array_ops_vm_impl.py

# Copyright 2020 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.
# ============================================================================
"""Generate vm_impl function for array ops"""
import numpy as np

import mindspore.common.dtype as mstype
from mindspore.common.tensor import Tensor
from mindspore.ops import operations as P
from mindspore.ops.operations import _grad_ops as G
from mindspore.ops.vm_impl_registry import vm_impl_registry as vm_impl_getters
from .vm_interface import vm


# pylint: disable=unused-argument


@vm_impl_getters.register(P.ExpandDims)
def vm_impl_expand_dims(self):
    """Generate vm_impl function for ExpandDims"""

    def vm_impl(x, axis):
        if isinstance(x, float):
            x = Tensor(np.array([x]))
        x = x.asnumpy()
        out = vm.expand_dims(x, axis)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.DType)
def vm_impl_dType(self):
    """Generate vm_impl function for DType"""

    def vm_impl(x):
        # update the src type
        return x.dtype()

    return vm_impl


@vm_impl_getters.register(P.Cast)
def vm_impl_cast(self):
    """Generate vm_impl function for Cast"""

    def vm_impl(x, t):
        if isinstance(t, type(mstype.tensor)):
            t = t.element_type()
        # update the src type
        x = x.asnumpy()
        out = x.astype(mstype.dtype_to_nptype(t))
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Reshape)
def vm_impl_reshape(self):
    """Generate vm_impl function for Reshape"""

    def vm_impl(x, shp):
        x = x.asnumpy()
        out = vm.reshape(x, shp)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Shape)
def vm_impl_shape(self):
    """Generate vm_impl function for Shape"""

    def vm_impl(x):
        shp = vm.shape(x.asnumpy())
        return shp

    return vm_impl


@vm_impl_getters.register(P.Squeeze)
def vm_impl_squeeze(self):
    """Generate vm_impl function for Squeeze"""

    def vm_impl(x):
        x = x.asnumpy()
        out = vm.squeeze(x, self.axis)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Transpose)
def vm_impl_transpose(self):
    """Generate vm_impl function for Transpose"""

    def vm_impl(x, perm=None):
        x = x.asnumpy()
        if perm is None:
            perm = [i for i in reversed(range(len(x.shape)))]
        out = vm.transpose(x, perm)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Split)
def vm_impl_split(self):
    """Generate vm_impl function for Split"""

    def vm_impl(x):
        x = x.asnumpy()
        output = np.array_split(x, (self.pos,))
        return Tensor(output[0]), Tensor(output[1])

    return vm_impl


@vm_impl_getters.register(P.Fill)
def vm_impl_fill(self):
    """Generate vm_impl function for Fill"""

    def vm_impl(dims, x):
        if isinstance(x, int):
            ret = np.full(dims, x, np.int32)
        else:
            ret = np.full(dims, x, np.float32)
        return Tensor(ret)

    return vm_impl


@vm_impl_getters.register(P.Eye)
def vm_impl_eye(self):
    """Generate vm_impl function for Eye"""

    def vm_impl(n, m, t):
        np_type = mstype.dtype_to_nptype(t)
        ret = np.eye(n, m, dtype=np_type)
        return Tensor(ret)

    return vm_impl


@vm_impl_getters.register(P.InvertPermutation)
def vm_impl_invert_permutation(self):
    """Generate vm_impl function for InvertPermutation"""

    def vm_impl(x):
        out = vm.invert_permutation(x)
        return out

    return vm_impl


@vm_impl_getters.register(P.Argmax)
def vm_impl_argmax(self):
    """Generate vm_impl function for Argmax"""

    def vm_impl(x):
        output = np.argmax(x.asnumpy(), axis=self.axis)
        return Tensor(output.ravel())

    return vm_impl


@vm_impl_getters.register(P.Tile)
def vm_impl_tile(self):
    """Generate vm_impl function for Tile"""

    def vm_impl(x, multiples):
        x = x.asnumpy()
        multiples = multiples.asnumpy()
        out = vm.Tile(x, multiples)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.ReduceAll)
def vm_impl_all(self):
    """Generate vm_impl function for All"""

    def vm_impl(x, axis):
        x = x.asnumpy()
        out = vm.all(x, axis)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Concat)
def vm_impl_concatV2(self):
    """Generate vm_impl function for Concat"""

    def vm_impl(x):
        x = x.asnumpy()
        out = vm.Concat(x, self.axis)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Slice)
def vm_impl_slice(self):
    """Generate vm_impl function for Slice"""

    def vm_impl(x, begin, size):
        x = x.asnumpy()
        begin = begin.asnumpy()
        size = size.asnumpy()
        out = vm.Slice(x, begin, size)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(G.ConcatOffset)
def vm_impl_concatOffset(self):
    """Generate vm_impl function for ConcatOffset"""

    def vm_impl(x):
        out = vm.ConcatOffset(x)  # out is tuple
        return out

    return vm_impl


@vm_impl_getters.register(P.ReduceSum)
def vm_impl_sum(self):
    """Generate vm_impl function for Sum"""

    def vm_impl(x, axis):
        x = x.asnumpy()
        out = vm.sum(x, axis)
        return Tensor(np.array(out))

    return vm_impl


@vm_impl_getters.register(P.Select)
def vm_impl_select(self):
    """Generate vm_impl function for Select"""

    def vm_impl(cond, x, y):
        """
        Args:
            cond: A `Tensor` of type `bool`
            x: A Tensor which may have the same shape as `condition`.
            y: A `Tensor` with the same shape and type as `x`.
        """
        cond = cond.asnumpy()
        x = x.asnumpy()
        y = y.asnumpy()
        out = vm.select(cond, x, y)
        return Tensor(out)

    return vm_impl


@vm_impl_getters.register(P.Square)
def vm_impl_square(self):
    """Generate vm_impl function for Square"""

    def vm_impl(x):
        x = x.asnumpy()
        return Tensor(x * x)

    return vm_impl