tf_decoder.py 20.1 KB
Newer Older
J
jiangjiajun 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
#   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.

from x2paddle.core.graph import GraphNode, Graph
J
jiangjiajun 已提交
16 17
from x2paddle.core.fluid_code import FluidCode
from tensorflow.python.framework import tensor_util
J
jiangjiajun 已提交
18
from tensorflow.core.framework import attr_value_pb2
J
jiangjiajun 已提交
19
import tensorflow as tf
J
jiangjiajun 已提交
20
import copy as cp
J
jiangjiajun 已提交
21
import numpy
J
jiangjiajun 已提交
22
import sys
J
jiangjiajun 已提交
23

24

J
jiangjiajun 已提交
25
class TFGraphNode(GraphNode):
J
jiangjiajun 已提交
26
    def __init__(self, layer, layer_name=None, data_format="NHWC"):
J
jiangjiajun 已提交
27
        if layer_name is None:
J
jiangjiajun 已提交
28 29 30
            super(TFGraphNode, self).__init__(
                layer,
                layer.name.replace('/', '_').replace('-', '_').replace('^', ''))
J
jiangjiajun 已提交
31
        else:
J
jiangjiajun 已提交
32 33 34
            super(TFGraphNode, self).__init__(
                layer,
                layer_name.replace('/', '_').replace('-', '_').replace('^', ''))
J
jiangjiajun 已提交
35

J
jiangjiajun 已提交
36
        self.layer_type = layer.op
J
jiangjiajun 已提交
37 38
        self.tf_data_format = data_format
        self.pd_data_format = "NCHW"
J
jiangjiajun 已提交
39
        self.fluid_code = FluidCode()
J
jiangjiajun 已提交
40

J
jiangjiajun 已提交
41 42 43 44 45 46 47
        self.dtype_map = {
            1: "float32",
            3: "int32",
            4: "uint8",
            9: "int64",
            10: "bool"
        }
48 49 50

    @property
    def out_shapes(self):
M
mamingjie-China 已提交
51
        if self.layer_type == "OneShotIterator" or self.layer_type == "IteratorV2":
J
jiangjiajun@baidu.com 已提交
52 53 54
            values = self.layer.attr["output_shapes"].list.shape
        else:
            values = self.layer.attr["_output_shapes"].list.shape
55 56 57 58 59 60 61 62
        out_shapes = list()
        for value in values:
            shape = [dim.size for dim in value.dim]
            out_shapes.append(shape)
        return out_shapes

    @property
    def dtype(self):
J
jiangjiajun 已提交
63
        keys = ['dtype', 'Tidx', 'T', 'DstT']
64 65 66 67
        for k in keys:
            dtype = self.layer.attr[k].type
            if dtype > 0:
                break
J
jiangjiajun@baidu.com 已提交
68 69
        if dtype == 0:
            dtype = self.layer.attr['output_types'].list.type[0]
70
        if dtype not in self.dtype_map:
M
mamingjie-China 已提交
71 72
            raise Exception("Dtype[{}] of node({}) not in dtype_map".format(
                dtype, self.layer.name))
73 74
        return self.dtype_map[dtype]

J
jiangjiajun 已提交
75 76 77 78 79 80 81 82 83
    @property
    def raw_dtype(self):
        keys = ['dtype', 'Tidx', 'T', 'DstT']
        for k in keys:
            dtype = self.layer.attr[k].type
            if dtype > 0:
                break
        return dtype

J
jiangjiajun 已提交
84 85 86 87 88 89 90 91
    @property
    def value(self):
        assert self.layer_type == "Const", "Only Const node has value."

        attr = self.layer.attr['value']
        field = getattr(attr, attr.WhichOneof('value'))
        return tensor_util.MakeNdarray(field)

J
jiangjiajun 已提交
92 93
    @property
    def name(self):
M
mamingjie-China 已提交
94 95
        if hasattr(self, 'index'):
            return self.layer_name + "_p{}".format(self.index)
J
jiangjiajun 已提交
96 97
        return self.layer_name

J
jiangjiajun 已提交
98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
    def get_attr(self, name):
        if name not in self.layer.attr:
            return None
        attr = self.layer.attr[name]
        field = attr.WhichOneof('value')
        value = getattr(attr, field) if field else None

        if isinstance(value, attr_value_pb2.AttrValue.ListValue):
            result = list(value.ListFields()[0][1])
            for i in range(len(result)):
                if isinstance(result[i], int):
                    result[i] = int(result[i])
                try:
                    if isinstance(result[i], long):
                        result[i] = int(result[i])
                except:
                    pass
            return result
        else:
            return value

J
jiangjiajun 已提交
119 120

class TFGraph(Graph):
J
jiangjiajun 已提交
121
    def __init__(self, model, data_format="NHWC"):
J
jiangjiajun 已提交
122
        super(TFGraph, self).__init__(model)
J
jiangjiajun 已提交
123
        self.identity_map = dict()
M
mamingjie-China 已提交
124
        self.multi_out_ops = ['Split', 'SplitV', 'IteratorV2']
J
jiangjiajun 已提交
125
        self.tf_data_format = data_format
J
jiangjiajun 已提交
126 127 128

    def build(self):
        for layer in self.model.node:
M
mamingjie-China 已提交
129 130
            if layer.op == 'Assert':
                continue
J
jiangjiajun 已提交
131
            self.node_map[layer.name.replace('/', '_').replace(
J
jiangjiajun 已提交
132 133
                '-', '_')] = TFGraphNode(
                    layer, data_format=self.tf_data_format)
J
jiangjiajun 已提交
134

J
jiangjiajun 已提交
135
        for layer_name, node in self.node_map.items():
M
mamingjie-China 已提交
136 137
            if node.layer_type == 'Const':
                continue
J
jiangjiajun 已提交
138
            for in_node in node.layer.input:
J
jiangjiajun 已提交
139 140
                in_node = in_node.replace('/', '_').replace('-', '_').replace(
                    '^', '')
J
jiangjiajun 已提交
141 142
                if in_node not in self.node_map:
                    if in_node.strip().split(':')[0] in self.node_map:
J
jiangjiajun 已提交
143
                        self.connect(in_node.strip().split(':')[0], layer_name)
J
jiangjiajun 已提交
144
                    else:
145 146 147
                        raise Exception(
                            'input[{}] of node[{}] does not exist in node_map'.
                            format(in_node, layer_name))
J
jiangjiajun 已提交
148 149 150
                else:
                    self.connect(in_node, layer_name)

151
        super(TFGraph, self).build()
J
jiangjiajun 已提交
152

M
mamingjie-China 已提交
153 154 155 156 157 158 159 160
        for layer in self.model.node:
            if layer.op == 'Assert':
                for ipt in layer.input:
                    ipt_name = ipt.replace('-', '_').replace('/', '_')
                    if ipt_name in self.output_nodes:
                        idx = self.output_nodes.index(ipt_name)
                        del self.output_nodes[idx]

J
jiangjiajun 已提交
161 162
        # tensorflow graph optimize
        self._remove_isolated_node()
J
jiangjiajun@baidu.com 已提交
163
        self._optimize_dialiation_conv()
J
jiangjiajun 已提交
164
        self._remove_identity_node()
J
jiangjiajun 已提交
165
        self._remove_cast_node()
J
jiangjiajun 已提交
166 167 168

    def get_node(self, node_name, copy=False):
        items = node_name.strip().split(':')
J
jiangjiajun 已提交
169
        items[0] = items[0].replace('/', '_').replace('-', '_')
J
jiangjiajun 已提交
170 171 172
        if items[0] in self.identity_map:
            items[0] = self.identity_map[items[0]]
        new_node_name = ":".join(items)
J
jiangjiajun 已提交
173
        node = super(TFGraph, self).get_node(new_node_name, copy)
J
jiangjiajun 已提交
174 175
        if node is None:
            return None
J
jiangjiajun 已提交
176 177 178
        if node.layer_type == "Switch":
            if hasattr(node, 'index'):
                del node.index
J
jiangjiajun 已提交
179 180 181
        if len(items) == 1 and node.layer_type in self.multi_out_ops:
            node.index = 0
        return node
J
jiangjiajun 已提交
182

J
jiangjiajun 已提交
183 184 185 186 187
    def remove_node(self, node_name):
        if node_name not in self.node_map:
            raise Exception("Node[{}] not in graph".format(node_name))
        inputs = self.node_map[node_name].inputs
        outputs = self.node_map[node_name].outputs
188
        #        assert len(inputs) == 1
J
jiangjiajun 已提交
189 190 191 192 193 194 195 196 197 198 199 200 201 202
        input_node = self.node_map[inputs[0]]
        idx = input_node.outputs.index(node_name)
        del input_node.outputs[idx]
        for output in outputs:
            node = self.node_map[output]
            idx = node.inputs.index(node_name)
            node.inputs[idx] = inputs[0]
            input_node.outputs.append(output)

        del self.node_map[node_name]

        idx = self.topo_sort.index(node_name)
        del self.topo_sort[idx]

J
jiangjiajun@baidu.com 已提交
203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230
    def _optimize_dialiation_conv(self):
        for name in list(self.node_map.keys()):
            node = self.node_map[name]
            if node.layer_type == "SpaceToBatchND":
                is_dilation = True
                out_node0 = self.node_map[node.outputs[0]]
                if out_node0.layer_type != 'ExpandDims':
                    is_dilation = False
                    continue
                out_node1 = self.node_map[out_node0.outputs[0]]
                if out_node1.layer_type != 'Conv2D':
                    is_dilation = False
                    continue
                out_node2 = self.node_map[out_node1.outputs[0]]
                if out_node2.layer_type != 'Squeeze':
                    is_dilation = False
                    continue
                out_node3 = self.node_map[out_node2.outputs[0]]
                if out_node3.layer_type != 'BatchToSpaceND':
                    is_dilation = False
                    continue

                if is_dilation:
                    node.skip = True
                    out_node3.skip = True
                    block_shape = self.node_map[node.inputs[1]]
                    out_node1.dilation = block_shape.value.tolist()

J
jiangjiajun 已提交
231 232 233 234
    def _remove_isolated_node(self):
        # delete isolated nodes
        isolated_nodes = list()
        for node_name in self.node_map.keys():
J
jiangjiajun 已提交
235
            if len(self.get_node(node_name).inputs) == 0 and len(
J
jiangjiajun 已提交
236 237 238
                    self.get_node(node_name).outputs) == 0:
                isolated_nodes.append(node_name)

J
jiangjiajun 已提交
239
        for node_name in isolated_nodes:
J
jiangjiajun 已提交
240 241 242 243 244 245 246 247 248
            del self.node_map[node_name]
            if node_name in self.input_nodes:
                idx = self.input_nodes.index(node_name)
                del self.input_nodes[idx]
            if node_name in self.output_nodes:
                idx = self.output_nodes.index(node_name)
                del self.output_nodes[idx]
            idx = self.topo_sort.index(node_name)
            del self.topo_sort[idx]
J
jiangjiajun 已提交
249 250

    def _remove_identity_node(self):
J
jiangjiajun 已提交
251 252
        identity_ops = [
            'Identity', 'StopGradient', 'Switch', 'Merge',
J
jiangjiajun@baidu.com 已提交
253
            'PlaceholderWithDefault', 'IteratorGetNext'
J
jiangjiajun 已提交
254
        ]
J
jiangjiajun 已提交
255 256
        identity_node = list()
        for node_name, node in self.node_map.items():
J
jiangjiajun 已提交
257
            if node.layer_type in identity_ops:
J
jiangjiajun 已提交
258 259 260 261 262
                identity_node.append(node_name)

        for node_name in identity_node:
            node = self.get_node(node_name)
            input_node = self.get_node(node.inputs[0])
J
jiangjiajun 已提交
263
            self.remove_node(node_name)
J
jiangjiajun 已提交
264 265 266

            self.identity_map[node_name] = input_node.layer_name

J
jiangjiajun 已提交
267 268 269 270
            if node_name in self.output_nodes:
                idx = self.output_nodes.index(node_name)
                self.output_nodes[idx] = input_node.layer_name

J
jiangjiajun 已提交
271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291
    def _remove_cast_node(self):
        cast_node = list()
        for node_name, node in self.node_map.items():
            if node.layer_type == "Cast":
                input = self.get_node(node.inputs[0])
                if input.layer_type != "Placeholder" or len(input.outputs) != 1:
                    continue
                cast_node.append(node_name)

        for node_name in cast_node:
            node = self.get_node(node_name)
            input_node = self.get_node(node.inputs[0])
            input_node.layer.attr["dtype"].type = node.raw_dtype
            self.remove_node(node_name)

            self.identity_map[node_name] = input_node.layer_name

            if node_name in self.output_nodes:
                idx = self.output_nodes.index(node_name)
                self.output_nodes[idx] = input_node.layer_name

J
jiangjiajun 已提交
292 293 294 295 296 297 298 299 300 301
    def data_format_propagation(self, node):
        current_node = self.node_map[node.layer_name]
        outputs = current_node.outputs
        if len(outputs) == 0:
            return
        for out in outputs:
            next_node = self.node_map[out]
            next_node.tf_data_format = node.tf_data_format
            self.data_format_propagation(next_node)

J
jiangjiajun 已提交
302

J
jiangjiajun 已提交
303
class TFDecoder(object):
304
    def __init__(self, pb_model, data_format="NHWC", define_input_shape=False):
305 306 307 308
        try:
            self.sess = tf.compat.v1.Session()
        except:
            self.sess = tf.Session()
J
jiangjiajun 已提交
309
        self.input_info = dict()
310
        self.define_input_shape = define_input_shape
311 312 313 314 315
        with open(pb_model, 'rb') as f:
            try:
                graph_def = tf.compat.v1.GraphDef()
            except:
                graph_def = tf.GraphDef()
J
jiangjiajun 已提交
316
            graph_def.ParseFromString(f.read())
J
jiangjiajun 已提交
317
            input_map = self._check_input_shape(graph_def)
J
jiangjiajun 已提交
318
            self._fix_output_shape(graph_def)
J
jiangjiajun 已提交
319
            self.sess.graph.as_default()
J
jiangjiajun 已提交
320
            tf.import_graph_def(graph_def, name='', input_map=input_map)
321

322 323 324 325 326
        try:
            initializer = tf.compat.v1.global_variables_initializer()
        except:
            initializer = tf.global_variables_initializer()
        self.sess.run(initializer)
J
jiangjiajun 已提交
327

J
jiangjiajun 已提交
328
        self.tf_graph = TFGraph(
J
jiangjiajun 已提交
329
            self.sess.graph._as_graph_def(add_shapes=True)[0], data_format)
J
jiangjiajun 已提交
330
        self.tf_graph.build()
J
jiangjiajun 已提交
331 332 333 334 335 336

    def _fix_output_shape(self, graph):
        for i in range(len(graph.node)):
            node = graph.node[i]
            if node.op == "swish_f32":
                graph.node[i].attr['_disable_call_shape_inference'].b = False
J
jiangjiajun 已提交
337 338

    def _check_input_shape(self, graph_def):
J
jiangjiajun 已提交
339
        numpy.random.seed(13)
J
jiangjiajun 已提交
340 341 342
        graph_def = cp.deepcopy(graph_def)
        input_map = dict()
        for layer in graph_def.node:
M
mamingjie-China 已提交
343
            if layer.op != "Placeholder" and layer.op != "OneShotIterator" and layer.op != "IteratorV2":
J
jiangjiajun 已提交
344 345
                continue
            graph_node = TFGraphNode(layer)
346
            dtype = graph_node.layer.attr['dtype'].type
J
jiangjiajun 已提交
347 348

            need_define_shape = 0
349 350 351 352 353
            if self.define_input_shape:
                need_define_shape = 3
            elif graph_node.layer.attr[
                    'shape'].shape.unknown_rank or not graph_node.get_attr(
                        "shape"):
J
jiangjiajun 已提交
354 355 356 357 358 359 360
                need_define_shape = 1
            else:
                value = graph_node.layer.attr["shape"].shape
                shape = [dim.size for dim in value.dim]
                if shape.count(-1) > 1:
                    need_define_shape = 2

J
jiangjiajun@baidu.com 已提交
361
            if need_define_shape == 1:
J
fix bug  
jiangjiajun 已提交
362 363 364 365 366 367
                try:
                    shape = graph_node.out_shapes[0]
                    if len(shape) > 0 and shape.count(-1) < 2:
                        need_define_shape = 0
                except:
                    pass
J
jiangjiajun@baidu.com 已提交
368

J
jiangjiajun 已提交
369
            if need_define_shape > 0:
370 371 372 373
                shape = None
                if graph_node.get_attr("shape"):
                    value = value = graph_node.layer.attr["shape"].shape
                    shape = [dim.size for dim in value.dim]
J
jiangjiajun 已提交
374
                if need_define_shape == 1:
J
jiangjiajun 已提交
375 376
                    print("Unknown shape for input tensor[tensor name: \"{}\"]".
                          format(layer.name))
377
                elif need_define_shape == 2:
J
jiangjiajun 已提交
378
                    print(
J
jiangjiajun 已提交
379 380
                        "\nShape[now is {}] for input tensor[tensor name: \"{}\"] not support yet"
                        .format(shape, layer.name))
381 382 383 384
                else:
                    print(
                        "Define shape[now is {}] for input tensor[tensor name: \"{}\']"
                        .format(shape, layer.name))
J
jiangjiajun 已提交
385
                print(
J
jiangjiajun 已提交
386 387 388 389
                    "Use your keyboard type the shape of input tensor below :)")

                right_shape_been_input = False
                while not right_shape_been_input:
M
mamingjie-China 已提交
390 391 392 393 394
                    try:
                        shape = raw_input(
                            "Shape of Input(e.g. None,224,224,3): ")
                    except:
                        shape = input("Shape of Input(e.g. None,224,224,3): ")
J
jiangjiajun 已提交
395
                    if shape.count("None") > 1:
J
jiangjiajun 已提交
396
                        print("Only 1 dimension can be None, type again:)")
J
jiangjiajun 已提交
397 398 399
                    else:
                        right_shape_been_input = True

J
jiangjiajun 已提交
400 401 402 403
                shape = [
                    None if dim == "None" else int(dim)
                    for dim in shape.strip().split(',')
                ]
J
jiangjiajun 已提交
404
                assert shape.count(None) <= 1, "Only one dimension can be None"
405 406 407 408 409 410
                try:
                    x2paddle_input = tf.compat.v1.placeholder(
                        dtype=dtype,
                        shape=shape,
                        name="x2paddle_{}".format(layer.name))
                except:
J
jiangjiajun 已提交
411 412 413 414
                    x2paddle_input = tf.placeholder(
                        dtype=dtype,
                        shape=shape,
                        name="x2paddle_{}".format(layer.name))
415

J
jiangjiajun 已提交
416
                input_map["{}:0".format(layer.name)] = x2paddle_input
417 418
                if shape.count(None) > 0:
                    shape[shape.index(None)] = -1
J
jiangjiajun 已提交
419 420 421 422 423
                self.input_info["x2paddle_{}".format(layer.name)] = (shape,
                                                                     dtype)
            else:
                value = graph_node.layer.attr["shape"].shape
                shape = [dim.size for dim in value.dim]
M
mamingjie-China 已提交
424
                self.input_info[layer.name] = (shape, dtype)
J
jiangjiajun 已提交
425

J
jiangjiajun 已提交
426
        return input_map
J
jiangjiajun 已提交
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486

    # trick method
    # should be removed after PaddlePaddle V1.6 been released
    def infer_tensor(self, graph_node):
        if hasattr(graph_node, "index"):
            tensor_name = graph_node.layer.name + ":{}".format(graph_node.index)
        else:
            tensor_name = graph_node.layer.name + ":0"
        feed = dict()
        for input_name, info in self.input_info.items():
            (shape, dtype) = cp.deepcopy(info)
            input_tensor = self.sess.graph.get_tensor_by_name(input_name + ":0")
            if shape.count(-1) > 0:
                shape[shape.index(-1)] = 2
            feed[input_tensor] = numpy.random.random_sample(shape)
        output_tensor = self.sess.graph.get_tensor_by_name(tensor_name)
        return self.sess.run([output_tensor], feed)[0]

    def infer_shape_tensor(self, graph_node, out_shape=None):
        if hasattr(graph_node, "index"):
            tensor_name = graph_node.layer.name + ":{}".format(graph_node.index)
        else:
            tensor_name = graph_node.layer.name + ":0"
        feed = dict()
        batch_size = [2, 3, 5]
        results = list()
        for b in batch_size:
            for input_name, info in self.input_info.items():
                (shape, dtype) = cp.deepcopy(info)
                input_tensor = self.sess.graph.get_tensor_by_name(input_name +
                                                                  ":0")
                if shape.count(-1) > 0:
                    shape[shape.index(-1)] = b
                feed[input_tensor] = numpy.random.random_sample(shape)
            output_tensor = self.sess.graph.get_tensor_by_name(tensor_name)
            results.append(self.sess.run([output_tensor], feed)[0].flatten())

        compare01 = (results[0] == results[1])
        compare12 = (results[1] == results[2])

        if compare01.all() and compare12.all():
            return results[0].tolist()

        if (compare01 == compare12).all():
            index = numpy.argwhere(compare01 == False).flatten()
            if index.shape[0] != 1:
                raise Exception("There's not only one unstable dimension")
            results[0][index[0]] = -1

            index = numpy.argwhere(results[0] < 0).flatten()
            if index.shape[0] > 2:
                print("Warning: More than two dimension less than zero")
            if index.shape[0] == 2 and out_shape is not None:
                if out_shape[index[1]] > 0:
                    results[0][index[1]] = out_shape[index[1]]
                else:
                    results[0][index[0]] = out_shape[index[0]]
            return results[0].tolist()
        else:
            raise Exception("Couldn't infer a stable shape shape tensor value")
J
jiangjiajun 已提交
487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521

    def infer_tensor_shape(self, graph_node):
        if hasattr(graph_node, "index"):
            tensor_name = graph_node.layer.name + ":{}".format(graph_node.index)
        else:
            tensor_name = graph_node.layer.name + ":0"
        feed = dict()
        batch_size = [2, 3, 5]
        shapes = list()
        for b in batch_size:
            for input_name, info in self.input_info.items():
                (shape, dtype) = cp.deepcopy(info)
                input_tensor = self.sess.graph.get_tensor_by_name(input_name +
                                                                  ":0")
                if shape.count(-1) > 0:
                    shape[shape.index(-1)] = b
                feed[input_tensor] = numpy.random.random_sample(shape)
            output_tensor = self.sess.graph.get_tensor_by_name(tensor_name)
            shape = self.sess.run([output_tensor], feed)[0].shape
            shapes.append(numpy.array(shape))

        compare01 = (shapes[0] == shapes[1])
        compare12 = (shapes[1] == shapes[2])

        if compare01.all() and compare12.all():
            return shape[0].tolist()

        if (compare01 == compare12).all():
            index = numpy.argwhere(compare01 == False).flatten()
            if index.shape[0] != 1:
                raise Exception("There's not only one unstable dimension")
            if index[0] != 0:
                raise Exception("Batch size not in the first dimension")
            shapes[0][0] = -1
            return shapes[0].tolist()