mindspore/tests/ut/python/pipeline/infer/test_range.py

# 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

from mindspore import Tensor
from mindspore.common.api import ms_function
from mindspore.ops import operations as P


def test_nest_range_transpose():
    batch_size = 2
    num_layers = 5
    batch_tuple = tuple(Tensor(np.array(np.ones((2, 3)) * 0.01)) for i in range(batch_size))
    layers_tuple = tuple(Tensor(np.array(np.ones((3, 4)) * 0.02)) for i in range(num_layers))
    transpose1 = P.Transpose()

    @ms_function()
    def invoke_range():
        out1 = ()
        for m in range(num_layers):
            out1 += (transpose1(layers_tuple[m], (1, 0)),)
        # Both for loop will the same range symbol as phi node, when range primitive is converted
        # to DoSigature MetaFuncGraph, that MetaFuncGraph will take 2 and 5 as argument, so there is
        # 2 entries in that MetaFuncGraphEvaluator, that will make Specialier try to use AnyValue to
        # FindGeneralized for S-make_range MetaFuncGraph but it will fail as AnyValue is not constant.
        for i in range(batch_size):
            out1 += (transpose1(batch_tuple[i], (1, 0)),)
            for j in range(num_layers):
                out1 += (transpose1(layers_tuple[j], (1, 0)),)
        return out1

    print(invoke_range())


def test_nest_range_simple():
    batch_size = 2
    num_layers = 5
    batch_tuple = tuple(Tensor(np.array(np.ones((2, 3)) * 0.01)) for i in range(batch_size))
    layers_tuple = tuple(Tensor(np.array(np.ones((3, 4)) * 0.02)) for i in range(num_layers))

    @ms_function()
    def invoke_range():
        out1 = ()
        for m in range(num_layers):
            out1 += (layers_tuple[m],)
        for i in range(batch_size):
            out1 += (batch_tuple[i],)
            for j in range(num_layers):
                out1 += (layers_tuple[j],)
        return out1

    print(invoke_range())