tf_op_mapper.py 51.0 KB
Newer Older
J
jiangjiajun 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13
#   Copyright (c) 2019  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.
J
jiangjiajun 已提交
14

J
jiangjiajun 已提交
15 16
from x2paddle.decoder.tf_decoder import TFGraph
from x2paddle.core.op_mapper import OpMapper
J
jiangjiajun 已提交
17
from x2paddle.core.util import *
J
jiangjiajun 已提交
18
import inspect
J
jiangjiajun 已提交
19
import numpy
J
jiangjiajun 已提交
20
import sys
21

J
jiangjiajun 已提交
22

J
jiangjiajun 已提交
23 24 25 26 27 28 29 30
# compute padding size for SAME mode
def get_same_padding(in_size, kernel_size, stride):
    new_size = int(math.ceil(in_size * 1.0 / stride))
    pad_size = (new_size - 1) * stride + kernel_size - in_size
    pad0 = int(pad_size / 2)
    pad1 = pad_size - pad0
    return [pad0, pad1]

J
jiangjiajun 已提交
31

J
jiangjiajun 已提交
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
def nhwc_dim_to_nchw(node, dim):
    tf_data_format = list(node.tf_data_format)
    pd_data_format = list(node.pd_data_format)
    if isinstance(dim, list):
        for i in range(len(dim)):
            char = tf_data_format[dim[i]]
            dim[i] = pd_data_format.index(char)
    else:
        char = tf_data_format[dim]
        dim = pd_data_format.index(char)
    return dim

    if dim < 0:
        dim += 4
    if dim > 0:
        dim = (dim + 1) % 4 + int((dim + 1) / 4)
    return dim


J
jiangjiajun 已提交
51
class TFOpMapper(OpMapper):
J
jiangjiajun 已提交
52 53 54 55 56 57 58
    directly_map_ops = {
        'Relu': ['relu'],
        'Relu6': ['relu6'],
        'Shape': ['shape'],
        'Abs': ['abs'],
        'Sigmoid': ['sigmoid'],
        'Exp': ['exp'],
J
jiangjiajun 已提交
59
        'Rsqrt': ['rsqrt'],
60
        'swish_f32': ['swish'],
J
jiangjiajun 已提交
61
        'Tanh': ['tanh'],
62 63 64
        'LeakyRelu': ['leaky_relu', {
            'alpha': 'alpha'
        }]
J
jiangjiajun 已提交
65 66 67 68 69 70
    }
    elementwise_ops = {
        'Add': 'elementwise_add',
        'RealDiv': 'elementwise_div',
        'Sub': 'elementwise_sub',
        'Maximum': 'elementwise_max',
71 72
        'Mul': 'elementwise_mul',
        'FloorDiv': 'elementwise_floordiv'
J
jiangjiajun 已提交
73 74
    }

J
jiangjiajun 已提交
75 76
    def __init__(self, decoder):
        super(TFOpMapper, self).__init__()
J
jiangjiajun 已提交
77
        self.decoder = decoder
J
jiangjiajun 已提交
78
        self.graph = decoder.tf_graph
79
        self.batch_node = None
J
jiangjiajun 已提交
80
        self.weights = dict()
J
jiangjiajun 已提交
81
        self.omit_nodes = list()
J
jiangjiajun 已提交
82
        self.used_custom_layers = dict()
83

J
jiangjiajun 已提交
84 85 86 87 88 89 90
        not_placeholder = list()
        for name in self.graph.input_nodes:
            if self.graph.get_node(name).layer_type != "Placeholder":
                not_placeholder.append(name)
        for name in not_placeholder:
            idx = self.graph.input_nodes.index(name)
            del self.graph.input_nodes[idx]
J
jiangjiajun 已提交
91

92
        sys.stderr.write("Total nodes: {}\n".format(len(self.graph.topo_sort)))
J
jiangjiajun 已提交
93
        unsupported_ops = set()
94 95
        for i, node_name in enumerate(self.graph.topo_sort):
            sys.stderr.write("\rConverting node {} ...    ".format(i + 1))
96 97
            node = self.graph.get_node(node_name)
            op = node.layer_type
J
jiangjiajun 已提交
98
            if op in self.directly_map_ops:
J
jiangjiajun 已提交
99 100
                if len(unsupported_ops) > 0:
                    continue
J
jiangjiajun 已提交
101 102
                self.directly_map(node)
            elif op in self.elementwise_ops:
J
jiangjiajun 已提交
103 104
                if len(unsupported_ops) > 0:
                    continue
J
jiangjiajun 已提交
105 106
                self.elementwise_map(node)
            elif hasattr(self, op):
J
jiangjiajun 已提交
107 108
                if len(unsupported_ops) > 0:
                    continue
J
jiangjiajun 已提交
109 110
                func = getattr(self, op)
                func(node)
J
jiangjiajun 已提交
111
            else:
J
jiangjiajun 已提交
112 113
                unsupported_ops.add(op)
        if len(unsupported_ops) > 0:
114 115 116
            sys.stderr.write(
                "=========={} Ops are not supported yet======\n".format(
                    len(unsupported_ops)))
J
jiangjiajun 已提交
117
            for op in unsupported_ops:
118
                sys.stderr.write("========== {} ==========\n".format(op))
J
jiangjiajun 已提交
119
            sys.exit(-1)
120
        sys.stderr.write('\nDone!\n')
121

J
jiangjiajun 已提交
122 123 124 125 126 127 128 129 130
    def add_omit_nodes(self, in_node_name, out_node_name):
        in_node = self.graph.get_node(in_node_name)
        out_node = self.graph.get_node(out_node_name)
        index = in_node.outputs.index(out_node_name)
        del in_node.outputs[index]
        index = out_node.inputs.index(in_node_name)
        del out_node.inputs[index]
        self.omit_nodes.append(in_node.layer_name)

J
jiangjiajun 已提交
131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148
    def directly_map(self, node):
        assert node.layer_type in self.directly_map_ops
        op_info = self.directly_map_ops[node.layer_type]
        input = self.graph.get_node(node.layer.input[0], copy=True)
        attr = dict()
        for param in op_info[1:]:
            tf_param_name = list(param.keys())[0]
            pd_param_name = list(param.values())[0]
            tf_param = node.get_attr(tf_param_name)
            attr[pd_param_name] = tf_param
        node.fluid_code.add_layer(op_info[0],
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def elementwise_map(self, node):
        assert node.layer_type in self.elementwise_ops
        op_type = self.elementwise_ops[node.layer_type]
J
jiangjiajun 已提交
149 150 151 152
        x = self.graph.get_node(node.layer.input[0], copy=True)
        y = self.graph.get_node(node.layer.input[1], copy=True)
        x_shape = x.out_shapes[0]
        y_shape = y.out_shapes[0]
153 154 155 156
        if len(x_shape) == 0:
            x_shape = [1]
        if len(y_shape) == 0:
            y_shape = [1]
J
jiangjiajun 已提交
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172
        # incomplement broadcasting support for paddle
        x_input = x
        y_input = y
        if len(x_shape) < len(y_shape):
            unrevertable_ops = [
                "elementwise_sub", "elementwise_div", "elementwise_floordiv",
                "elementwise_mod", "elementwise_pow"
            ]
            if op_type not in unrevertable_ops:
                x_input = y
                y_input = x
                x_shape = y.out_shapes[0]
                y_shape = x.out_shapes[0]
            else:
                raise Exception("Unexpected situation happend")

J
jiangjiajun 已提交
173 174 175 176 177 178 179 180 181 182 183 184 185
        if len(x_shape) == 4 and len(y_shape) == 1:
            if x_input.tf_data_format == "NHWC":
                axis = 1
            else:
                axis = -1
            attr = {"axis": axis}
            inputs = {"x": x_input, "y": y_input}
            node.fluid_code.add_layer(op_type,
                                      inputs=inputs,
                                      output=node,
                                      param_attr=attr)
            return

J
jiangjiajun 已提交
186 187 188 189 190 191
        is_sub_seq = True
        for i in range(len(y_shape)):
            index = -1 * i - 1
            if y_shape[index] != x_shape[index]:
                is_sub_seq = False
        if not is_sub_seq:
J
jiangjiajun 已提交
192 193 194 195
            if x_shape.count(-1) > 2:
                x_shape = self.decoder.infer_tensor_shape(x_input)
            if y_shape.count(-1) > 2:
                y_shape = self.decoder.infer_tensor_shape(y_input)
J
jiangjiajun 已提交
196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
            x_expand_times = [1] * len(x_shape)
            y_expand_times = [1] * len(y_shape)
            x_need_expand = False
            y_need_expand = False
            for i in range(len(y_shape)):
                index = -1 * i - 1
                if y_shape[index] != x_shape[index]:
                    if y_shape[index] == 1:
                        y_expand_times[index] = x_shape[index]
                        y_need_expand = True
                    elif x_shape[index] == 1:
                        x_expand_times[index] = y_shape[index]
                        x_need_expand = True
                    else:
                        raise Exception("Unexpected situation happend")
            if x_need_expand:
J
jiangjiajun 已提交
212 213 214 215
                if len(x_expand_times) == 3 and x.tf_data_format == "NHWC":
                    x_expand_times = [x_expand_times[i] for i in [2, 0, 1]]
                if len(x_expand_times) == 4 and x.tf_data_format == "NHWC":
                    x_expand_times = [x_expand_times[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
216 217 218 219 220 221 222
                attr = {"expand_times": x_expand_times}
                node.fluid_code.add_layer("expand",
                                          inputs=x_input,
                                          output="x_tmp",
                                          param_attr=attr)
                x_input = "x_tmp"
            if y_need_expand:
J
jiangjiajun 已提交
223 224 225 226
                if len(y_expand_times) == 3 and y.tf_data_format == "NHWC":
                    y_expand_times = [y_expand_times[i] for i in [2, 0, 1]]
                if len(y_expand_times) == 4 and y.tf_data_format == "NHWC":
                    y_expand_times = [y_expand_times[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
227 228 229 230 231 232 233 234 235 236 237 238
                attr = {"expand_times": y_expand_times}
                node.fluid_code.add_layer("expand",
                                          inputs=y_input,
                                          output="y_tmp",
                                          param_attr=attr)
                y_input = "y_tmp"
        inputs = {"x": x_input, "y": y_input}
        node.fluid_code.add_layer(op_type,
                                  inputs=inputs,
                                  output=node,
                                  param_attr=None)

239 240
    def Placeholder(self, node):
        shape = node.out_shapes[0]
J
jiangjiajun 已提交
241 242
        assert len(shape) != 0, "Unknown shape of input nodes[{}].".format(
            node.layer_name)
J
jiangjiajun 已提交
243 244 245 246
        if node.tf_data_format == "NHWC" and len(shape) == 4:
            shape = [shape[i] for i in [0, 3, 1, 2]]
        elif node.tf_data_format == "NCHW" and len(shape) == 4:
            self.graph.data_format_propagation(node)
247 248
        dtype = node.dtype
        attr = {
J
jiangjiajun 已提交
249
            'dtype': string(dtype),
250
            'shape': shape,
J
jiangjiajun 已提交
251 252
            'name': string(node.layer_name),
            'append_batch_size': False
253
        }
254 255 256
        if shape[0] < 0:
            self.batch_node = node

J
jiangjiajun 已提交
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271
        node.fluid_code.add_layer("data",
                                  inputs=None,
                                  output=node,
                                  param_attr=attr)

    def Const(self, node):
        shape = node.out_shapes[0]
        dtype = node.dtype
        value = node.value
        initializer = "Constant(0.0)"
        if len(shape) == 0:
            assert value.size == 1, "Unexpected situation happend"
            shape = [1]
            initializer = "Constant({})".format(value)

J
jiangjiajun 已提交
272 273 274 275 276 277 278 279 280 281 282 283 284
        self.weights[node.layer_name] = node.value

        if node.tf_data_format == "NHWC":
            if len(shape) == 4:
                shape = [shape[i] for i in [0, 3, 1, 2]]
            if len(shape) == 3:
                shape = [shape[i] for i in [2, 0, 1]]
                self.weights[node.layer_name] = numpy.transpose(
                    node.value, (2, 0, 1))
        elif node.tf_data_format == "NCHW":
            if len(shape) == 4:
                self.graph.data_format_propagation(node)

J
jiangjiajun 已提交
285 286 287 288 289 290 291 292 293 294 295 296
        attr = {
            'dtype': string(dtype),
            'shape': shape,
            'name': string(node.layer_name),
            'default_initializer': initializer
        }
        node.fluid_code.add_layer("create_parameter",
                                  inputs=None,
                                  output=node,
                                  param_attr=attr)

    def Transpose(self, node):
J
jiangjiajun 已提交
297 298
        input = self.graph.get_node(node.layer.input[0], copy=True)
        perm = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
299
        assert perm.layer_type == "Const", "Perm of transpose OP should be Const"
300
        del self.weights[perm.layer_name.replace('/', '_')]
J
jiangjiajun 已提交
301 302 303
        perm.fluid_code.clear()
        perm = perm.value.tolist()

J
jiangjiajun 已提交
304
        if perm == [0, 3, 1, 2] and input.data_format == "NHWC":
305 306 307 308 309
            input_name = input.layer_name
            if hasattr(input, "index"):
                input_name = input_name + "[{}]".format(input.index)
            node.fluid_code.add_layer("{} = {}").format(node.layer_name,
                                                        input_name)
J
jiangjiajun 已提交
310 311 312
            node.tf_data_format = "NCHW"
            self.graph.data_format_propagation(node)
        elif perm == [0, 2, 3, 1] and input.tf_data_format == "NCHW":
313 314 315 316 317
            input_name = input.layer_name
            if hasattr(input, "index"):
                input_name = input_name + "[{}]".format(input.index)
            node.fluid_code.add_layer("{} = {}").format(node.layer_name,
                                                        input_name)
J
jiangjiajun 已提交
318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
            node.tf_data_format = "NHWC"
            self.graph.data_format_propagation(node)
        elif len(input.out_shapes[0]) > 4:
            tf_data_format = list(input.tf_data_format)
            pd_data_format = list(input.pd_data_format)
            new_perm = [i for i in range(len(perm))]
            for i in range(len(perm)):
                char0 = tf_data_format[i]
                char1 = tf_data_format[perm[i]]
                index0 = pd_data_format.index(char0)
                index1 = pd_data_format.index(char1)
                new_perm[index0] = index1
            node.tf_data_format = [tf_data_format[i] for i in perm]
            node.pd_data_format = [pd_data_format[i] for i in perm]
            attr = {'perm': new_perm}
            node.fluid_code.add_layer("transpose",
                                      inputs=input,
                                      output=node,
                                      param_attr=attr)
        elif len(node.out_shapes[0]) != 4:
            attr = {'perm': perm}
            node.fluid_code.add_layer("transpose",
                                      inputs=input,
                                      output=node,
                                      param_attr=attr)
        else:
            raise Exception("Unexpected situation happend in Transpose OP")
J
jiangjiajun 已提交
345

J
jiangjiajun 已提交
346 347
    def MaxPool(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
J
jiangjiajun 已提交
348

J
jiangjiajun 已提交
349
        in_shape = input.out_shapes[0]
J
jiangjiajun 已提交
350 351 352
        if in_shape.count(-1) > 2:
            in_shape = self.decoder.infer_tensor(input).shape

J
jiangjiajun 已提交
353 354 355 356
        k_size = node.get_attr("ksize")
        strides = node.get_attr("strides")
        data_format = node.get_attr("data_format").decode()
        pad_mode = node.get_attr("padding").decode()
J
jiangjiajun 已提交
357
        channel_first = data_format == "NCHW"
J
jiangjiajun 已提交
358
        padding = 0
J
jiangjiajun 已提交
359

J
jiangjiajun 已提交
360
        if not channel_first:
J
jiangjiajun 已提交
361 362
            in_shape = [in_shape[i] for i in [0, 3, 1, 2]]
            strides = [strides[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
363
            k_size = [k_size[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
364 365
        else:
            self.graph.data_format_propagation(node)
J
jiangjiajun 已提交
366 367

        if pad_mode == "SAME":
J
jiangjiajun 已提交
368 369
            pad_h = get_same_padding(in_shape[2], k_size[2], strides[2])
            pad_w = get_same_padding(in_shape[3], k_size[3], strides[3])
J
jiangjiajun 已提交
370 371 372
            pad_h = pad_h[0] + pad_h[1]
            pad_w = pad_w[0] + pad_w[1]
            attr = {"paddings": [0, pad_h, 0, pad_w], "pad_value": -10000.0}
J
jiangjiajun 已提交
373 374 375 376 377
            node.fluid_code.add_layer("pad2d",
                                      inputs=input,
                                      output=node,
                                      param_attr=attr)
            input = node
J
jiangjiajun 已提交
378
        attr = {
J
jiangjiajun 已提交
379
            "pool_size": k_size[2:4],
J
jiangjiajun 已提交
380
            "pool_type": string("max"),
J
jiangjiajun 已提交
381
            "pool_padding": padding,
J
jiangjiajun 已提交
382
            "pool_stride": strides[2:4]
J
jiangjiajun 已提交
383
        }
J
jiangjiajun 已提交
384 385 386 387
        node.fluid_code.add_layer("pool2d",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
388 389 390 391 392

    def Conv2D(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        kernel = self.graph.get_node(node.layer.input[1], copy=True)
        assert kernel.layer_type == "Const", "Kernel of Conv2D should be Const"
J
jiangjiajun 已提交
393
        self.add_omit_nodes(kernel.layer_name, node.layer_name)
J
jiangjiajun 已提交
394

J
jiangjiajun 已提交
395
        in_shape = input.out_shapes[0]
J
jiangjiajun 已提交
396 397
        if in_shape.count(-1) > 2:
            in_shape = self.decoder.infer_tensor(input).shape
J
jiangjiajun 已提交
398
        k_size = kernel.out_shapes[0]
J
jiangjiajun 已提交
399 400 401
        if k_size.count(-1) > 2:
            k_size = self.decoder.infer_tensor(kernel).shape

J
jiangjiajun 已提交
402 403 404 405 406
        strides = node.get_attr("strides")
        dilations = node.get_attr("dilations")
        data_format = node.get_attr("data_format").decode()
        pad_mode = node.get_attr("padding").decode()
        channel_first = data_format == "NCHW"
J
jiangjiajun 已提交
407 408 409 410
        padding = 0

        self.weights[kernel.layer_name.replace('/', '_')] = numpy.transpose(
            kernel.value, (3, 2, 0, 1))
J
jiangjiajun 已提交
411 412 413 414 415

        if not channel_first:
            in_shape = [in_shape[i] for i in [0, 3, 1, 2]]
            strides = [strides[i] for i in [0, 3, 1, 2]]
            dilations = [dilations[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
416 417
        else:
            self.graph.data_format_propagation(node)
J
jiangjiajun 已提交
418

J
jiangjiajun 已提交
419 420 421
        if pad_mode == "SAME":
            pad_h = get_same_padding(in_shape[2], k_size[0], strides[2])
            pad_w = get_same_padding(in_shape[3], k_size[1], strides[3])
J
jiangjiajun 已提交
422 423 424 425 426 427 428 429 430
            if pad_h[0] == pad_h[1] and pad_w[0] == pad_w[1]:
                padding = [pad_h[0], pad_w[0]]
            else:
                attr = {"paddings": pad_h + pad_w, "pad_value": 0.0}
                node.fluid_code.add_layer("pad2d",
                                          inputs=input,
                                          output=node,
                                          param_attr=attr)
                input = node
J
jiangjiajun 已提交
431 432 433 434 435 436
        attr = {
            "bias_attr": False,
            "param_attr": string(kernel.layer_name),
            "num_filters": k_size[3],
            "filter_size": k_size[0:2],
            "stride": strides[2:4],
J
jiangjiajun 已提交
437 438
            "dilation": dilations[2:4],
            "padding": padding
J
jiangjiajun 已提交
439
        }
J
jiangjiajun 已提交
440 441 442 443
        node.fluid_code.add_layer("conv2d",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
444

J
jiangjiajun 已提交
445 446 447 448 449 450 451 452 453 454 455 456
    def BiasAdd(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        bias = self.graph.get_node(node.layer.input[1], copy=True)
        axis = -1
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            axis = 1
        inputs = {"x": input, "y": bias}
        attr = {"axis": axis}
        node.fluid_code.add_layer("elementwise_add",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
457 458 459 460 461 462 463

    def FusedBatchNorm(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        gamma = self.graph.get_node(node.layer.input[1], copy=True)
        beta = self.graph.get_node(node.layer.input[2], copy=True)
        moving_mean = self.graph.get_node(node.layer.input[3], copy=True)
        moving_var = self.graph.get_node(node.layer.input[4], copy=True)
J
jiangjiajun 已提交
464 465
        data_format = node.get_attr("data_format").decode()
        channel_first = data_format == "NCHW"
J
jiangjiajun 已提交
466 467 468 469 470

        assert gamma.layer_type == "Const"
        assert beta.layer_type == "Const"
        assert moving_mean.layer_type == "Const"
        assert moving_var.layer_type == "Const"
J
jiangjiajun 已提交
471 472 473 474
        self.add_omit_nodes(gamma.layer_name, node.layer_name)
        self.add_omit_nodes(beta.layer_name, node.layer_name)
        self.add_omit_nodes(moving_mean.layer_name, node.layer_name)
        self.add_omit_nodes(moving_var.layer_name, node.layer_name)
J
jiangjiajun 已提交
475 476
        if channel_first:
            self.data_format_propagation(node)
J
jiangjiajun 已提交
477

J
jiangjiajun 已提交
478 479 480 481 482 483 484 485 486 487
        attr = {
            "epsilon": node.get_attr("epsilon"),
            "param_attr": string(gamma.layer_name),
            "bias_attr": string(beta.layer_name),
            "moving_mean_name": string(moving_mean.layer_name),
            "moving_variance_name": string(moving_var.layer_name),
            "is_test": True
        }

        node.fluid_code.add_layer("batch_norm",
J
jiangjiajun 已提交
488
                                  inputs=input,
J
jiangjiajun 已提交
489 490 491 492 493 494 495
                                  output=node,
                                  param_attr=attr)

    def DepthwiseConv2dNative(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        kernel = self.graph.get_node(node.layer.input[1], copy=True)
        assert kernel.layer_type == "Const", "Kernel of DepthwiseConv2DNative should be Const"
J
jiangjiajun 已提交
496
        self.add_omit_nodes(kernel.layer_name, node.layer_name)
J
jiangjiajun 已提交
497

J
jiangjiajun 已提交
498
        in_shape = input.out_shapes[0]
J
jiangjiajun 已提交
499 500
        if in_shape.count(-1) > 2:
            in_shape = self.decoder.infer_tensor(input).shape
J
jiangjiajun 已提交
501
        k_size = kernel.out_shapes[0]
J
jiangjiajun 已提交
502 503 504
        if k_size.count(-1) > 2:
            k_size = self.decoder.infer_tensor(kernel).shape

J
jiangjiajun 已提交
505 506 507 508 509
        strides = node.get_attr("strides")
        dilations = node.get_attr("dilations")
        data_format = node.get_attr("data_format").decode()
        pad_mode = node.get_attr("padding").decode()
        channel_first = data_format == "NCHW"
J
jiangjiajun 已提交
510 511 512 513
        padding = 0

        self.weights[kernel.layer_name.replace('/', '_')] = numpy.transpose(
            kernel.value, (2, 3, 0, 1))
J
jiangjiajun 已提交
514 515 516 517 518

        if not channel_first:
            in_shape = [in_shape[i] for i in [0, 3, 1, 2]]
            strides = [strides[i] for i in [0, 3, 1, 2]]
            dilations = [dilations[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
519 520
        else:
            self.data_format_propagation(node)
J
jiangjiajun 已提交
521 522 523 524

        if pad_mode == "SAME":
            pad_h = get_same_padding(in_shape[2], k_size[0], strides[2])
            pad_w = get_same_padding(in_shape[3], k_size[1], strides[3])
J
jiangjiajun 已提交
525 526 527 528
            if pad_h[0] == pad_h[1] and pad_w[0] == pad_w[1]:
                padding = [pad_h[0], pad_w[0]]
            else:
                attr = {"paddings": pad_h + pad_w, "pad_value": 0.0}
J
jiangjiajun 已提交
529
                node.fluid_code.add_layer("pad2d",
J
jiangjiajun 已提交
530
                                          inputs=input,
J
jiangjiajun 已提交
531 532
                                          output=node,
                                          param_attr=attr)
J
jiangjiajun 已提交
533 534
                input = node

J
jiangjiajun 已提交
535 536 537 538 539 540 541
        attr = {
            "bias_attr": False,
            "param_attr": string(kernel.layer_name),
            "num_filters": in_shape[1],
            "filter_size": k_size[0:2],
            "stride": strides[2:4],
            "dilation": dilations[2:4],
J
jiangjiajun 已提交
542
            "groups": k_size[3] * in_shape[1],
J
jiangjiajun 已提交
543
            "use_cudnn": False,
J
jiangjiajun 已提交
544
            "padding": padding
J
jiangjiajun 已提交
545
        }
J
jiangjiajun 已提交
546
        node.fluid_code.add_layer("conv2d",
J
jiangjiajun 已提交
547
                                  inputs=input,
J
jiangjiajun 已提交
548 549
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
550

J
jiangjiajun 已提交
551 552 553 554 555
    def Reshape(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        param = self.graph.get_node(node.layer.input[1], copy=True)
        if param.layer_type == "Const":
            attr = {"shape": param.value.tolist()}
J
jiangjiajun 已提交
556
            self.add_omit_nodes(param.layer_name, node.layer_name)
J
jiangjiajun 已提交
557 558
        else:
            # Here is a trick method to solove tensor parameter in tensorflow
J
jiangjiajun 已提交
559 560 561
            shape = self.decoder.infer_shape_tensor(param, node.out_shapes[0])
            if shape.count(-1) <= 1:
                attr = {"shape": shape}
J
jiangjiajun 已提交
562 563 564 565 566
                self.add_omit_nodes(param.layer_name, node.layer_name)
            elif shape.count(-1) == 2 and shape[0] == -1:
                shape[0] = 0
                attr = {"shape": shape}
                self.add_omit_nodes(param.layer_name, node.layer_name)
J
jiangjiajun 已提交
567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
            else:
                assert len(param.out_shapes[0]
                           ) == 1, "Unexpected situation of shape parameter"
                attr = {"shape": [-1]}
                node.fluid_code.add_layer("reshape",
                                          inputs=param,
                                          output="shape_param",
                                          param_attr=attr)
                attr = {"num_or_sections": param.out_shapes[0][0], "dim": 0}
                node.fluid_code.add_layer("split",
                                          inputs="shape_param",
                                          output=node,
                                          param_attr=attr)
                new_param = "["
                for i in range(param.out_shapes[0][0]):
                    new_param += (node.layer_name + "[{}]".format(i) + ", ")
                new_param = new_param.strip(", ") + "]"
                attr = {"shape": new_param}
585 586 587 588 589 590 591 592 593 594 595 596 597 598

        if len(input.out_shapes[0]) == 4 and node.tf_data_format == "NHWC":
            if len(attr["shape"]) < 3:
                perm = {"perm": [0, 2, 3, 1]}
                node.fluid_code.add_layer("transpose",
                                          inputs=input,
                                          output=node,
                                          param_attr=perm)
                node.fluid_code.add_layer("reshape",
                                          inputs=node,
                                          output=node,
                                          param_attr=attr)
                return

J
jiangjiajun 已提交
599
        if len(attr["shape"]) == 4 and node.tf_data_format == "NHWC":
J
jiangjiajun 已提交
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
            input_shape = self.decoder.infer_tensor(input).shape
            if input_shape[1] == attr["shape"][1]:
                attr["shape"] = [attr["shape"][i] for i in [0, 3, 1, 2]]
            else:
                perm = {"perm": [0, 2, 3, 1]}
                node.fluid_code.add_layer("transpose",
                                          inputs=input,
                                          output=node,
                                          param_attr=perm)
                node.fluid_code.add_layer("reshape",
                                          inputs=node,
                                          output=node,
                                          param_attr=attr)
                perm = {"perm": [0, 3, 1, 2]}
                node.fluid_code.add_layer("transpose",
                                          inputs=node,
                                          output=node,
                                          param_attr=perm)
                return
J
jiangjiajun 已提交
619 620 621 622 623 624 625
        node.fluid_code.add_layer("reshape",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def AvgPool(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
J
jiangjiajun 已提交
626

J
jiangjiajun 已提交
627
        in_shape = input.out_shapes[0]
J
jiangjiajun 已提交
628 629 630
        if in_shape.count(-1) > 2:
            in_shape = self.decoder.infer_tensor(input).shape

J
jiangjiajun 已提交
631 632 633 634 635 636 637 638 639
        k_size = node.get_attr("ksize")
        strides = node.get_attr("strides")
        data_format = node.get_attr("data_format").decode()
        pad_mode = node.get_attr("padding").decode()
        channel_first = data_format == "NCHW"

        if not channel_first:
            in_shape = [in_shape[i] for i in [0, 3, 1, 2]]
            strides = [strides[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
640
            k_size = [k_size[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
641 642
        else:
            self.graph.data_format_propagation(node)
J
jiangjiajun 已提交
643 644

        attr = {
J
jiangjiajun 已提交
645
            "pool_size": k_size[2:4],
J
jiangjiajun 已提交
646 647 648 649
            "pool_type": string("avg"),
            "pool_stride": strides[2:4]
        }
        if pad_mode == "SAME":
J
jiangjiajun 已提交
650 651
            pad_h = get_same_padding(in_shape[2], k_size[2], strides[2])
            pad_w = get_same_padding(in_shape[3], k_size[3], strides[3])
J
jiangjiajun 已提交
652 653 654 655
            assert pad_h[0] == pad_h[1] and pad_w[0] == pad_w[
                1], "Cannot map AvgPool"
            attr["pool_padding"] = [pad_h[0], pad_w[0]]
        node.fluid_code.add_layer("pool2d",
J
jiangjiajun 已提交
656
                                  inputs=input,
J
jiangjiajun 已提交
657 658 659
                                  output=node,
                                  param_attr=attr)

J
jiangjiajun 已提交
660 661 662 663 664 665
    def SplitV(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        num_sections = self.graph.get_node(node.layer.input[1], copy=True)
        dim = self.graph.get_node(node.layer.input[2], copy=True)
        assert num_sections.layer_type == "Const"
        assert dim.layer_type == "Const"
J
jiangjiajun 已提交
666 667
        self.add_omit_nodes(num_sections.layer_name, node.layer_name)
        self.add_omit_nodes(dim.layer_name, node.layer_name)
J
jiangjiajun 已提交
668 669 670
        dim = dim.value
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            dim = nhwc_dim_to_nchw(input, dim)
J
jiangjiajun 已提交
671 672 673 674 675 676 677 678
        attr = {
            "num_or_sections": num_sections.value.tolist(),
            "dim": dim.value
        }
        node.fluid_code.add_layer("split",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
679 680

    def ConcatV2(self, node):
J
jiangjiajun 已提交
681 682 683 684
        inputs = [
            self.graph.get_node(name, copy=True)
            for name in node.layer.input[:-1]
        ]
J
jiangjiajun 已提交
685 686
        axis = self.graph.get_node(node.layer.input[-1], copy=True)
        assert axis.layer_type == "Const"
J
jiangjiajun 已提交
687
        self.add_omit_nodes(axis.layer_name, node.layer_name)
J
jiangjiajun 已提交
688 689 690 691 692
        axis = axis.value
        if inputs[0].tf_data_format == "NHWC" and len(
                inputs[0].out_shapes[0]) == 4:
            axis = nhwc_dim_to_nchw(inputs[0], axis)
        attr = {"axis": axis}
J
jiangjiajun 已提交
693 694 695 696
        node.fluid_code.add_layer("concat",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
697 698 699 700

    def Tile(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        expand_times = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
701
        self.add_omit_nodes(expand_times.layer_name, node.layer_name)
702 703 704 705
        if expand_times.layer_type == "Const":
            expand_times = expand_times.value.tolist()
        else:
            expand_times = self.decoder.infer_shape_tensor(expand_times)
J
jiangjiajun 已提交
706 707 708 709 710
        if input.tf_data_format == "NHWC":
            if len(input.out_shapes[0]) == 4:
                expand_times = [expand_times[i] for i in [0, 3, 1, 2]]
            elif len(input.out_shape[0]) == 3:
                expand_times = [expand_times[i] for i in [2, 0, 1]]
711 712 713 714
        for i in range(len(expand_times)):
            if expand_times[i] < 0:
                expand_times[i] = 1

J
jiangjiajun 已提交
715
        attr = {"expand_times": expand_times}
J
jiangjiajun 已提交
716 717 718 719
        node.fluid_code.add_layer("expand",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
720 721

    def Pack(self, node):
J
jiangjiajun 已提交
722 723 724
        inputs = [
            self.graph.get_node(name, copy=True) for name in node.layer.input
        ]
J
jiangjiajun 已提交
725 726 727 728 729 730 731 732 733 734 735 736
        axis = node.get_attr("axis")
        if inputs[0].tf_data_format == "NHWC" and len(
                inputs[0].out_shapes[0]) == 4:
            tf_data_format = list(inputs[0].tf_data_format)
            tf_data_format.insert(axis, str(len(tf_data_format)))
            axis = nhwc_dim_to_nchw(inputs[0], axis)
            pd_data_format = list(inputs[0].pd_data_format)
            pd_data_format.insert(axis, str(len(pd_data_format)))
            node.tf_data_format = "".join(tf_data_format)
            node.pd_data_format = "".join(pd_data_format)

        attr = {"axis": axis}
J
jiangjiajun 已提交
737 738 739
        node.fluid_code.add_layer("stack",
                                  inputs=inputs,
                                  output=node,
J
jiangjiajun 已提交
740
                                  param_attr=attr)
J
jiangjiajun 已提交
741 742 743

    def Pad(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
J
jiangjiajun 已提交
744
        paddings = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
745
        assert paddings.layer_type == "Const", "Padding should be Const"
J
jiangjiajun 已提交
746
        self.add_omit_nodes(paddings.layer_name, node.layer_name)
J
jiangjiajun 已提交
747 748 749
        paddings = paddings.value.flatten().tolist()
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            paddings = [paddings[i] for i in [0, 1, 6, 7, 2, 3, 4, 5]]
J
jiangjiajun 已提交
750 751 752 753 754 755

        pad_op = "pad"
        if len(input.out_shapes[0]) == 4:
            if paddings[0] + paddings[1] + paddings[2] + paddings[3] == 0:
                paddings = paddings[4:]
                pad_op = "pad2d"
J
jiangjiajun 已提交
756
        attr = {"paddings": paddings}
J
jiangjiajun 已提交
757
        node.fluid_code.add_layer(pad_op,
J
jiangjiajun 已提交
758 759 760
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
761 762 763 764 765 766 767

    def Range(self, node):
        start = self.graph.get_node(node.layer.input[0], copy=True)
        limit = self.graph.get_node(node.layer.input[1], copy=True)
        delta = self.graph.get_node(node.layer.input[2], copy=True)
        if start.layer_type == "Const":
            start = start.value
768 769
        else:
            start = self.decoder.infer_tensor(start)
J
jiangjiajun 已提交
770 771
        if limit.layer_type == "Const":
            limit = limit.value
772 773
        else:
            limit = self.decoder.infer_tensor(limit)
J
jiangjiajun 已提交
774 775
        if delta.layer_type == "Const":
            delta = delta.value
776 777
        else:
            delta = self.decoder.infer_tensor(delta)
J
jiangjiajun 已提交
778 779 780
        self.add_omit_nodes(start.layer_name, node.layer_name)
        self.add_omit_nodes(limit.layer_name, node.layer_name)
        self.add_omit_nodes(delta.layer_name, node.layer_name)
781

J
jiangjiajun 已提交
782
        inputs = {"start": start, "end": limit, "step": delta}
J
jiangjiajun 已提交
783
        attr = {"dtype": string(node.dtype)}
784 785 786 787
        node.fluid_code.add_layer("range",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=None)
J
jiangjiajun 已提交
788 789 790 791 792

    def Mean(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        reduce_idx = self.graph.get_node(node.layer.input[1], copy=True)
        assert reduce_idx.layer_type == "Const", "Only support Const parameter[reduce_idx]"
J
jiangjiajun 已提交
793
        dims = reduce_idx.value.tolist()
J
jiangjiajun 已提交
794
        keep_dims = node.get_attr("keep_dims")
J
jiangjiajun 已提交
795 796 797 798 799 800

        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            for i in range(len(dims)):
                dims[i] = nhwc_dim_to_nchw(input, dims[i])

        attr = {"dim": dims, "keep_dim": keep_dims}
J
jiangjiajun 已提交
801 802 803 804 805 806 807 808 809 810 811
        node.fluid_code.add_layer("reduce_mean",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def MatMul(self, node):
        x = self.graph.get_node(node.layer.input[0], copy=True)
        y = self.graph.get_node(node.layer.input[1], copy=True)
        transpose_a = node.get_attr('transpose_a')
        transpose_b = node.get_attr('transpose_b')
        inputs = {"x": x, "y": y}
J
jiangjiajun 已提交
812 813 814 815 816 817 818 819 820 821
        # fix paddle shape infer problem
        # should be removed after paddle 1.6
        if x.out_shapes[0][-1] < 0 and y.out_shapes[0][0] > 0:
            shape = x.out_shapes[0]
            shape[-1] = y.out_shapes[0][0]
            attr = {"shape": shape}
            node.fluid_code.add_layer("reshape",
                                      inputs=x,
                                      output=x,
                                      param_attr=attr)
J
jiangjiajun 已提交
822 823 824 825 826 827 828 829 830 831
        attr = {"transpose_x": transpose_a, "transpose_y": transpose_b}
        node.fluid_code.add_layer("matmul",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=attr)

    def ArgMax(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        axis = self.graph.get_node(node.layer.input[1], copy=True)
        assert axis.layer_type == "Const", "ArgMax only support Const parameter"
J
jiangjiajun 已提交
832
        self.add_omit_nodes(axis.layer_name, node.layer_name)
J
jiangjiajun 已提交
833 834 835 836
        axis = axis.value
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            axis = nhwc_dim_to_nchw(input, axis)
        attr = {"axis": axis}
J
jiangjiajun 已提交
837 838 839 840 841 842 843 844 845 846 847 848 849
        node.fluid_code.add_layer("argmax",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def StridedSlice(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        begin = self.graph.get_node(node.layer.input[1], copy=True)
        end = self.graph.get_node(node.layer.input[2], copy=True)
        strides = self.graph.get_node(node.layer.input[3], copy=True)
        assert begin.layer_type == "Const"
        assert end.layer_type == "Const"
        assert strides.layer_type == "Const"
J
jiangjiajun 已提交
850 851 852
        self.add_omit_nodes(begin.layer_name, node.layer_name)
        self.add_omit_nodes(end.layer_name, node.layer_name)
        self.add_omit_nodes(strides.layer_name, node.layer_name)
J
jiangjiajun 已提交
853 854 855
        strides = strides.value.tolist()
        assert len(set(strides)) == 1 and strides[0] == 1

J
jiangjiajun 已提交
856 857 858 859 860 861
        begin = begin.value.tolist()
        end = end.value.tolist()
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            begin = [begin[i] for i in [0, 3, 1, 2]]
            end = [end[i] for i in [0, 3, 1, 2]]

J
jiangjiajun 已提交
862 863 864 865 866 867 868 869 870
        for i in range(len(end)):
            if end[i] == 0:
                end[i] = 999999

        attr = {
            "axes": [i for i in range(len(strides))],
            "starts": begin,
            "ends": end
        }
J
jiangjiajun 已提交
871 872 873 874
        node.fluid_code.add_layer("slice",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
875 876 877 878 879

    def Slice(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        begin = self.graph.get_node(node.layer.input[1], copy=True)
        size = self.graph.get_node(node.layer.input[2], copy=True)
J
jiangjiajun 已提交
880 881
        self.add_omit_nodes(begin.layer_name, node.layer_name)
        self.add_omit_nodes(size.layer_name, node.layer_name)
J
jiangjiajun 已提交
882 883 884 885 886 887 888 889
        if begin.layer_type == "Const":
            begin = begin.value.tolist()
        else:
            begin = self.decoder.infer_tensor(begin).tolist()
        if size.layer_type == "const":
            size = size.value.tolist()
        else:
            size = self.decoder.infer_tensor(size).tolist()
890

J
jiangjiajun 已提交
891 892 893 894
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            size = [size[i] for i in [0, 3, 1, 2]]
            begin = [begin[i] for i in [0, 3, 1, 2]]

895 896 897 898 899 900 901 902 903 904 905 906
        for i in range(len(size)):
            if size[i] < 0:
                size[i] = 99999999
            else:
                size[i] = size[i] + begin[i]

        attr = {
            "axes": [i for i in range(len(size))],
            "starts": begin,
            "ends": size
        }
        node.fluid_code.add_layer("slice",
J
jiangjiajun 已提交
907 908 909
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
910 911

    def Conv2DBackpropInput(self, node):
912
        out_shape = self.graph.get_node(node.layer.input[0], copy=True)
913
        kernel = self.graph.get_node(node.layer.input[1], copy=True)
914 915
        input = self.graph.get_node(node.layer.input[2], copy=True)

916
        assert kernel.layer_type == "Const", "Kernel of Conv2DBackpropInput should be Const"
917

J
jiangjiajun 已提交
918
        self.add_omit_nodes(kernel.layer_name, node.layer_name)
919 920
        self.add_omit_nodes(out_shape.layer_name, node.layer_name)

J
jiangjiajun 已提交
921 922 923 924 925 926
        if out_shape.layer_type == "Const":
            out_shape = out_shape.value.tolist()
        else:
            out_shape = self.decoder.infer_shape_tensor(out_shape,
                                                        node.out_shapes[0])

927
        in_shape = input.out_shapes[0]
J
jiangjiajun 已提交
928 929
        if in_shape.count(-1) > 2:
            in_shape = self.decoder.infer_tensor(input).shape
930
        k_size = kernel.out_shapes[0]
J
jiangjiajun 已提交
931 932 933
        if k_size.count(-1) > 2:
            k_size = self.decoder.infer_tensor(kernel).shape

J
jiangjiajun 已提交
934
        pad_mode = node.get_attr("padding").decode()
935 936 937 938
        strides = node.get_attr("strides")
        dilations = node.get_attr("dilations")
        data_format = node.get_attr("data_format").decode()
        channel_first = data_format == "NCHW"
939

J
jiangjiajun 已提交
940 941
        self.weights[kernel.layer_name.replace('/', '_')] = numpy.transpose(
            kernel.value, (3, 2, 0, 1))
942 943 944 945
        if not channel_first:
            in_shape = [in_shape[i] for i in [0, 3, 1, 2]]
            strides = [strides[i] for i in [0, 3, 1, 2]]
            dilations = [dilations[i] for i in [0, 3, 1, 2]]
J
jiangjiajun 已提交
946 947
        else:
            self.data_format_propagation(node)
948

J
jiangjiajun 已提交
949
        padding = 0
950 951 952
        if pad_mode == "SAME":
            pad_h = get_same_padding(in_shape[2], k_size[0], strides[2])
            pad_w = get_same_padding(in_shape[3], k_size[1], strides[3])
J
jiangjiajun 已提交
953 954 955 956 957 958 959 960 961
            if pad_h[0] == pad_h[1] and pad_w[0] == pad_w[1]:
                padding = [pad_h[0], pad_w[0]]
            else:
                attr = {"paddings": pad_h + pad_w, "pad_value": 0.0}
                node.fluid_code.add_layer("pad2d",
                                          inputs=input,
                                          output=node,
                                          param_attr=attr)
                input = node
962

963 964 965 966 967 968
        attr = {
            "bias_attr": False,
            "param_attr": string(kernel.layer_name),
            "num_filters": k_size[3],
            "filter_size": k_size[0:2],
            "stride": strides[2:4],
J
jiangjiajun 已提交
969 970
            "dilation": dilations[2:4],
            "padding": padding
971
        }
972 973 974 975
        node.fluid_code.add_layer("conv2d_transpose",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
976

J
jiangjiajun 已提交
977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992
        if pad_mode == "SAME":
            if node.tf_data_format == "NHWC":
                out_shape = [out_shape[i] for i in [0, 3, 1, 2]]
            for i in range(4):
                if out_shape[i] < 0:
                    out_shape[i] = 999999
            attr = {
                "axes": [0, 1, 2, 3],
                "starts": [0, 0, 0, 0],
                "ends": out_shape
            }
            node.fluid_code.add_layer("slice",
                                      inputs=node,
                                      output=node,
                                      param_attr=attr)

993 994 995 996 997
    def Max(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        reduce_idx = self.graph.get_node(node.layer.input[1], copy=True)
        assert reduce_idx.layer_type == "Const", "Only support Const parameter[reduce_idx]"
        keep_dims = node.get_attr("keep_dims")
J
jiangjiajun 已提交
998 999 1000 1001 1002
        dim = reduce_idx.value.tolist()
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            dim = nhwc_dim_to_nchw(input, dim)

        attr = {"dim": dim, "keep_dim": keep_dims}
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
        node.fluid_code.add_layer("reduce_max",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def Sum(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        reduce_idx = self.graph.get_node(node.layer.input[1], copy=True)
        assert reduce_idx.layer_type == "Const", "Only support Const parameter[reduce_idx]"
        keep_dims = node.get_attr("keep_dims")
J
jiangjiajun 已提交
1013 1014 1015 1016 1017
        dim = reduce_idx.value.tolist()
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            dim = nhwc_dim_to_nchw(input, dim)

        attr = {"dim": dim, "keep_dim": keep_dims}
1018 1019 1020 1021 1022
        node.fluid_code.add_layer("reduce_sum",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

J
jiangjiajun 已提交
1023 1024 1025 1026 1027 1028 1029 1030
    def Cast(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        dtype = node.dtype_map[node.get_attr('DstT')]
        attr = {"dtype": string(dtype)}
        node.fluid_code.add_layer("cast",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
1031

J
jiangjiajun 已提交
1032 1033 1034
    def Split(self, node):
        dim = self.graph.get_node(node.layer.input[0], copy=True)
        input = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
1035
        self.add_omit_nodes(dim.layer_name, node.layer_name)
J
jiangjiajun 已提交
1036
        num_split = node.get_attr('num_split')
J
jiangjiajun 已提交
1037 1038 1039 1040 1041
        dim = dim.value
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            dim = nhwc_dim_to_nchw(input, dim)

        attr = {"num_or_sections": num_split, "dim": dim}
J
jiangjiajun 已提交
1042 1043 1044 1045
        node.fluid_code.add_layer("split",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061

    def Squeeze(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        squeeze_dims = node.get_attr('squeeze_dims')
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            for i in range(len(squeeze_dims)):
                squeeze_dims[i] = nhwc_dim_to_nchw(input, squeeze_dims[i])
        attr = {"axes": squeeze_dims}
        node.fluid_code.add_layer("squeeze",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def Softmax(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        axis = node.get_attr("axis")
J
jiangjiajun 已提交
1062 1063
        if axis is None:
            axis = -1 + len(input.out_shapes[0])
J
jiangjiajun 已提交
1064 1065 1066 1067 1068 1069 1070
        if input.tf_data_format == "NHWC" and len(input.out_shapes[0]) == 4:
            axis = nhwc_dim_to_nchw(input, axis)
        attr = {"axis": axis}
        node.fluid_code.add_layer("softmax",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
J
jiangjiajun 已提交
1071 1072 1073 1074

    def ResizeNearestNeighbor(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        resize_shape = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
1075
        self.add_omit_nodes(resize_shape.layer_name, node.layer_name)
J
jiangjiajun 已提交
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
        if resize_shape.layer_type == "Const":
            resize_shape = resize_shape.value.tolist()
        else:
            resize_shape = self.decoder.infer_shape_tensor(resize_shape)
        align_corners = node.get_attr("align_corners")
        attr = {"align_corners": align_corners, "out_shape": resize_shape}
        node.fluid_code.add_layer("resize_nearest",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def ResizeBilinear(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        resize_shape = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
1090
        self.add_omit_nodes(resize_shape.layer_name, node.layer_name)
J
jiangjiajun 已提交
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
        if resize_shape.layer_type == "Const":
            resize_shape = resize_shape.value.tolist()
        else:
            resize_shape = self.decoder.infer_shape_tensor(resize_shape)
        align_corners = node.get_attr("align_corners")
        attr = {
            "align_corners": align_corners,
            "out_shape": resize_shape,
            "align_mode": 1
        }
        node.fluid_code.add_layer("resize_bilinear",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
1105 1106 1107 1108

    def ResizeNearestNeighbor(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        resize_shape = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
1109
        self.add_omit_nodes(resize_shape.layer_name, node.layer_name)
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
        if resize_shape.layer_type == "Const":
            resize_shape = resize_shape.value.tolist()
        else:
            resize_shape = self.decoder.infer_shape_tensor(
                resize_shape, node.out_shapes[0])
        align_corners = node.get_attr("align_corners")
        attr = {"align_corners": align_corners, "out_shape": resize_shape}
        node.fluid_code.add_layer("resize_nearest",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)

    def ResizeBilinear(self, node):
        input = self.graph.get_node(node.layer.input[0], copy=True)
        resize_shape = self.graph.get_node(node.layer.input[1], copy=True)
J
jiangjiajun 已提交
1125
        self.add_omit_nodes(resize_shape.layer_name, node.layer_name)
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
        if resize_shape.layer_type == "Const":
            resize_shape = resize_shape.value.tolist()
        else:
            resize_shape = self.decoder.infer_shape_tensor(
                resize_shape, node.out_shapes[0])
        align_corners = node.get_attr("align_corners")
        attr = {
            "align_corners": align_corners,
            "out_shape": resize_shape,
            "align_mode": 1
        }
        node.fluid_code.add_layer("resize_bilinear",
                                  inputs=input,
                                  output=node,
                                  param_attr=attr)
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171

    def GreaterEqual(self, node):
        x = self.graph.get_node(node.layer.input[0], copy=True)
        y = self.graph.get_node(node.layer.input[1], copy=True)
        inputs = {"x": x, "y": y}
        node.fluid_code.add_layer("greater_equal",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=None)

    def RandomUniform(self, node):
        shape = self.graph.get_node(node.layer.input[0], copy=True)
        self.add_omit_nodes(shape.layer_name, node.layer_name)
        if shape.layer_type == "Const":
            shape = shape.value.tolist()
        else:
            shape = self.decoder.infer_shape_tensor(shape)
        if node.tf_data_format == "NHWC" and len(shape) == 4:
            shape = [shape[i] for i in [0, 3, 1, 2]]
        attr = {"shape": shape, "min": 0.0, "max": 0.9999}
        if shape[0] < 0:
            input = self.batch_node
            node.fluid_code.add_layer("uniform_random_batch_size_like",
                                      inputs=input,
                                      output=node,
                                      param_attr=attr)
        else:
            node.fluid_code.add_layer("uniform_random",
                                      inputs=None,
                                      output=node,
                                      param_attr=attr)
J
jiangjiajun 已提交
1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202

    def GreaterEqual(self, node):
        x = self.graph.get_node(node.layer.input[0], copy=True)
        y = self.graph.get_node(node.layer.input[1], copy=True)
        inputs = {"x": x, "y": y}
        node.fluid_code.add_layer("greater_equal",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=None)

    def RandomUniform(self, node):
        shape = self.graph.get_node(node.layer.input[0], copy=True)
        self.add_omit_nodes(shape.layer_name, node.layer_name)
        if shape.layer_type == "Const":
            shape = shape.value.tolist()
        else:
            shape = self.decoder.infer_shape_tensor(shape)
        if len(shape) == 4 and node.tf_data_format == "NHWC":
            shape = [shape[i] for i in [0, 3, 1, 2]]
        attr = {"shape": shape, "min": 0.0, "max": 0.9999}
        if shape[0] < 0:
            input = self.batch_node
            node.fluid_code.add_layer("uniform_random_batch_size_like",
                                      inputs=input,
                                      output=node,
                                      param_attr=attr)
        else:
            node.fluid_code.add_layer("uniform_random",
                                      inputs=None,
                                      output=node,
                                      param_attr=attr)
J
jiangjiajun 已提交
1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216

    def SquaredDifference(self, node):
        x = self.graph.get_node(node.layer.input[0], copy=True)
        y = self.graph.get_node(node.layer.input[1], copy=True)
        inputs = {"x": x, "y": y}
        node.fluid_code.add_layer("elementwise_sub",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=None)
        inputs = {"x": node, "y": node}
        node.fluid_code.add_layer("elementwise_mul",
                                  inputs=inputs,
                                  output=node,
                                  param_attr=None)