mindspore/tests/st/scipy_st/test_grad.py

256 lines
9.5 KiB
Python

# Copyright 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.
# ============================================================================
"""st for scipy.ops_grad."""
import pytest
import numpy as onp
import mindspore.nn as nn
import mindspore.ops as ops
from mindspore import context, Tensor
from mindspore.scipy.linalg import cho_factor, cho_solve
from mindspore.scipy.ops import Eigh, Cholesky, SolveTriangular
from tests.st.scipy_st.utils import create_random_rank_matrix, create_sym_pos_matrix, gradient_check
@pytest.mark.level0
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
@pytest.mark.parametrize('shape', [(8, 8)])
@pytest.mark.parametrize('data_type', [(onp.float32, 1e-2, 1e-3), (onp.float64, 1e-4, 1e-7)])
def test_cholesky_grad(shape, data_type):
"""
Feature: ALL TO ALL
Description: test cases for grad implementation of cholesky operator in graph mode and pynative mode.
Expectation: the result match gradient checking.
"""
onp.random.seed(0)
context.set_context(mode=context.GRAPH_MODE)
dtype, epsilon, error = data_type
class CholeskyNet(nn.Cell):
def __init__(self):
super(CholeskyNet, self).__init__()
self.mean = ops.ReduceMean()
# Input arg clean not supports grad right now, just default clean to True.
self.cholesky = Cholesky(clean=True)
def construct(self, a):
c = self.cholesky(a)
return self.mean(c)
cholesky_net = CholeskyNet()
a = create_sym_pos_matrix(shape, dtype)
cholesky_net(Tensor(a))
assert gradient_check(Tensor(a), cholesky_net, epsilon, symmetric=True) < error
context.set_context(mode=context.PYNATIVE_MODE)
cholesky_net(Tensor(a))
assert gradient_check(Tensor(a), cholesky_net, epsilon, symmetric=True) < error
@pytest.mark.level0
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
@pytest.mark.parametrize('lower', [True, False])
@pytest.mark.parametrize('shape', [(8, 8)])
@pytest.mark.parametrize('data_type', [(onp.float32, 1e-2, 1e-3), (onp.float64, 1e-4, 1e-7)])
def test_cho_factor_grad(lower, shape, data_type):
"""
Feature: ALL TO ALL
Description: test cases for grad implementation of cho_factor in graph mode and pynative mode.
Expectation: the result match gradient checking.
"""
onp.random.seed(0)
context.set_context(mode=context.GRAPH_MODE)
dtype, epsilon, error = data_type
class ChoFactorNet(nn.Cell):
def __init__(self, lower):
super(ChoFactorNet, self).__init__()
self.mean = ops.ReduceMean()
self.lower = lower
def construct(self, a):
c, _ = cho_factor(a, self.lower)
return self.mean(c)
def _enumerate_fn(x):
for inner, _ in onp.ndenumerate(x):
if inner[-1] > inner[-2]:
continue
yield inner, _
cho_factor_net = ChoFactorNet(lower)
a = create_sym_pos_matrix(shape, dtype)
assert gradient_check(Tensor(a), cho_factor_net, epsilon, symmetric=True, enumerate_fn=_enumerate_fn) < error
context.set_context(mode=context.PYNATIVE_MODE)
assert gradient_check(Tensor(a), cho_factor_net, epsilon, symmetric=True, enumerate_fn=_enumerate_fn) < error
@pytest.mark.level0
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
@pytest.mark.parametrize('lower', [True, False])
@pytest.mark.parametrize('shape', [(8, 8)])
@pytest.mark.parametrize('data_type', [(onp.float32, 1e-2, 1e-3), (onp.float64, 1e-4, 1e-7)])
def test_cho_solve_grad(lower, shape, data_type):
"""
Feature: ALL TO ALL
Description: test cases for grad implementation of cho_solve in graph mode and pynative mode.
Expectation: the result match gradient checking.
"""
onp.random.seed(0)
context.set_context(mode=context.GRAPH_MODE)
dtype, epsilon, error = data_type
class ChoSolveNet(nn.Cell):
def __init__(self, lower):
super(ChoSolveNet, self).__init__()
self.mean = ops.ReduceMean()
self.lower = lower
def construct(self, c, b):
c_lower = (c, self.lower)
output = cho_solve(c_lower, b)
return self.mean(output)
a = create_sym_pos_matrix(shape, dtype)
n = shape[-1]
b = onp.ones((n, 1), dtype=dtype)
msp_c, msp_lower = cho_factor(Tensor(a), lower)
cho_solve_net = ChoSolveNet(msp_lower)
assert gradient_check([msp_c, Tensor(b)], cho_solve_net, epsilon) < error
context.set_context(mode=context.PYNATIVE_MODE)
assert gradient_check([msp_c, Tensor(b)], cho_solve_net, epsilon) < error
@pytest.mark.level0
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
@pytest.mark.parametrize('compute_eigenvectors', [True, False])
@pytest.mark.parametrize('lower', [True, False])
@pytest.mark.parametrize('shape', [(8, 8)])
@pytest.mark.parametrize('data_type', [(onp.float32, 1e-3, 1e-3), (onp.float64, 1e-4, 1e-7)])
def test_eigh_grad(compute_eigenvectors, lower, shape, data_type):
"""
Feature: ALL TO ALL
Description: test cases for grad implementation of Eigh operator
Expectation: the result match gradient checking.
"""
onp.random.seed(0)
dtype, epsilon, error = data_type
class Net(nn.Cell):
def __init__(self):
super(Net, self).__init__()
self.mean = ops.ReduceMean()
self.sum = ops.ReduceSum()
self.compute_eigenvectors = compute_eigenvectors
self.lower = lower
self.eigh = Eigh(compute_eigenvectors, lower)
def construct(self, a):
res = None
if self.compute_eigenvectors:
w, v = self.eigh(a)
res = self.sum(w) + self.mean(v)
else:
w = self.eigh(a)
res = self.mean(w)
return res
net = Net()
a = create_random_rank_matrix(shape, dtype)
context.set_context(mode=context.GRAPH_MODE)
assert gradient_check(Tensor(a), net, epsilon) < error
context.set_context(mode=context.PYNATIVE_MODE)
assert gradient_check(Tensor(a), net, epsilon) < error
@pytest.mark.level0
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
@pytest.mark.parametrize('shapes', [((8, 8), (8, 8)), ((8, 8), (8, 2)), ((8, 8), (8,))])
@pytest.mark.parametrize('trans', ["N", "T", "C"])
@pytest.mark.parametrize('lower', [False, True])
@pytest.mark.parametrize('unit_diagonal', [True, False])
@pytest.mark.parametrize('data_type', [(onp.float32, 1e-3, 1e-3), (onp.float64, 1e-4, 1e-7)])
def test_trsm_grad_pynative(shapes, trans, lower, unit_diagonal, data_type):
"""
Feature: ALL TO ALL
Description: test cases for grad implementation of SolveTriangular operator in PYNATIVE mode.
Expectation: the result match gradient checking.
"""
a_shape, b_shape = shapes
onp.random.seed(0)
dtype, epsilon, error = data_type
class Net(nn.Cell):
def __init__(self):
super(Net, self).__init__()
self.mean = ops.ReduceMean()
self.sum = ops.ReduceSum()
self.trsm = SolveTriangular(lower, unit_diagonal, trans)
def construct(self, a, b):
x = self.trsm(a, b)
return self.sum(x) + self.mean(x)
net = Net()
a = (onp.random.random(a_shape) + onp.eye(a_shape[-1])).astype(dtype)
b = onp.random.random(b_shape).astype(dtype)
context.set_context(mode=context.PYNATIVE_MODE)
assert gradient_check([Tensor(a), Tensor(b)], net, epsilon) < error
@pytest.mark.level0
@pytest.mark.platform_x86_cpu
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
@pytest.mark.parametrize('shapes', [((8, 8), (8, 8)), ((8, 8), (8, 2)), ((8, 8), (8,))])
@pytest.mark.parametrize('trans', ["N", "T", "C"])
@pytest.mark.parametrize('lower', [False, True])
@pytest.mark.parametrize('unit_diagonal', [True, False])
@pytest.mark.parametrize('data_type', [(onp.float32, 1e-3, 1e-3), (onp.float64, 1e-4, 1e-7)])
def test_trsm_grad_graph(shapes, trans, lower, unit_diagonal, data_type):
"""
Feature: ALL TO ALL
Description: test cases for grad implementation of SolveTriangular operator in GRAPH mode.
Expectation: the result match gradient checking.
"""
a_shape, b_shape = shapes
onp.random.seed(0)
dtype, epsilon, error = data_type
class Net(nn.Cell):
def __init__(self):
super(Net, self).__init__()
self.mean = ops.ReduceMean()
self.sum = ops.ReduceSum()
self.trsm = SolveTriangular(lower, unit_diagonal, trans)
def construct(self, a, b):
x = self.trsm(a, b)
return self.sum(x) + self.mean(x)
net = Net()
a = (onp.random.random(a_shape) + onp.eye(a_shape[-1])).astype(dtype)
b = onp.random.random(b_shape).astype(dtype)
context.set_context(mode=context.GRAPH_MODE)
assert gradient_check([Tensor(a), Tensor(b)], net, epsilon) < error