x'], name=self.inputs[0])
y = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape_y'], name=self.inputs[1])
tf.matmul(x, y, transpose_a=params['transpose_a'], transpose_b=params['transpose_b'],
name=self.outputs[0])
tf.matmul(x, y, transpose_a=False, transpose_b=True, name=self.outputs[0])
net = sess.graph_def
return net

40
tests/layer_tests/tensorflow_lite_tests/test_tfl_HardSwish.py Normal file
View File

@ -0,0 +1,40 @@
import pytest
import tensorflow as tf
import numpy as np
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': [1]},
{'shape': [2, 3]},
{'shape': [1, 1, 1, 1]},
{'shape': [1, 3, 4, 3]},
{'shape': [3, 15, 14, 3]},
]
class TestTFLiteHardSwishLayerTest(TFLiteLayerTest):
inputs = ["Input"]
outputs = ["HardSwish"]
allowed_ops = ['HARD_SWISH']
def _prepare_input(self, inputs_dict, generator=None):
inputs_dict['Input'] = np.float32((10 - (-10)) * np.random.random_sample(inputs_dict['Input']) + (-10))
return inputs_dict
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape'})) == 1, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
placeholder = tf.compat.v1.placeholder(tf.float32, params['shape'],
name=self.inputs[0])
hs = placeholder * tf.nn.relu6(placeholder + np.float32(3)) * np.float32(1. / 6.)
tf.identity(hs, name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_hardswish(self, params, ie_device, precision, temp_dir):
self._test(ie_device, precision, temp_dir, params)

50
tests/layer_tests/tensorflow_lite_tests/test_tfl_Pad.py
View File

@ -4,40 +4,58 @@ import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': [1, 1, 2, 1, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0], [1, 0]]},
{'shape': [2, 1, 1, 1, 1], 'paddings': [[0, 1], [0, 0], [0, 0], [2, 3], [1, 0]]},
{'op_name': 'PAD', 'shape': [1, 1, 2, 1, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0], [1, 0]]},
{'op_name': 'PAD', 'shape': [2, 1, 1, 1, 1], 'paddings': [[0, 1], [0, 0], [0, 0], [2, 3], [1, 0]]},
{'shape': [1, 1, 2, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0]]},
{'shape': [1, 1, 2, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0]]},
{'op_name': 'PAD', 'shape': [1, 1, 2, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0]]},
{'op_name': 'PAD', 'shape': [1, 1, 2, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0]]},
{'shape': [1, 2], 'paddings': [[0, 1], [2, 1]]},
{'shape': [1, 2], 'paddings': [[2, 3], [0, 1]]},
{'op_name': 'PAD', 'shape': [1, 2], 'paddings': [[0, 1], [2, 1]]},
{'op_name': 'PAD', 'shape': [1, 2], 'paddings': [[2, 3], [0, 1]]},
{'shape': [1], 'paddings': [[1, 2]]},
{'op_name': 'PAD', 'shape': [1], 'paddings': [[1, 2]]},
{'op_name': 'PADV2', 'shape': [1, 1, 2, 1, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0], [1, 0]],
'constant_value': -1},
{'op_name': 'PADV2', 'shape': [2, 1, 1, 1, 1], 'paddings': [[0, 1], [0, 0], [0, 0], [2, 3], [1, 0]],
'constant_value': 1},
{'op_name': 'PADV2', 'shape': [1, 1, 2, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0]], 'constant_value': -1},
{'op_name': 'PADV2', 'shape': [1, 1, 2, 1], 'paddings': [[0, 0], [0, 1], [2, 3], [0, 0]], 'constant_value': 1},
{'op_name': 'PADV2', 'shape': [1, 2], 'paddings': [[0, 1], [2, 1]], 'constant_value': 1},
{'op_name': 'PADV2', 'shape': [1, 2], 'paddings': [[2, 3], [0, 1]], 'constant_value': -1},
{'op_name': 'PADV2', 'shape': [1], 'paddings': [[1, 2]], 'constant_value': -1},
]
class TestTFLitePadLayerTest(TFLiteLayerTest):
inputs = ["Input", 'Paddings']
outputs = ["Pad"]
allowed_ops = ['PAD']
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'paddings'})) == 2, \
assert len(set(params.keys()).intersection({'op_name', 'shape', 'paddings'})) == 3, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
self.allowed_ops = [params['op_name']]
self.inputs = ["Input"]
with tf.compat.v1.Session() as sess:
place_holder = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape'],
name=self.inputs[0])
shape = [len(params["paddings"]), 2]
paddings = tf.compat.v1.placeholder(dtype=tf.int32, name=self.inputs[1], shape=shape)
tf.pad(tensor=place_holder, paddings=paddings, name=self.outputs[0])
place_holder = tf.compat.v1.placeholder(tf.float32, params['shape'], name=self.inputs[0])
if params['op_name'] == 'PADV2':
tf.pad(tensor=place_holder, paddings=params['paddings'], constant_values=params['constant_value'],
name=self.outputs[0])
else:
self.inputs.append('Paddings')
shape = [len(params["paddings"]), 2]
paddings = tf.compat.v1.placeholder(dtype=tf.int32, name=self.inputs[1], shape=shape)
tf.pad(tensor=place_holder, paddings=paddings, name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_pad(self, params, ie_device, precision, temp_dir):
pytest.xfail("CVS-110828")
if params['op_name'] == 'PAD':
pytest.xfail("CVS-110828")
self._test(ie_device, precision, temp_dir, params)

13
tests/layer_tests/tensorflow_lite_tests/test_tfl_Pool.py
View File

@ -10,11 +10,10 @@ test_ops = [
]
test_params = [
# TFLite doesn't support avgpool with 'NCHW' format
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': 2, 'padding': 'SAME', 'data_format': 'NHWC'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': (2, 1), 'padding': 'SAME', 'data_format': 'NHWC'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': 2, 'padding': 'VALID', 'data_format': 'NHWC'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': (3, 4), 'padding': 'VALID', 'data_format': 'NHWC'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': 2, 'padding': 'SAME'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': (2, 1), 'padding': 'SAME'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': 2, 'padding': 'VALID'},
{'shape': [1, 22, 22, 8], 'ksize': [3, 3], 'strides': (3, 4), 'padding': 'VALID'},
]
test_data = parametrize_tests(test_ops, test_params)
@ -26,7 +25,7 @@ class TestTFLitePoolLayerTest(TFLiteLayerTest):
def make_model(self, params):
assert len(set(params.keys()).intersection({'op_name', 'op_func', 'shape', 'ksize', 'strides',
'padding', 'data_format'})) == 7, \
'padding'})) == 6, \
'Unexpected parameters for test: ' + ','.join(params.keys())
self.allowed_ops = [params['op_name']]
tf.compat.v1.reset_default_graph()
@ -34,7 +33,7 @@ class TestTFLitePoolLayerTest(TFLiteLayerTest):
place_holder = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape'],
name=self.inputs[0])
params['op_func'](place_holder, params['ksize'], params['strides'],
params['padding'], params['data_format'], name=self.outputs[0])
params['padding'], 'NHWC', name=self.outputs[0])
net = sess.graph_def
return net

51
tests/layer_tests/tensorflow_lite_tests/test_tfl_Shape.py Normal file
View File

@ -0,0 +1,51 @@
# Copyright (C) 2018-2023 Intel Corporation
# SPDX-License-Identifier: Apache-2.0
from functools import partial
import numpy as np
import pytest
import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
from common.utils.tflite_utils import parametrize_tests
test_params = [
{'shape': [1, 4], 'new_shape': [1, 4]},
{'shape': [1, 4], 'new_shape': [4, 1]},
{'shape': [1, 4], 'new_shape': [2, 2]},
]
class TestTFLiteShapeLayerTest(TFLiteLayerTest):
inputs = ["Input", "Input1"]
outputs = ["Shape"]
allowed_ops = ['RESHAPE', 'SHAPE']
def _prepare_input(self, inputs_dict, generator=None):
inputs_dict['Input'] = np.random.randint(0, 100, self.shape).astype(np.int32)
inputs_dict['Input1'] = np.array(self.new_shape).astype(np.int32)
return inputs_dict
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'new_shape'})) == 2, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
self.shape = params['shape']
self.new_shape = params['new_shape']
with tf.compat.v1.Session() as sess:
placeholder = tf.compat.v1.placeholder(tf.int32, self.shape, name=self.inputs[0])
shape_of_new_shape = [len(self.new_shape)]
new_shape = tf.compat.v1.placeholder(tf.int32, shape_of_new_shape, name=self.inputs[1])
reshaped = tf.reshape(placeholder, shape=new_shape)
tf.shape(input=reshaped)
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_shape(self, params, ie_device, precision, temp_dir):
self._test(ie_device, precision, temp_dir, params)

8
tests/layer_tests/tensorflow_lite_tests/test_tfl_SpaceToDepth.py
View File

@ -4,8 +4,8 @@ import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': [8, 10, 10, 16], 'block_size': 2, 'data_format': 'NHWC'},
{'shape': [24, 10, 10, 50], 'block_size': 5, 'data_format': 'NHWC'},
{'shape': [8, 10, 10, 16], 'block_size': 2},
{'shape': [24, 10, 10, 50], 'block_size': 5},
]
@ -15,13 +15,13 @@ class TestTFLiteSpaceToDepthLayerTest(TFLiteLayerTest):
allowed_ops = ['SPACE_TO_DEPTH']
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'block_size', 'data_format'})) == 3, \
assert len(set(params.keys()).intersection({'shape', 'block_size'})) == 2, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
place_holder = tf.compat.v1.placeholder(params.get('dtype', tf.float32), params['shape'],
name=self.inputs[0])
tf.nn.space_to_depth(place_holder, params['block_size'], params['data_format'], name=self.outputs[0])
tf.nn.space_to_depth(place_holder, params['block_size'], 'NHWC', name=self.outputs[0])
net = sess.graph_def
return net

36
tests/layer_tests/tensorflow_lite_tests/test_tfl_Squeeze.py Normal file
View File

@ -0,0 +1,36 @@
import pytest
import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': [1, 3], 'axis': [0]},
{'shape': [2, 1], 'axis': [1]},
{'shape': [1, 1, 2], 'axis': [0, 1]},
{'shape': [1, 1, 2, 2], 'axis': [1]},
{'shape': [1, 1, 2, 2], 'axis': [1, 1]},
{'shape': [5, 1, 1, 1], 'axis': [-1, -2, -3]},
]
class TestTFLiteSqueezeLayerTest(TFLiteLayerTest):
inputs = ["Input"]
outputs = ["Squeeze"]
# TFLite returns SQUEEZE only when it has undetermined rank, but OV doesn't support SQUEEZE op with such rank
allowed_ops = ['RESHAPE']
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'axis'})) == 2, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
input_value1 = tf.compat.v1.placeholder(tf.float32, params['shape'], name=self.inputs[0])
tf.squeeze(input_value1, params['axis'], name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_squeeze_dims(self, params, ie_device, precision, temp_dir):
self._test(ie_device, precision, temp_dir, params)

15
tests/layer_tests/tensorflow_lite_tests/test_tfl_TransposeConv.py
View File

@ -8,14 +8,13 @@ np.random.seed(42)
test_params = [
{'shape': [1, 3, 4, 1], 'ksize': [1, 1, 1, 1], 'output_shape': [1, 3, 4, 1], 'strides': [1, 1, 1, 1],
'padding': 'SAME', 'data_format': 'NHWC', 'dilations': [1, 1, 1, 1]},
'padding': 'SAME', 'dilations': [1, 1, 1, 1]},
{'shape': [1, 4, 4, 1], 'ksize': [1, 1, 1, 1], 'output_shape': [1, 4, 4, 1], 'strides': [1, 1], 'padding': 'SAME',
'data_format': 'NHWC', 'dilations': [1, 2, 2, 1]},
#
'dilations': [1, 2, 2, 1]},
{'shape': [1, 22, 22, 3], 'ksize': [1, 1, 6, 3], 'output_shape': [1, 22, 22, 6], 'strides': [1, 1],
'padding': 'VALID', 'data_format': 'NHWC', 'dilations': [1, 1, 1, 1]},
'padding': 'VALID', 'dilations': [1, 1, 1, 1]},
{'shape': [1, 22, 22, 3], 'ksize': [3, 3, 3, 3], 'output_shape': [1, 24, 24, 3], 'strides': [1, 1],
'padding': 'VALID', 'data_format': 'NHWC', 'dilations': [1, 1, 1, 1]},
'padding': 'VALID', 'dilations': [1, 1, 1, 1]},
]
@ -26,7 +25,7 @@ class TestTFLiteTransposeConvLayerTest(TFLiteLayerTest):
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape', 'ksize', 'strides',
'padding', 'data_format', 'dilations', 'output_shape'})) == 7, \
'padding', 'dilations', 'output_shape'})) == 6, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
@ -34,8 +33,8 @@ class TestTFLiteTransposeConvLayerTest(TFLiteLayerTest):
name=self.inputs[0])
filter_input = tf.constant(np.random.randint(-1, 1, size=(params['ksize'])), dtype=tf.float32)
tf.nn.conv2d_transpose(placeholder, filter_input, params['output_shape'], params["strides"],
params["padding"],
params["data_format"], name=self.outputs[0])
params["padding"], 'NHWC', name=self.outputs[0])
net = sess.graph_def
return net

3
tests/layer_tests/tensorflow_lite_tests/test_tfl_Unary.py
View File

@ -41,9 +41,6 @@ test_ops = [
# This op could not be converted standalone -- tries to become FlexOp (offload from tfl to tf)
# {'op_name': 'SIGN', 'op_func': tf.math.sign},
# TF has no such standalone operation
# {'op_name': 'HARD_SWISH'}
]
test_params = [

39
tests/layer_tests/tensorflow_lite_tests/test_tfl_Where.py Normal file
View File

@ -0,0 +1,39 @@
import numpy as np
import pytest
import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
np.random.seed(42)
test_params = [
{'shape': [10]},
{'shape': [1, 2, 3, 4]},
{'shape': [8, 7, 6, 5, 4, 3, 2, 1]}
]
class TestTFLiteWhereLayerTest(TFLiteLayerTest):
inputs = ["Input"]
outputs = ["Where"]
allowed_ops = ['WHERE']
def _prepare_input(self, inputs_dict, generator=None):
inputs_dict['Input'] = np.random.randint(0, 1, inputs_dict['Input']) < 1
return inputs_dict
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape'})) == 1, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
with tf.compat.v1.Session() as sess:
input_value1 = tf.compat.v1.placeholder(tf.bool, params['shape'], name=self.inputs[0])
tf.where(input_value1, name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_where(self, params, ie_device, precision, temp_dir):
self._test(ie_device, precision, temp_dir, params)

41
tests/layer_tests/tensorflow_lite_tests/test_tfl_ZerosLike.py Normal file
View File

@ -0,0 +1,41 @@
import pytest
import tensorflow as tf
from common.tflite_layer_test_class import TFLiteLayerTest
test_params = [
{'shape': ([None], [10])},
{'shape': ([None, 10], [1, 10])},
{'shape': ([1, 10, None], [1, 10, 10])},
{'shape': ([None, 2, 3, 4], [5, 2, 3, 4])},
{'shape': ([5, 2, None, 4], [5, 2, 3, 4])}
]
class TestTFLiteBatchMatmulLayerTest(TFLiteLayerTest):
inputs = ["Input"]
outputs = ["ZerosLike"]
allowed_ops = ['MAXIMUM', 'ZEROS_LIKE']
def _prepare_input(self, inputs_dict, generator=None):
import numpy as np
inputs_dict['Input'] = np.float32(np.random.randint(0, 100, self.shapes_set[1]))
return inputs_dict
def make_model(self, params):
assert len(set(params.keys()).intersection({'shape'})) == 1, \
'Unexpected parameters for test: ' + ','.join(params.keys())
tf.compat.v1.reset_default_graph()
self.shapes_set = params['shape']
with tf.compat.v1.Session() as sess:
placeholder = tf.compat.v1.placeholder(dtype=tf.float32, shape=self.shapes_set[0], name=self.inputs[0])
zeros = tf.zeros_like(placeholder)
tf.maximum(zeros, placeholder, name=self.outputs[0])
net = sess.graph_def
return net
@pytest.mark.parametrize("params", test_params)
@pytest.mark.nightly
def test_zeros_like(self, params, ie_device, precision, temp_dir):
self._test(ie_device, precision, temp_dir, params)

2
tests/layer_tests/tensorflow_tests/test_tf_GatherNd.py
View File

@ -12,7 +12,7 @@ class TestGatherNd(CommonTFLayerTest):
assert 'params' in inputs_info
assert 'indices' in inputs_info
params_shape = inputs_info['params']
indices_shape = inputs_info['indices']
indices_shape = list(inputs_info['indices'])
inputs_data = {}
inputs_data['params'] = np.random.randint(-50, 50, params_shape).astype(self.params_type)
# generate indices for each slice and concatenate