batchnorm2d_fuser.py

#   Copyright (c) 2020  PaddlePaddle Authors. All Rights Reserved.
#
# 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 x2paddle.optimizer.pytorch_optimizer.pattern_matcher import FuseBase
from x2paddle.core.program import PaddleGraph, PaddleLayer
from x2paddle.core.util import *


class BatchNorm2dFuser(FuseBase):
    def __init__(self):
        super(BatchNorm2dFuser, self).__init__(graph_type="dygraph")

    def build_pattern(self):
        """ 描述需要替换的batchnorm2d图结构。
        batchnorm2d层模式python实现代码示例:
            x336 = fluid.layers.shape(input=x334)
            x336 = len(x336)
            x337 = x336 != 4
            if x337 :
                raise RaiseException('Exception')
            if False :
                x351 = fluid.layers.shape(input=x334)
                x352 = x351[0]
                x353 = len(x351)
                x354 = x353 - 2
                x357 = x352
                for _x356 in range(x354):
                    x358 = _x356 + 2
                    x359 = x351[x358]
                    x360 = x357 * x359
                    x355 = x360
                x361 = x355 == 1
                if x361 :
                    raise RaiseException('Exception')
            x364 = self.batchnorm7(x334)
        """

        def gen_name(id):
            return "x" + str(id)

        self.pattern.add_layer(
            "fluid.layers.shape",
            inputs={'input': "bn-input-0"},
            outputs=[gen_name(0)])
        self.pattern.add_layer(
            "prim.len", inputs={'input': gen_name(0)}, outputs=[gen_name(0)])
        self.pattern.add_layer(
            "prim.ne", inputs={"x": gen_name(0)}, outputs=[gen_name(1)], y=4)
        self.pattern.add_layer("prim.if", {'input': gen_name(1)}, [gen_name(2)])
        if_layer1 = self.pattern.layers[list(self.pattern.layers.keys())[-1]]
        pattern_block0 = PaddleGraph(if_layer1, graph_type="dygraph")
        pattern_block0.add_layer(
            "prim.exception",
            inputs={},
            outputs=[gen_name(3)],
            input="Exception")
        if_layer1.add_block(pattern_block0)
        pattern_block1 = PaddleGraph(if_layer1, graph_type="dygraph")
        if_layer1.add_block(pattern_block1)
        self.pattern.add_layer("prim.if", {}, [gen_name(4)], input=False)
        if_layer2 = self.pattern.layers[list(self.pattern.layers.keys())[-1]]
        pattern_block0 = PaddleGraph(if_layer2, graph_type="dygraph")
        pattern_block0.add_layer(
            "fluid.layers.shape",
            inputs={'input': "bn-input-0"},
            outputs=[gen_name(5)])
        pattern_block0.add_layer(
            "prim.getitem",
            inputs={"list": gen_name(5)},
            outputs=[gen_name(6)],
            index=0)
        pattern_block0.add_layer(
            "prim.len", inputs={"input": gen_name(5)}, outputs=[gen_name(7)])
        pattern_block0.add_layer(
            "prim.sub", inputs={"x": gen_name(7)}, outputs=[gen_name(8)], y=2)
        pattern_block0.add_layer(
            "prim.equal", inputs={"input": gen_name(6)}, outputs=[gen_name(9)])
        pattern_block0.add_layer(
            "prim.loop",
            inputs={"input": gen_name(8)},
            outputs=[gen_name(8.1), gen_name(10)])
        loop_layer = pattern_block0.layers[list(pattern_block0.layers.keys())[
            -1]]
        pattern_block0_block0 = PaddleGraph(loop_layer, graph_type="dygraph")
        pattern_block0_block0.add_layer(
            "prim.add", inputs={"x": gen_name(10)}, outputs=[gen_name(11)], y=2)
        pattern_block0_block0.add_layer(
            "prim.getitem",
            inputs={"list": gen_name(5),
                    "index": gen_name(11)},
            outputs=[gen_name(12)])
        pattern_block0_block0.add_layer(
            "prim.mul",
            inputs={"x": gen_name(9),
                    "y": gen_name(12)},
            outputs=[gen_name(13)])
        pattern_block0_block0.add_layer(
            "prim.equal",
            inputs={"input": gen_name(13)},
            outputs=[gen_name(8.1)])
        loop_layer.inputs["input-1"] = gen_name(5)
        loop_layer.inputs["input-2"] = gen_name(9)
        loop_layer.add_block(pattern_block0_block0)
        pattern_block0.add_layer(
            "prim.eq", inputs={"x": gen_name(8.1)}, outputs=[gen_name(14)], y=1)
        pattern_block0.add_layer(
            "prim.if", inputs={"input": gen_name(14)}, outputs=[gen_name(15)])
        if_layer21 = pattern_block0.layers[list(pattern_block0.layers.keys())[
            -1]]
        pattern_block0_block0 = PaddleGraph(if_layer21, graph_type="dygraph")
        pattern_block0_block0.add_layer(
            "prim.exception",
            inputs={},
            outputs=[gen_name(15)],
            input="Exception")
        if_layer21.add_block(pattern_block0_block0)
        pattern_block0_block1 = PaddleGraph(if_layer21, graph_type="dygraph")
        if_layer21.add_block(pattern_block0_block1)
        if_layer2.add_block(pattern_block0)
        pattern_block1 = PaddleGraph(if_layer2, graph_type="dygraph")
        if_layer2.add_block(pattern_block1)
        if_layer2.inputs["input-0"] = "bn-input-0"
        self.pattern.add_layer(
            "paddle.nn.BatchNorm",
            inputs={"input": "bn-input-0"},
            outputs=[gen_name(16), gen_name(17)],
            is_test=True,
            num_channels=160,
            momentum=0.1,
            epsilon=0.001)
        self.pattern.build(inputs={"input-0": "bn-input-0"})

    def insert_new_layer(self, graph, parameters, matches):
        new_layer = self.gen_new_layer(parameters, matches)
        new_layer_id = list(matches.keys())[0]
        graph.layers[new_layer_id] = new_layer
        matches.pop(new_layer_id)

#         for layer in matches.values():
#             print(layer.outputs)
#         print("-------")

    def gen_new_layer(self, parameters, matches):
        layers_id = list(matches.keys())
        layer = matches[layers_id[-1]]
        return layer