mindspore/tests/ut/python/dataset/test_pyfunc.py

625 lines
24 KiB
Python

# Copyright 2019-2022 Huawei Technologies Co., Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""
Test Python Function (PyFunc) Support in Dataset
"""
import numpy as np
import pytest
import mindspore.dataset as ds
import mindspore.dataset.engine.iterators as it
from mindspore import log as logger
DATA_DIR = ["../data/dataset/testPyfuncMap/data.data"]
SCHEMA_DIR = "../data/dataset/testPyfuncMap/schema.json"
def test_case_0():
"""
Feature: PyFunc in Map op
Description: Test 1-1 PyFunc : lambda x : x + x
Expectation: Output is equal to the expected output
"""
logger.info("Test 1-1 PyFunc : lambda x : x + x")
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x: x + x), input_columns="col0", output_columns="out")
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out"], golden)
i = i + 4
def test_case_1():
"""
Feature: PyFunc in Map op
Description: Test 1-n PyFunc : lambda x : (x, x + x)
Expectation: Output is equal to the expected output
"""
logger.info("Test 1-n PyFunc : lambda x : (x , x + x) ")
col = "col0"
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x: (x, x + x)), input_columns=col, output_columns=["out0", "out1"])
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i, i + 1], [i + 2, i + 3]])
np.testing.assert_array_equal(item["out0"], golden)
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out1"], golden)
i = i + 4
def test_case_2():
"""
Feature: PyFunc in Map op
Description: Test n-1 PyFunc : lambda x, y : x + y
Expectation: Output is equal to the expected output
"""
logger.info("Test n-1 PyFunc : lambda x, y : x + y ")
col = ["col0", "col1"]
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x, y: x + y), input_columns=col, output_columns="out")
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out"], golden)
i = i + 4
def test_case_3():
"""
Feature: PyFunc in Map op
Description: Test n-m PyFunc : lambda x, y : (x, x + 1, x + y)
Expectation: Output is equal to the expected output
"""
logger.info("Test n-m PyFunc : lambda x, y : (x , x + 1, x + y)")
col = ["col0", "col1"]
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x, y: (x, x + y, x + y + 1)), input_columns=col,
output_columns=["out0", "out1", "out2"])
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i, i + 1], [i + 2, i + 3]])
np.testing.assert_array_equal(item["out0"], golden)
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out1"], golden)
golden = np.array([[i * 2 + 1, (i + 1) * 2 + 1], [(i + 2) * 2 + 1, (i + 3) * 2 + 1]])
np.testing.assert_array_equal(item["out2"], golden)
i = i + 4
def test_case_4():
"""
Feature: PyFunc in Map op
Description: Test parallel n-m PyFunc : lambda x, y : (x, x + 1, x + y)
Expectation: Output is equal to the expected output
"""
logger.info("Test Parallel n-m PyFunc : lambda x, y : (x , x + 1, x + y)")
col = ["col0", "col1"]
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x, y: (x, x + y, x + y + 1)), input_columns=col,
output_columns=["out0", "out1", "out2"], num_parallel_workers=4)
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i, i + 1], [i + 2, i + 3]])
np.testing.assert_array_equal(item["out0"], golden)
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out1"], golden)
golden = np.array([[i * 2 + 1, (i + 1) * 2 + 1], [(i + 2) * 2 + 1, (i + 3) * 2 + 1]])
np.testing.assert_array_equal(item["out2"], golden)
i = i + 4
# The execution of this function will acquire GIL
def func_5(x):
return np.ones(x.shape, dtype=x.dtype)
def test_case_5():
"""
Feature: PyFunc in Map op
Description: Test 1-1 PyFunc : lambda x : np.ones(x.shape)
Expectation: Output is equal to the expected output
"""
logger.info("Test 1-1 PyFunc : lambda x: np.ones(x.shape)")
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=func_5, input_columns="col0", output_columns="out")
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[1, 1], [1, 1]])
np.testing.assert_array_equal(item["out"], golden)
def test_case_6():
"""
Feature: PyFunc in Map op
Description: Test PyFunc Compose : (lambda x : x + x), (lambda x : x + x)
Expectation: Output is equal to the expected output
"""
logger.info("Test PyFunc Compose : (lambda x : x + x), (lambda x : x + x)")
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=[(lambda x: x + x), (lambda x: x + x)], input_columns="col0", output_columns="out")
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i * 4, (i + 1) * 4], [(i + 2) * 4, (i + 3) * 4]])
np.testing.assert_array_equal(item["out"], golden)
i = i + 4
def test_case_7():
"""
Feature: PyFunc in Map op
Description: Test 1-1 PyFunc with python_multiprocessing=True : lambda x : x + x
Expectation: Output is equal to the expected output
"""
logger.info("Test 1-1 PyFunc Multiprocess: lambda x : x + x")
# Reduce memory required by disabling the shared memory optimization
mem_original = ds.config.get_enable_shared_mem()
ds.config.set_enable_shared_mem(False)
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x: x + x), input_columns="col0", output_columns="out",
num_parallel_workers=4, python_multiprocessing=True)
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out"], golden)
i = i + 4
ds.config.set_enable_shared_mem(mem_original)
def test_case_8():
"""
Feature: PyFunc in Map op
Description: Test n-m PyFunc with python_multiprocessing=True : lambda x, y : (x, x + 1, x + y)
Expectation: Output is equal to the expected output
"""
logger.info("Test Multiprocess n-m PyFunc : lambda x, y : (x , x + 1, x + y)")
# Reduce memory required by disabling the shared memory optimization
mem_original = ds.config.get_enable_shared_mem()
ds.config.set_enable_shared_mem(False)
col = ["col0", "col1"]
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=(lambda x, y: (x, x + y, x + y + 1)), input_columns=col,
output_columns=["out0", "out1", "out2"], num_parallel_workers=4,
python_multiprocessing=True)
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i, i + 1], [i + 2, i + 3]])
np.testing.assert_array_equal(item["out0"], golden)
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out1"], golden)
golden = np.array([[i * 2 + 1, (i + 1) * 2 + 1], [(i + 2) * 2 + 1, (i + 3) * 2 + 1]])
np.testing.assert_array_equal(item["out2"], golden)
i = i + 4
ds.config.set_enable_shared_mem(mem_original)
def test_case_9():
"""
Feature: PyFunc in Map op
Description: Test multiple 1-1 PyFunc with python_multiprocessing=True : lambda x : x + x
Expectation: Output is equal to the expected output
"""
logger.info("Test multiple 1-1 PyFunc Multiprocess: lambda x : x + x")
# Reduce memory required by disabling the shared memory optimization
mem_original = ds.config.get_enable_shared_mem()
ds.config.set_enable_shared_mem(False)
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=[(lambda x: x + x), (lambda x: x + 1), (lambda x: x + 2)], input_columns="col0",
output_columns="out", num_parallel_workers=4, python_multiprocessing=True)
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i * 2 + 3, (i + 1) * 2 + 3], [(i + 2) * 2 + 3, (i + 3) * 2 + 3]])
np.testing.assert_array_equal(item["out"], golden)
i = i + 4
ds.config.set_enable_shared_mem(mem_original)
def test_case_10():
"""
Feature: PyFunc in Map op
Description: Test multiple map with python_multiprocessing=True : lambda x : x + x
Expectation: Output is equal to the expected output
"""
logger.info("Test multiple map with multiprocess: lambda x : x + x")
# Reduce memory required by disabling the shared memory optimization
mem_original = ds.config.get_enable_shared_mem()
ds.config.set_enable_shared_mem(False)
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=[(lambda x: x * 10)], input_columns="col0",
output_columns="out", num_parallel_workers=4)
data1 = data1.map(operations=[(lambda x: x + x), (lambda x: x + 1), (lambda x: x + 2)], input_columns="out",
output_columns="out", num_parallel_workers=4, python_multiprocessing=True)
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i * 20 + 3, (i + 1) * 20 + 3], [(i + 2) * 20 + 3, (i + 3) * 20 + 3]])
np.testing.assert_array_equal(item["out"], golden)
i = i + 4
ds.config.set_enable_shared_mem(mem_original)
def test_pyfunc_implicit_compose():
"""
Feature: PyFunc in Map op
Description: Test implicit compose with n-m PyFunc : lambda x, y : (x, x + 1, x + y)
Expectation: Output is equal to the expected output
"""
logger.info("Test n-m PyFunc : lambda x, y : (x , x + 1, x + y)")
col = ["col0", "col1"]
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=[(lambda x, y: (x, x + y, x + y + 1)), (lambda x, y, z: (x, y, z))], input_columns=col,
output_columns=["out0", "out1", "out2"])
i = 0
for item in data1.create_dict_iterator(num_epochs=1, output_numpy=True): # each data is a dictionary
# In this test, the dataset is 2x2 sequential tensors
golden = np.array([[i, i + 1], [i + 2, i + 3]])
np.testing.assert_array_equal(item["out0"], golden)
golden = np.array([[i * 2, (i + 1) * 2], [(i + 2) * 2, (i + 3) * 2]])
np.testing.assert_array_equal(item["out1"], golden)
golden = np.array([[i * 2 + 1, (i + 1) * 2 + 1], [(i + 2) * 2 + 1, (i + 3) * 2 + 1]])
np.testing.assert_array_equal(item["out2"], golden)
i = i + 4
def test_pyfunc_exception():
"""
Feature: PyFunc in Map op
Description: Test PyFunc with exception in child pyfunc process
Expectation: Exception is received and test ends gracefully
"""
logger.info("Test PyFunc Exception Throw: lambda x : raise Exception()")
# Sometimes there are some ITERATORS left in ITERATORS_LIST when run all UTs together,
# and cause core dump and blocking in this UT. Add cleanup() here to fix it.
it._cleanup() # pylint: disable=W0212
def pyfunc():
raise Exception("Pyfunc Throw")
with pytest.raises(RuntimeError) as info:
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=pyfunc, input_columns="col0", output_columns="out",
num_parallel_workers=4)
for _ in data1:
pass
assert "Pyfunc Throw" in str(info.value)
def test_pyfunc_exception_multiprocess():
"""
Feature: PyFunc in Map op
Description: Test python_multiprocessing=True with exception in child pyfunc process
Expectation: Exception is received and test ends gracefully
"""
logger.info("Test Multiprocess PyFunc Exception Throw: lambda x : raise Exception()")
def pyfunc():
raise Exception("MP Pyfunc Throw")
# Reduce memory required by disabling the shared memory optimization
mem_original = ds.config.get_enable_shared_mem()
ds.config.set_enable_shared_mem(False)
with pytest.raises(RuntimeError) as info:
# apply dataset operations
data1 = ds.TFRecordDataset(DATA_DIR, SCHEMA_DIR, shuffle=False)
data1 = data1.map(operations=pyfunc, input_columns="col0", output_columns="out",
num_parallel_workers=4, python_multiprocessing=True)
for _ in data1:
pass
assert "MP Pyfunc Throw" in str(info.value)
ds.config.set_enable_shared_mem(mem_original)
def test_func_with_yield_manifest_dataset_01():
"""
Feature: PyFunc in Map op
Description: Test PyFunc mapping on ManifestDataset
Expectation: Error is raised as expected
"""
def pass_func(_):
for i in range(10):
yield (np.array([i]),)
# Sometimes there are some ITERATORS left in ITERATORS_LIST when run all UTs together,
# and cause core dump and blocking in this UT. Add cleanup() here to fix it.
it._cleanup() # pylint: disable=W0212
manifest_data_file = "../data/dataset/testManifestData/test.manifest"
data = ds.ManifestDataset(manifest_data_file)
data = data.map(operations=pass_func, input_columns=["image"], num_parallel_workers=1, python_multiprocessing=True,
max_rowsize=1)
num_iter = 0
try:
for _ in data.create_dict_iterator(num_epochs=1, output_numpy=True):
num_iter += 1
except RuntimeError as e:
assert " Cannot pickle <class 'generator'> object, please verify pyfunc return with numpy array" in str(e)
def test_func_mixed_with_ops():
"""
Feature: PyFunc in Map op
Description: Test adding computing operator into user defined python function
Expectation: Dataset pipeline has num_parallel_workers decreased/set to 1
"""
logger.info("test_func_mixed_with_ops - PyFunc testing")
def generator_func():
for i in range(1, 5):
yield (np.ones(shape=[2, i]),)
def func(x):
import mindspore.ops as ops
import mindspore
from mindspore import Tensor
flatten = ops.Flatten()
output = flatten(Tensor(x, dtype=mindspore.float32))
return output.asnumpy()
dataset = ds.GeneratorDataset(generator_func, ["data"])
dataset = dataset.map(operations=func, input_columns=["data"])
assert dataset.num_parallel_workers == 1
for _ in dataset.create_dict_iterator(num_epochs=1):
pass
def test_pyfunc_returned_types_basic():
"""
Feature: PyFunc in Map op
Description: Test different returned types from a PyFunc
Expectation: Output is equal to the expected output
"""
logger.info("test_pyfunc_returned_types_basic - PyFunc testing")
def pipeline_testing(returned_data, ms_tensor_dtype):
"""Test stub to validate different returned types from a PyFunc"""
def myfunc(_):
return returned_data
data1 = ds.NumpySlicesDataset([1, 2], num_parallel_workers=1, shuffle=False)
data1 = data1.map(operations=myfunc, num_parallel_workers=2)
# Execute dataset pipeline, request output in numpy format
num_itr = 0
for d in data1.create_tuple_iterator(output_numpy=True, num_epochs=1):
num_itr += 1
logger.info("{}".format(d))
logger.info("{}".format(d[0].dtype))
# Verify NumPy output
np.testing.assert_array_equal(d[0], np.array(returned_data))
assert num_itr == 2
num_itr = 0
# Execute dataset pipeline, request output in non-numpy format
for d in data1.create_tuple_iterator(output_numpy=False, num_epochs=1):
num_itr += 1
logger.info("{}".format(d))
# Verify Non-NumPy MindSpore tensor data type
assert str(d[0].dtype) == ms_tensor_dtype
np.testing.assert_string_equal(str(d[0].dtype), ms_tensor_dtype)
assert num_itr == 2
# Test integer type
pipeline_testing(1, 'Int64')
pipeline_testing((1), 'Int64')
pipeline_testing([-1, 0, 2], 'Int64')
# Test float type
pipeline_testing(1.5, 'Float64')
pipeline_testing(0.0, 'Float64')
pipeline_testing((9.9), 'Float64')
pipeline_testing([-1.5, 3.14], 'Float64')
# Test boolean type
pipeline_testing(True, 'Bool')
pipeline_testing(False, 'Bool')
pipeline_testing((False), 'Bool')
pipeline_testing([False, True, False], 'Bool')
# Test bytes type - expect String type returned
pipeline_testing(b'123', 'String')
pipeline_testing(b'\x00\x01\x02\x01', 'String')
pipeline_testing((b'456'), 'String')
pipeline_testing([b'123', b'456'], 'String')
# Test string type
pipeline_testing('a', 'String')
pipeline_testing('ABC', 'String')
pipeline_testing("YZ", 'String')
pipeline_testing(("MNOP"), 'String')
pipeline_testing(['a', 'ABC', "YZ"], 'String')
@pytest.mark.parametrize("python_multiproc", (False, True))
def test_pyfunc_returned_list_types_mixed(python_multiproc):
"""
Feature: PyFunc in Map op
Description: Test PyFunc which returns list of mixed types
Expectation: Output is equal to the expected output
"""
logger.info("test_pyfunc_returned_list_types_mixed - PyFunc testing")
def test_config_mixed(returned_data, ms_tensor_dtype, python_multiproc, num_workers=2):
"""Test stub to validate returned list with mixed types from a PyFunc"""
def myfunc(_):
return returned_data
if python_multiproc:
# Reduce memory required by disabling the shared memory optimization
mem_original = ds.config.get_enable_shared_mem()
ds.config.set_enable_shared_mem(False)
data1 = ds.NumpySlicesDataset([1, 2, 3, 4, 5, 6, 7, 8], num_parallel_workers=1, shuffle=False)
data1 = data1.map(operations=myfunc, num_parallel_workers=num_workers, python_multiprocessing=python_multiproc)
# Execute dataset pipeline, request output in numpy format
num_itr = 0
for d in data1.create_tuple_iterator(output_numpy=True, num_epochs=1):
num_itr += 1
logger.info("{}".format(d))
logger.info("{}".format(d[0].dtype))
# Verify NumPy output
np.testing.assert_array_equal(d[0], np.array(returned_data))
assert num_itr == 8
num_itr = 0
# Execute dataset pipeline, request output in non-numpy format
for d in data1.create_tuple_iterator(output_numpy=False, num_epochs=1):
num_itr += 1
logger.info("{}".format(d))
# Verify Non-NumPy MindSpore tensor data type
assert str(d[0].dtype) == ms_tensor_dtype
np.testing.assert_string_equal(str(d[0].dtype), ms_tensor_dtype)
assert num_itr == 8
if python_multiproc:
# Restore configuration
ds.config.set_enable_shared_mem(mem_original)
# Test mixed types in list
test_config_mixed([1, True], 'Int64', python_multiproc)
test_config_mixed([False, 0], 'Int64', python_multiproc)
test_config_mixed([1, 2.0], 'Float64', python_multiproc)
test_config_mixed([-3.0, 4], 'Float64', python_multiproc, num_workers=4)
test_config_mixed([True, 20, 30.0], 'Float64', python_multiproc, num_workers=1)
test_config_mixed([5, "five"], 'String', python_multiproc)
test_config_mixed(["five", 5], 'String', python_multiproc)
test_config_mixed([-10.0, 20, True, "four"], 'String', python_multiproc)
def test_pyfunc_returned_types_exception():
"""
Feature: PyFunc in Map op
Description: Test PyFunc with number of column mismatch
Expectation: RuntimeError is detected
"""
logger.info("test_pyfunc_returned_types_exception - PyFunc testing")
def myfunc2(_):
return (1, 2)
def test_config_columns_mismatch(mypyfunc):
"""Test stub to test mismatch in number of columns with PyFunc"""
data1 = ds.NumpySlicesDataset([1, 2, 3], shuffle=False)
data1 = data1.map(operations=mypyfunc)
with pytest.raises(RuntimeError) as error_info:
for _ in data1.create_tuple_iterator(output_numpy=True, num_epochs=1):
pass
assert "Invalid columns, the number of columns returned" in str(error_info.value)
test_config_columns_mismatch(myfunc2)
if __name__ == "__main__":
test_case_0()
test_case_1()
test_case_2()
test_case_3()
test_case_4()
test_case_5()
test_case_6()
test_case_7()
test_case_8()
test_case_9()
test_case_10()
test_pyfunc_implicit_compose()
test_pyfunc_exception()
test_pyfunc_exception_multiprocess()
test_func_with_yield_manifest_dataset_01()
test_func_mixed_with_ops()
test_pyfunc_returned_types_basic()
test_pyfunc_returned_list_types_mixed(python_multiproc=False)
test_pyfunc_returned_types_exception()