# Copyright 2019 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.

import numpy as np

import mindspore as ms
import mindspore.nn as nn
from mindspore import Tensor
from mindspore import context
from mindspore.common.api import _cell_graph_executor
from mindspore.ops import composite as C
from mindspore.ops import operations as P
from tests.ut.python.ops.test_math_ops import VirtualLoss


grad_all = C.GradOperation(get_all=True)


class NetWithLoss(nn.Cell):
    def __init__(self, network):
        super(NetWithLoss, self).__init__()
        self.loss = VirtualLoss()
        self.network = network

    def construct(self, x, y, b):
        predict = self.network(x, y, b)
        return self.loss(predict)


class GradWrap(nn.Cell):
    def __init__(self, network):
        super(GradWrap, self).__init__()
        self.network = network

    def construct(self, x, y, b):
        return grad_all(self.network)(x, y, b)


def compile_net(net, x, y, b):
    net.set_auto_parallel()
    net.set_train()
    _cell_graph_executor.compile(net, x, y, b)


def test_matmul_equal():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.equal = P.Equal().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.equal(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([128, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([128, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_not_equal():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.notequal = P.NotEqual().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.notequal(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([128, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([128, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_approximateEqual():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.approximateEqual = P.ApproximateEqual(tolerance=0.5).shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.approximateEqual(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0)
    context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_greater():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.greater = P.Greater().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.greater(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0)
    context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_greaterEqual():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.greaterEqual = P.GreaterEqual().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.greaterEqual(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0)
    context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_less():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.less = P.Less().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.less(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0)
    context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_lessEqual():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.lessEqual = P.LessEqual().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.lessEqual(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0)
    context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_not_equal_repeated_calculation():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.notequal = P.NotEqual().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.notequal(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 1), (4, 1))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([128, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([128, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_maximum():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.maximum = P.Maximum().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.maximum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_maximum_broadcast():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.maximum = P.Maximum().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.maximum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (2,))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_maximum_broadcast2():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.maximum = P.Maximum().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.maximum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 4), (4, 1))
    strategy2 = ((4, 1), (1, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 1]), dtype=ms.float32)
    b = Tensor(np.ones([1, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_minimum():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.minimum = P.Minimum().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.minimum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (4, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_minimum_broadcast():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.minimum = P.Maximum().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.minimum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 2), (2, 2))
    strategy2 = ((4, 2), (2,))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 64]), dtype=ms.float32)
    b = Tensor(np.ones([64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_minimum_broadcast2():
    class Net(nn.Cell):
        def __init__(self, strategy1, strategy2):
            super().__init__()
            self.matmul = P.MatMul().shard(strategy1)
            self.minimum = P.Minimum().shard(strategy2)

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.minimum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="semi_auto_parallel")
    strategy1 = ((2, 4), (4, 1))
    strategy2 = ((4, 1), (1, 2))
    net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 1]), dtype=ms.float32)
    b = Tensor(np.ones([1, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)


def test_matmul_minimum_auto_parallel():
    class Net(nn.Cell):
        def __init__(self):
            super().__init__()
            self.matmul = P.MatMul()
            self.minimum = P.Minimum()

        def construct(self, x, y, b):
            out = self.matmul(x, y)
            out = self.minimum(out, b)
            return out

    context.set_auto_parallel_context(device_num=8, global_rank=0, parallel_mode="auto_parallel")
    net = GradWrap(NetWithLoss(Net()))

    x = Tensor(np.ones([64, 32]), dtype=ms.float32)
    y = Tensor(np.ones([32, 1]), dtype=ms.float32)
    b = Tensor(np.ones([1, 64]), dtype=ms.float32)
    compile_net(net, x, y, b)