post_training_quantization.py 46.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
#   Copyright (c) 2018 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 math
15 16
import os
import re
17 18 19 20 21
import logging
import numpy as np
from .... import io
from .... import core
from .... import framework
22
from ....executor import global_scope, Executor
23 24 25 26 27
from ....framework import IrGraph
from ....log_helper import get_logger
from .quantization_pass import QuantizationTransformPass
from .quantization_pass import QuantizationFreezePass
from .quantization_pass import AddQuantDequantPass
28 29 30
from .quantization_pass import _out_scale_op_list
from .quantization_pass import _get_op_input_var_names
from .quantization_pass import _get_op_output_var_names
31

32
__all__ = ['PostTrainingQuantization', 'WeightQuantization']
33 34 35 36 37

_logger = get_logger(
    __name__, logging.INFO, fmt='%(asctime)s-%(levelname)s: %(message)s')


38 39 40 41
def _load_variable_data(scope, var_name):
    '''
    Load variable value from scope
    '''
42 43 44 45
    var_node = scope.find_var(var_name)
    assert var_node is not None, \
        "Cannot find " + var_name + " in scope."
    return np.array(var_node.get_tensor())
46 47 48 49 50 51 52 53 54 55 56 57 58 59


def _set_variable_data(scope, place, var_name, np_value):
    '''
    Set the value of var node by name, if the node exits,
    '''
    assert isinstance(np_value, np.ndarray), \
        'The type of value should be numpy array.'
    var_node = scope.find_var(var_name)
    if var_node != None:
        tensor = var_node.get_tensor()
        tensor.set(np_value, place)


60 61 62 63 64 65 66 67
def _all_persistable_var_names(program):
    persistable_var_names = []
    for var in program.list_vars():
        if var.persistable:
            persistable_var_names.append(var.name)
    return persistable_var_names


68
class PostTrainingQuantization(object):
69 70 71 72 73 74
    """
    Utilizing post training quantization methon to quantize the FP32 model,
    and it uses calibrate data to get the quantization information for all 
    quantized variables.
    """

75
    def __init__(self,
76 77 78
                 executor=None,
                 scope=None,
                 model_dir=None,
79 80
                 model_filename=None,
                 params_filename=None,
81
                 batch_generator=None,
82
                 sample_generator=None,
83 84 85
                 batch_size=10,
                 batch_nums=None,
                 algo="KL",
86
                 quantizable_op_type=["conv2d", "depthwise_conv2d", "mul"],
87
                 is_full_quantize=False,
88
                 activation_bits=8,
89 90 91
                 weight_bits=8,
                 activation_quantize_type='range_abs_max',
                 weight_quantize_type='channel_wise_abs_max',
92 93
                 is_use_cache_file=False,
                 cache_dir="./temp_post_training"):
94
        '''
95
        Constructor.
96 97

        Args:
98
            executor(fluid.Executor): The executor to load, run and save the
99
                quantized model.
100 101
            scope(fluid.Scope, optional): The scope of the program, use it to load 
                and save variables. If scope=None, get scope by global_scope(). 
102 103 104 105 106 107 108 109 110
            model_dir(str): The path of the fp32 model that will be quantized, 
                and the model and params files are under the path.
            model_filename(str, optional): The name of file to load the inference 
                program. If it is None, the default filename '__model__' will 
                be used. Default is 'None'.
            params_filename(str, optional): The name of file to load all parameters.
                When all parameters were saved in a single binary file, set it 
                as the real filename. If parameters were saved in separate files, 
                set it as 'None'. Default is 'None'.
111 112 113 114 115 116 117 118
            batch_generator(Python Generator): The batch generator provides 
                calibrate data for DataLoader, and it returns a batch every
                time. Note that, sample_generator and batch_generator, only one
                should be set. Beisdes, batch_generator supports lod tensor.
            sample_generator(Python Generator): The sample generator provides
                calibrate data for DataLoader, and it only returns a sample every
                time. Note that, sample_generator and batch_generator, only one
                should be set. Beisdes, sample_generator dose not support lod tensor.
119 120 121 122
            batch_size(int, optional): The batch size of DataLoader. Default is 10.
            batch_nums(int, optional): If batch_nums is not None, the number of 
                calibrate data is batch_size*batch_nums. If batch_nums is None, use 
                all data provided by sample_generator as calibrate data.
123 124 125 126 127
            algo(str, optional): If algo='KL', use KL-divergenc method to
                get the KL threshold for quantized activations and get the abs_max
                value for quantized weights. If algo='abs_max', get the abs max 
                value for activations and weights. If algo= 'min_max', get the min 
                and max value for quantized activations and weights. Default is KL.
128 129
            quantizable_op_type(list[str], optional): List the type of ops 
                that will be quantized. Default is ["conv2d", "depthwise_conv2d", 
130 131
                "mul"].
            is_full_quantized(bool, optional): If set is_full_quantized as True, 
132
                apply quantization to all supported quantizable op type. If set
133 134
                is_full_quantized as False, only apply quantization to the op type 
                according to the input quantizable_op_type.
135
            activation_bits(int): quantization bit number for activation.
136 137 138 139 140 141 142 143 144 145 146 147
            weight_bits(int, optional): quantization bit number for weights.
            activation_quantize_type(str): quantization type for activation,
                now support 'range_abs_max', 'moving_average_abs_max' and 'abs_max'.
                This param only specifies the fake ops in saving quantized model.
                If it is 'range_abs_max' or 'moving_average_abs_max', we save the scale
                obtained by post training quantization in fake ops. Note that, if it
                is 'abs_max', the scale will not be saved in fake ops.
            weight_quantize_type(str): quantization type for weights,
                support 'abs_max' and 'channel_wise_abs_max'. This param only specifies
                the fake ops in saving quantized model, and we save the scale obtained
                by post training quantization in fake ops. Compared to 'abs_max',
                the model accuracy is usually higher when it is 'channel_wise_abs_max'.
148 149 150 151 152 153 154
            is_use_cache_file(bool, optional): If set is_use_cache_file as False,
                all temp data will be saved in memory. If set is_use_cache_file as True,
                it will save temp data to disk. When the fp32 model is complex or
                the number of calibrate data is large, we should set is_use_cache_file
                as True. Defalut is False.
            cache_dir(str, optional): When is_use_cache_file is True, set cache_dir as
                the directory for saving temp data. Default is ./temp_post_training.
155 156 157
        Returns:
            None

158 159 160 161 162 163
        Examples:
        .. code-block:: python
            import paddle.fluid as fluid
            from paddle.fluid.contrib.slim.quantization import PostTrainingQuantization
            
            exe = fluid.Executor(fluid.CPUPlace())
164 165 166 167 168 169 170 171 172
            model_dir = path/to/fp32_model_params
            # set model_filename as None when the filename is __model__, 
            # otherwise set it as the real filename
            model_filename = None 
            # set params_filename as None when all parameters were saved in 
            # separate files, otherwise set it as the real filename
            params_filename = None
            save_model_path = path/to/save_model_path
            # prepare the sample generator according to the model, and the 
173
            # sample generator must return a sample every time. The reference
174 175 176
            # document: https://www.paddlepaddle.org.cn/documentation/docs/zh
            # /user_guides/howto/prepare_data/use_py_reader.html
            sample_generator = your_sample_generator
177 178 179
            batch_size = 10
            batch_nums = 10
            algo = "KL"
180
            quantizable_op_type = ["conv2d", "depthwise_conv2d", "mul"]
181 182
            ptq = PostTrainingQuantization(
                        executor=exe,
183 184 185 186
                        sample_generator=sample_generator,
                        model_dir=model_dir,
                        model_filename=model_filename,
                        params_filename=params_filename,
187 188 189 190 191 192 193
                        batch_size=batch_size,
                        batch_nums=batch_nums,
                        algo=algo,
                        quantizable_op_type=quantizable_op_type)
            ptq.quantize()
            ptq.save_quantized_model(save_model_path)
        '''
194

195 196 197 198 199 200 201 202 203 204
        self._support_activation_quantize_type = [
            'range_abs_max', 'moving_average_abs_max', 'abs_max'
        ]
        self._support_weight_quantize_type = ['abs_max', 'channel_wise_abs_max']
        self._support_algo_type = ['KL', 'abs_max', 'min_max']
        self._support_quantize_op_type = \
            list(set(QuantizationTransformPass._supported_quantizable_op_type +
                AddQuantDequantPass._supported_quantizable_op_type))

        # Check inputs
205 206
        assert executor is not None, "The executor cannot be None."
        assert model_dir is not None, "The model_dir cannot be None."
207 208 209 210 211
        assert any([gen is not None] for gen in [sample_generator,
            batch_generator]), "The sample_generator and batch_generator " \
            "cannot be None in the same time."
        assert batch_size > 0, "The batch_size should be greater than 0."
        assert algo in self._support_algo_type, \
212
            "The algo should be KL, abs_max or min_max."
213 214 215 216 217 218 219 220
        assert activation_quantize_type in self._support_activation_quantize_type, \
            "The activation_quantize_type ({}) should in ({}).".format(
            activation_quantize_type, self._support_activation_quantize_type)
        assert weight_quantize_type in self._support_weight_quantize_type, \
            "The weight_quantize_type ({}) shoud in ({}).".format(
            weight_quantize_type, self._support_weight_quantize_type)

        # Save input params
221
        self._executor = executor
222
        self._scope = global_scope() if scope == None else scope
223 224 225
        self._model_dir = model_dir
        self._model_filename = model_filename
        self._params_filename = params_filename
226
        self._sample_generator = sample_generator
227
        self._batch_generator = batch_generator
228 229 230
        self._batch_size = batch_size
        self._batch_nums = batch_nums
        self._algo = algo
231 232 233 234 235
        self._activation_bits = activation_bits
        self._weight_bits = weight_bits
        self._activation_quantize_type = activation_quantize_type
        self._weight_quantize_type = weight_quantize_type
        self._is_full_quantize = is_full_quantize
236
        if is_full_quantize:
237
            self._quantizable_op_type = self._support_quantize_op_type
238 239 240
        else:
            self._quantizable_op_type = quantizable_op_type
            for op_type in self._quantizable_op_type:
241
                assert op_type in self._support_quantize_op_type, \
242
                    op_type + " is not supported for quantization."
243 244 245 246
        self._is_use_cache_file = is_use_cache_file
        self._cache_dir = cache_dir
        if self._is_use_cache_file and not os.path.exists(self._cache_dir):
            os.mkdir(self._cache_dir)
247

248
        # Define variables
249 250 251 252 253 254
        self._place = self._executor.place
        self._program = None
        self._feed_list = None
        self._fetch_list = None
        self._data_loader = None

255
        self._out_scale_op_list = _out_scale_op_list
256 257
        self._quantized_weight_var_name = set()
        self._quantized_act_var_name = set()
258
        self._sampling_data = {}
259 260 261 262
        self._quantized_var_kl_threshold = {}
        self._quantized_var_min = {}
        self._quantized_var_max = {}
        self._quantized_var_abs_max = {}
263 264 265

    def quantize(self):
        '''
266 267 268
        Load the FP32 model, and use the calibrate data to calculate the forward-stage.
        Based on the sample data, we can get the quantization information, and obtain
        the final quantized model.
269 270 271 272

        Args:
            None
        Returns:
273 274
            the program of quantized model.
        '''
275
        self._load_model_data()
276
        self._collect_target_varnames()
277
        self._set_activation_persistable()
278 279 280 281 282

        batch_id = 0
        for data in self._data_loader():
            self._executor.run(program=self._program,
                               feed=data,
283 284
                               fetch_list=self._fetch_list,
                               return_numpy=False)
285 286 287 288
            if self._algo == "KL":
                self._sample_data(batch_id)
            else:
                self._sample_threshold()
289

290
            if batch_id % 5 == 0:
291
                _logger.info("Run batch: " + str(batch_id))
292 293 294
            batch_id += 1
            if self._batch_nums and batch_id >= self._batch_nums:
                break
295
        _logger.info("Finish all batch: " + str(batch_id))
296

297
        self._reset_activation_persistable()
298

299 300
        if self._algo == "KL":
            self._calculate_kl_threshold()
301

302 303 304 305 306 307
        if self._algo in ["KL", "abs_max"]:
            self._update_program()
        else:
            self._save_input_threhold()

        self._save_output_threshold()
308 309
        return self._program

310 311 312 313
    def save_quantized_model(self,
                             save_model_path,
                             model_filename=None,
                             params_filename=None):
314 315 316 317
        '''
        Save the quantized model to the disk.

        Args:
318 319 320 321 322 323 324
            save_model_path(str): The path to save the quantized model.
            model_filename(str, optional): If the model_filename is None,
                save the model to '__model__'. Otherwise, save the model
                to the specified filename. Default: None.
            params_filename(str, optional): If the params_filename is None,
                save params to separted files. Otherwise, save all params
                to the specified filename.
325
        Returns:
326 327 328 329
            None
        '''
        io.save_inference_model(
            dirname=save_model_path,
330 331
            model_filename=model_filename,
            params_filename=params_filename,
332 333 334 335 336
            feeded_var_names=self._feed_list,
            target_vars=self._fetch_list,
            executor=self._executor,
            main_program=self._program)

337
    def _load_model_data(self):
338
        '''
339
        Load model and set data loader.
340
        '''
341
        _logger.info("Load model and set data loader ...")
342
        [self._program, self._feed_list, self._fetch_list] = \
343 344 345 346
            io.load_inference_model(dirname=self._model_dir,
                                    executor=self._executor,
                                    model_filename=self._model_filename,
                                    params_filename=self._params_filename)
347 348 349 350
        feed_vars = [framework._get_var(str(var_name), self._program) \
            for var_name in self._feed_list]
        self._data_loader = io.DataLoader.from_generator(
            feed_list=feed_vars, capacity=3 * self._batch_size, iterable=True)
351 352 353 354 355 356 357 358 359 360 361
        if self._sample_generator is not None:
            self._data_loader.set_sample_generator(
                self._sample_generator,
                batch_size=self._batch_size,
                drop_last=True,
                places=self._place)
        elif self._batch_generator is not None:
            self._data_loader.set_batch_generator(
                self._batch_generator, places=self._place)

    def _collect_target_varnames(self):
362 363 364 365
        '''
        Collect the variable names for sampling, and set activation
        variables to be persistable.
        '''
366
        # TODO(juncaipeng), consider the name_scope of skip_quant
367
        _logger.info("Collect quantized variable names ...")
368 369 370 371 372 373 374 375

        def collect_var_name(var_name_list, persistable_var_names):
            for var_name in var_name_list:
                if var_name in persistable_var_names:
                    self._quantized_weight_var_name.add(var_name)
                else:
                    self._quantized_act_var_name.add(var_name)

376
        persistable_var_names = _all_persistable_var_names(self._program)
377
        for op in self._program.global_block().ops:
378
            op_type = op.type
379
            # For quantized ops, sample inputs and outputs
380
            if op_type in self._quantizable_op_type:
381 382 383 384 385 386 387 388
                collect_var_name(
                    _get_op_input_var_names(op), persistable_var_names)
                collect_var_name(
                    _get_op_output_var_names(op), persistable_var_names)
            # For other op, only sample output scale
            elif op_type in self._out_scale_op_list:
                collect_var_name(
                    _get_op_output_var_names(op), persistable_var_names)
389 390 391 392 393 394

    def _set_activation_persistable(self):
        '''
        Set activation variables to be persistable, so can obtain 
        the tensor data in sample_data
        '''
395 396 397 398
        for var in self._program.list_vars():
            if var.name in self._quantized_act_var_name:
                var.persistable = True

399 400 401 402 403 404 405 406 407 408 409 410 411 412
    def _reset_activation_persistable(self):
        '''
        Reset activations to be not persistable.
        '''
        for var in self._program.list_vars():
            if var.name in self._quantized_act_var_name:
                var.persistable = False

    def _sample_threshold(self):
        '''
        Sample the input threshold(min, max, or abs_max) in every iterations.
        '''
        assert self._algo in ["abs_max", "min_max"], \
            "The algo should be abs_max or min_max to sample min max value."
413

414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
        if self._algo == "abs_max":
            # Only calculate abs_max value for weight for once
            if self._quantized_var_abs_max == {}:
                for var_name in self._quantized_weight_var_name:
                    var_tensor = _load_variable_data(self._scope, var_name)
                    abs_max_per_channel = []
                    for i in range(var_tensor.shape[0]):
                        abs_max_per_channel.append(
                            float(np.max(np.abs(var_tensor[i]))))
                    self._quantized_var_abs_max[var_name] = abs_max_per_channel
            for var_name in self._quantized_act_var_name:
                var_tensor = _load_variable_data(self._scope, var_name)
                abs_max_value = float(np.max(np.abs(var_tensor)))
                if (var_name not in self._quantized_var_abs_max) or \
                    (abs_max_value > self._quantized_var_abs_max[var_name]):
                    self._quantized_var_abs_max[var_name] = abs_max_value
        elif self._algo == "min_max":
            if self._quantized_var_min == {} and self._quantized_var_max == {}:
                for var_name in self._quantized_weight_var_name:
                    var_tensor = _load_variable_data(self._scope, var_name)
                    min_per_channel = []
                    max_per_channle = []
                    for i in range(var_tensor.shape[0]):
                        min_per_channel.append(float(np.min(var_tensor[i])))
                        max_per_channle.append(float(np.max(var_tensor[i])))
                    self._quantized_var_min[var_name] = min_per_channel
                    self._quantized_var_max[var_name] = max_per_channle
            for var_name in self._quantized_act_var_name:
                var_tensor = _load_variable_data(self._scope, var_name)
                min_value = float(np.min(var_tensor))
                max_value = float(np.max(var_tensor))
                if (var_name not in self._quantized_var_min) or \
                    (min_value < self._quantized_var_min[var_name]):
                    self._quantized_var_min[var_name] = min_value
                if (var_name not in self._quantized_var_max) or \
                    (max_value > self._quantized_var_max[var_name]):
                    self._quantized_var_max[var_name] = max_value

    def _save_input_threhold(self):
        '''
        Save input threshold to the quantized op.
        '''
        assert self._algo == "min_max", \
            "The algo should be min_max to save input threshold."
        for op in self._program.global_block().ops:
            if op.type in self._quantizable_op_type:
460 461 462 463 464 465 466
                for var_name in _get_op_input_var_names(op):
                    assert var_name in self._quantized_var_min
                    assert var_name in self._quantized_var_max
                    op._set_attr(var_name + ".min",
                                 self._quantized_var_min[var_name])
                    op._set_attr(var_name + ".max",
                                 self._quantized_var_max[var_name])
467

468
    def _sample_data(self, iter):
469 470 471 472
        '''
        Sample the tensor data of quantized variables, 
        applied in every iteration.
        '''
473
        assert self._algo == "KL", "The algo should be KL to sample data."
474 475
        for var_name in self._quantized_weight_var_name:
            if var_name not in self._sampling_data:
476
                var_tensor = _load_variable_data(self._scope, var_name)
477 478
                self._sampling_data[var_name] = var_tensor

479 480
        if self._is_use_cache_file:
            for var_name in self._quantized_act_var_name:
481
                var_tensor = _load_variable_data(self._scope, var_name)
482 483 484 485 486 487 488 489
                var_tensor = var_tensor.ravel()
                save_path = os.path.join(self._cache_dir,
                                         var_name + "_" + str(iter) + ".npy")
                np.save(save_path, var_tensor)
        else:
            for var_name in self._quantized_act_var_name:
                if var_name not in self._sampling_data:
                    self._sampling_data[var_name] = []
490
                var_tensor = _load_variable_data(self._scope, var_name)
491 492
                var_tensor = var_tensor.ravel()
                self._sampling_data[var_name].append(var_tensor)
493

494
    def _calculate_kl_threshold(self):
495
        '''
496
        Calculate the KL threshold of quantized variables.
497
        '''
498 499
        _logger.info("Calculate KL threshold ...")
        assert self._algo == "KL", "The algo should be KL to calculate kl threshold."
500 501

        # Abs_max threshold for weights
502
        for var_name in self._quantized_weight_var_name:
503 504 505 506 507 508 509 510 511 512 513 514
            weight_data = self._sampling_data[var_name]
            weight_threshold = None
            if self._weight_quantize_type == "abs_max":
                weight_threshold = np.max(np.abs(weight_data))
            elif self._weight_quantize_type == "channel_wise_abs_max":
                weight_threshold = []
                for i in range(weight_data.shape[0]):
                    abs_max_value = np.max(np.abs(weight_data[i]))
                    weight_threshold.append(abs_max_value)
            self._quantized_var_kl_threshold[var_name] = weight_threshold

        # KL threshold for activations
515 516 517 518 519 520 521 522 523 524
        if self._is_use_cache_file:
            for var_name in self._quantized_act_var_name:
                sampling_data = []
                filenames = [f for f in os.listdir(self._cache_dir) \
                    if re.match(var_name + '_[0-9]+.npy', f)]
                for filename in filenames:
                    file_path = os.path.join(self._cache_dir, filename)
                    sampling_data.append(np.load(file_path))
                    os.remove(file_path)
                sampling_data = np.concatenate(sampling_data)
525 526
                self._quantized_var_kl_threshold[var_name] = \
                    self._get_kl_scaling_factor(np.abs(sampling_data))
527 528 529 530
        else:
            for var_name in self._quantized_act_var_name:
                self._sampling_data[var_name] = np.concatenate(
                    self._sampling_data[var_name])
531 532
                self._quantized_var_kl_threshold[var_name] = \
                    self._get_kl_scaling_factor(np.abs(self._sampling_data[var_name]))
533 534 535

    def _update_program(self):
        '''
536 537 538
        Use QuantizationTransformPass and AddQuantDequantPass to insert 
        fake_quantize, fake_dequantize and fake_quant_dequant op. 
        Besides, save all kl threshold to the scale var node.
539
        '''
540
        _logger.info("Update the program ...")
541 542
        graph = IrGraph(core.Graph(self._program.desc), for_test=True)

543
        # use QuantizationTransformPass to insert fake_quant/fake_dequantize op
544 545
        major_quantizable_op_types = []
        for op_type in QuantizationTransformPass._supported_quantizable_op_type:
546
            if op_type in self._quantizable_op_type:
547
                major_quantizable_op_types.append(op_type)
548 549 550
        transform_pass = QuantizationTransformPass(
            scope=self._scope,
            place=self._place,
551 552 553 554
            weight_bits=self._weight_bits,
            activation_bits=self._activation_bits,
            activation_quantize_type=self._activation_quantize_type,
            weight_quantize_type=self._weight_quantize_type,
555
            quantizable_op_type=major_quantizable_op_types)
556 557 558
        transform_pass.apply(graph)

        # use AddQuantDequantPass to insert fake_quant_dequant op
559 560
        minor_quantizable_op_types = []
        for op_type in AddQuantDequantPass._supported_quantizable_op_type:
561
            if op_type in self._quantizable_op_type:
562
                minor_quantizable_op_types.append(op_type)
563 564 565
        add_quant_dequant_pass = AddQuantDequantPass(
            scope=self._scope,
            place=self._place,
566
            quantizable_op_type=minor_quantizable_op_types)
567 568
        add_quant_dequant_pass.apply(graph)

569 570 571 572 573 574
        # save abs_max or KL threshold to scale var node
        if self._algo == "KL":
            scale_dict = self._quantized_var_kl_threshold
        else:
            scale_dict = self._quantized_var_abs_max
        for key, val in scale_dict.items():
575 576 577 578 579
            _set_variable_data(
                self._scope,
                self._place,
                key + ".scale",
                np.array(
580
                    [val], dtype=np.float32))
581 582 583 584 585
            _set_variable_data(
                self._scope,
                self._place,
                key + ".quant_dequant.scale",
                np.array(
586 587 588 589 590 591
                    [val], dtype=np.float32))

        # apply QuantizationFreezePass, and obtain the final quant model
        freeze_pass = QuantizationFreezePass(
            scope=self._scope,
            place=self._place,
592 593 594
            weight_bits=self._weight_bits,
            activation_bits=self._activation_bits,
            weight_quantize_type=self._weight_quantize_type,
595
            quantizable_op_type=major_quantizable_op_types)
596 597 598
        freeze_pass.apply(graph)
        self._program = graph.to_program()

599
    def _save_output_threshold(self):
600
        '''
601
        Save output threshold to the quantized op.
602
        '''
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626

        def save_info(op_node, out_var_name, threshold_map, out_info_name,
                      quantized_type):
            assert out_var_name in threshold_map, \
                "The output ({}) of {} node does not have threshold.".format(
                out_var_name, op_node.type)
            op_node._set_attr(out_info_name, threshold_map[var_name])
            if op_node.type in self._quantizable_op_type:
                op._set_attr("quantization_type", quantized_type)

        def analysis_and_save_info(op_node, out_var_name):
            if self._algo == "KL":
                save_info(op_node, out_var_name,
                          self._quantized_var_kl_threshold, "out_threshold",
                          "post_kl")
            elif self._algo == "abs_max":
                save_info(op_node, out_var_name, self._quantized_var_abs_max,
                          "out_threshold", "post_abs_max")
            elif self._algo == "min_max":
                save_info(op_node, out_var_name, self._quantized_var_min,
                          "out_min", "post_min_max")
                save_info(op_node, out_var_name, self._quantized_var_max,
                          "out_max", "post_min_max")

627
        for op in self._program.global_block().ops:
628 629 630 631 632 633
            if op.type in (self._quantizable_op_type + self._out_scale_op_list):
                out_var_names = _get_op_output_var_names(op)
                assert len(out_var_names) == 1, "Post training " + \
                    "quantization only support one output for " + op.type
                for var_name in out_var_names:
                    analysis_and_save_info(op, var_name)
634

635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735
    def _get_kl_scaling_factor(self, activation_blob, num_quantized_bins=255):
        '''
        Using the KL-divergenc method to get the more precise scaling factor.
        '''
        max_val = np.max(activation_blob)
        min_val = np.min(activation_blob)
        if min_val >= 0:
            hist, hist_edeges = np.histogram(
                activation_blob, bins=2048, range=(min_val, max_val))
            ending_iter = 2047
            starting_iter = int(ending_iter * 0.7)
        else:
            _logger.error("Please first apply abs to activation_blob.")
        bin_width = hist_edeges[1] - hist_edeges[0]

        P_sum = len(np.array(activation_blob).ravel())
        min_kl_divergence = 0
        min_kl_index = 0
        kl_inited = False
        for i in range(starting_iter, ending_iter + 1):
            reference_distr_P = hist[0:i].tolist()
            outliers_count = sum(hist[i:2048])
            if reference_distr_P[i - 1] == 0:
                continue
            reference_distr_P[i - 1] += outliers_count
            reference_distr_bins = reference_distr_P[:]
            candidate_distr_Q = hist[0:i].tolist()
            num_merged_bins = int(i / num_quantized_bins)
            candidate_distr_Q_quantized = [0] * num_quantized_bins
            j_start = 0
            j_end = num_merged_bins
            for idx in range(num_quantized_bins):
                candidate_distr_Q_quantized[idx] = sum(candidate_distr_Q[
                    j_start:j_end])
                j_start += num_merged_bins
                j_end += num_merged_bins
                if (idx + 1) == num_quantized_bins - 1:
                    j_end = i
            candidate_distr_Q = self._expand_quantized_bins(
                candidate_distr_Q_quantized, reference_distr_bins)
            Q_sum = sum(candidate_distr_Q)
            kl_divergence = self._safe_entropy(reference_distr_P, P_sum,
                                               candidate_distr_Q, Q_sum)
            if not kl_inited:
                min_kl_divergence = kl_divergence
                min_kl_index = i
                kl_inited = True
            elif kl_divergence < min_kl_divergence:
                min_kl_divergence = kl_divergence
                min_kl_index = i
            else:
                pass
        if min_kl_index == 0:
            while starting_iter > 0:
                if hist[starting_iter] == 0:
                    starting_iter -= 1
                    continue
                else:
                    break
            min_kl_index = starting_iter
        return (min_kl_index + 0.5) * bin_width

    def _expand_quantized_bins(self, quantized_bins, reference_bins):
        '''
        '''
        expanded_quantized_bins = [0] * len(reference_bins)
        num_merged_bins = int(len(reference_bins) / len(quantized_bins))
        j_start = 0
        j_end = num_merged_bins
        for idx in range(len(quantized_bins)):
            zero_count = reference_bins[j_start:j_end].count(0)
            num_merged_bins = j_end - j_start
            if zero_count == num_merged_bins:
                avg_bin_ele = 0
            else:
                avg_bin_ele = quantized_bins[idx] / (
                    num_merged_bins - zero_count + 0.0)
            for idx1 in range(j_start, j_end):
                expanded_quantized_bins[idx1] = (0 if reference_bins[idx1] == 0
                                                 else avg_bin_ele)
            j_start += num_merged_bins
            j_end += num_merged_bins
            if (idx + 1) == len(quantized_bins) - 1:
                j_end = len(reference_bins)
        return expanded_quantized_bins

    def _safe_entropy(self, reference_distr_P, P_sum, candidate_distr_Q, Q_sum):
        '''
        Calculate the entropy.
        '''
        assert len(reference_distr_P) == len(candidate_distr_Q)
        tmp_sum1 = 0
        tmp_sum2 = 0
        for idx in range(len(reference_distr_P)):
            p_idx = reference_distr_P[idx]
            q_idx = candidate_distr_Q[idx]
            if p_idx == 0:
                tmp_sum1 += 0
                tmp_sum2 += 0
            else:
                if q_idx == 0:
736 737
                    _logger.error("Fatal error!, idx = " + str(idx) +
                                  " qindex = 0! p_idx = " + str(p_idx))
738 739 740
                tmp_sum1 += p_idx * (math.log(Q_sum * p_idx))
                tmp_sum2 += p_idx * (math.log(P_sum * q_idx))
        return (tmp_sum1 - tmp_sum2) / P_sum
741 742 743 744


class WeightQuantization(object):
    _supported_quantizable_op_type = ['conv2d', 'depthwise_conv2d', 'mul']
745
    _supported_weight_quantize_type = ['channel_wise_abs_max', 'abs_max']
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771

    def __init__(self, model_dir, model_filename=None, params_filename=None):
        '''
        This class quantizes the weight of some ops to reduce the size of model
        or improve the perforemace.

        Args:
            model_dir(str): The path of the fp32 model that will be quantized,
                and the model and params files are under the path.
            model_filename(str, optional): The name of file to load the inference
                program. If it is None, the default filename '__model__' will
                be used. Default is 'None'.
            params_filename(str, optional): The name of file to load all parameters.
                When all parameters were saved in a single binary file, set it
                as the real filename. If parameters were saved in separate files,
                set it as 'None'. Default is 'None'.
        '''
        self._model_dir = model_dir
        self._model_filename = model_filename
        self._params_filename = params_filename

    def quantize_weight_to_int(self,
                               save_model_dir,
                               save_model_filename=None,
                               save_params_filename=None,
                               quantizable_op_type=["conv2d", "mul"],
772
                               weight_bits=8,
773 774
                               weight_quantize_type="channel_wise_abs_max",
                               generate_test_model=False,
775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793
                               threshold_rate=0.0):
        '''
        In order to reduce the size of model, this api quantizes the weight
        of some ops from float32 to int8/16. In the inference stage, the 
        quantized weight will be dequantized to float32 again.
        
        Args:
            save_model_dir(str): The path to save the quantized model.
            save_model_filename(str, optional): The name of file to 
                save the inference program. If it is None, the default 
                filename '__model__' will be used. Default is 'None'.
            save_params_filename(str, optional): The name of file to 
                save all parameters. If it is None, parameters were 
                saved in separate files. If it is not None, all 
                parameters were saved in a single binary file.
            quantizable_op_type(list[str], optional): The list of ops 
                that will be quantized, and the quantized ops should be
                contained in ["conv2d", "depthwise_conv2d", "mul"]. 
                Default is ["conv2d","mul"].
794 795
            weight_bits(int, optional): The bits for the quantized weight, 
                and it should be 8 or 16. Default is 8.
796 797 798 799 800 801 802
            weight_quantize_type(str, optional): quantization type for weights,
                support 'channel_wise_abs_max' and 'abs_max'. Set it as
                'channel_wise_abs_max', the accuracy performs better.
            generate_test_model(bool, optional): If set generate_test_model 
                as True, it saves a fake quantized model, in which the weights 
                are quantized and dequantized. We can use PaddlePaddle to load 
                the fake quantized model and test the accuracy on GPU or CPU.
803 804 805 806 807 808 809 810 811
            threshold_rate(float, optional): This api uses abs_max methd to 
                quantize the weight from float32 to int8/16, and the abs max 
                value is important for quantization diff. When the abs_max 
                value is far away from the center of the numerical distribution, 
                we can set threshold_rate between 1e-6 and 1e-8, so the abs max 
                value will be optimized. Default is 0.0.
        '''
        for op_type in quantizable_op_type:
            assert op_type in self._supported_quantizable_op_type, \
812
                "Input error:" + op_type + \
813
                " is not supported for weight quantization."
814
        assert weight_bits in [8, 16], \
815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840
            "Input error: weight_bits should be 8 or 16."
        assert weight_quantize_type in self._supported_weight_quantize_type, \
            "Input error: weight_quantize_type should in {}".format(
                self._supported_weight_quantize_type)

        quantized_model_dir = os.path.join(save_model_dir, "quantized_model")
        self._quantize_weight_to_int(quantized_model_dir, save_model_filename,
                                     save_params_filename, quantizable_op_type,
                                     weight_bits, weight_quantize_type, False,
                                     threshold_rate)

        if generate_test_model:
            test_model_dir = os.path.join(save_model_dir, "test_model")
            self._quantize_weight_to_int(
                test_model_dir, save_model_filename, save_params_filename,
                quantizable_op_type, weight_bits, weight_quantize_type, True,
                threshold_rate)

    def _quantize_weight_to_int(self, save_model_dir, save_model_filename,
                                save_params_filename, quantizable_op_type,
                                weight_bits, weight_quantize_type, for_test,
                                threshold_rate):
        """
        Generate quantized model or fake quantized model.
        """
        # Load model
841 842 843 844 845 846 847 848 849
        place = core.CPUPlace()
        exe = Executor(place)
        scope = global_scope()
        [program, feed_list, fetch_list] = \
            io.load_inference_model(dirname=self._model_dir,
                                    executor=exe,
                                    model_filename=self._model_filename,
                                    params_filename=self._params_filename)

850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868
        quantized_ops = []
        for index in range(program.num_blocks):
            block = program.block(index)
            for op in block.ops:
                if op.type in quantizable_op_type:
                    quantized_ops.append(op)

        # Quantize weights
        persistable_var_names = _all_persistable_var_names(program)
        for op in quantized_ops:
            for var_name in op.input_arg_names:
                if var_name in persistable_var_names:
                    if weight_quantize_type == "abs_max":
                        self._weight_abs_max_quantization(
                            scope, place, weight_bits, threshold_rate, op,
                            var_name, for_test)
                    elif weight_quantize_type == "channel_wise_abs_max":
                        self._weight_channel_wise_abs_max_quantization(
                            scope, place, weight_bits, op, var_name, for_test)
869 870 871 872 873 874 875 876 877 878

        io.save_inference_model(
            dirname=save_model_dir,
            feeded_var_names=feed_list,
            target_vars=fetch_list,
            executor=exe,
            main_program=program,
            model_filename=save_model_filename,
            params_filename=save_params_filename)

879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
    def _weight_abs_max_quantization(self, scope, place, weight_bits,
                                     threshold_rate, op, var_name, for_test):
        '''
        Use abs_max method to quantize weight.
        '''
        quantize_range = (1 << (weight_bits - 1)) - 1
        save_weight_dtype = np.int8 if weight_bits == 8 else np.int16

        # Get quantized scale and weight data
        weight_data = _load_variable_data(scope, var_name)
        if abs(threshold_rate) < 1e-10:
            threshold_value = np.max(np.abs(weight_data))
        else:
            threshold_value = self._calculate_threshold(\
                weight_data, threshold_rate)
            weight_data[weight_data > threshold_value] = threshold_value
            weight_data[weight_data < -threshold_value] = -threshold_value
        scale = threshold_value / quantize_range
        quantized_weight_data = \
            np.around(weight_data / scale).astype(save_weight_dtype)

        # Set weight data
        if not for_test:
            _set_variable_data(scope, place, var_name, quantized_weight_data)
        else:
            dequantized_weight_data = \
                (quantized_weight_data * scale).astype(np.float32)
            _set_variable_data(scope, place, var_name, dequantized_weight_data)

        # Save info
        op._set_attr('quantization_type', 'post_weight_abs_max')
        op._set_attr('quantize_weight_bits', weight_bits)
        op._set_attr(var_name + "_quant_scale", [scale])  # Save as list

    def _weight_channel_wise_abs_max_quantization(
            self, scope, place, weight_bits, op, var_name, for_test):
        ''' 
        Use channel_wise_abs_max method to quantize weight.
        '''
        quantize_range = (1 << (weight_bits - 1)) - 1
        save_weight_dtype = np.int8 if weight_bits == 8 else np.int16

        # Get quantized scale and weight data
        weight_data = _load_variable_data(scope, var_name)
        if op.type == "mul":
            scales, quantized_weight_data = \
                self._mul_channel_wise_quantization(weight_data,
                    quantize_range, save_weight_dtype)
        elif op.type in ["conv2d", "depthwise_conv2d"]:
            scales, quantized_weight_data = \
                self._conv_channel_wise_quantization(weight_data,
                    quantize_range, save_weight_dtype)
        else:
            _logger.error(op.type + " is not supported by weight quantization")

        # Set weight data
        if not for_test:
            _set_variable_data(scope, place, var_name, quantized_weight_data)
        else:
            if op.type == "mul":
                dequantized_weight_data = \
                    self._mul_channel_wise_dequantization(quantized_weight_data, scales)
            elif op.type in ["conv2d", "depthwise_conv2d"]:
                dequantized_weight_data = \
                    self._conv_channel_wise_dequantization(quantized_weight_data, scales)
            else:
                _logger.error(op.type +
                              " is not supported by weight quantization")
            _set_variable_data(scope, place, var_name, dequantized_weight_data)

        # Save info
        op._set_attr('quantization_type', 'post_weight_channel_wise_abs_max')
        op._set_attr('quantize_weight_bits', weight_bits)
        op._set_attr(var_name + "_quant_scale", scales)

    def _conv_channel_wise_quantization(self, weight_data, quantize_range,
                                        save_weight_dtype):
        '''
        Get channel wise scale for the weights of conv2d and depthwise_conv2d,
        and quantize the weights.
        '''
        scales = []
        quantized_weight_data = np.zeros_like(
            weight_data, dtype=save_weight_dtype)
        channel_num = weight_data.shape[0]
        for i in range(channel_num):
            scale = np.max(np.abs(weight_data[i])) / quantize_range
            scales.append(scale)
            quantized_weight_data[i] = \
                np.around(weight_data[i] / scale).astype(save_weight_dtype)
        return scales, quantized_weight_data

    def _conv_channel_wise_dequantization(self, quantized_weight_data, scales):
        '''
        For conv2d and depthwise_conv2d, dequantize the weights to fp32.
        '''
        dequantized_weight_data = np.zeros_like(
            quantized_weight_data, dtype=np.float32)
        for i in range(len(scales)):
            dequantized_weight_data[i] = \
                (quantized_weight_data[i] * scales[i]).astype(np.float32)
        return dequantized_weight_data

    def _mul_channel_wise_quantization(self, weight_data, quantize_range,
                                       save_weight_dtype):
        '''
        Get channel wise scale for the weights of conv2d and depthwise_conv2d,
        and quantize the weights.
        '''
        scales = []
        quantized_weight_data = np.zeros_like(
            weight_data, dtype=save_weight_dtype)
        channel_num = weight_data.shape[-1]
        for i in range(channel_num):
            scale = np.max(np.abs(weight_data[:, i])) / quantize_range
            scales.append(scale)
            quantized_weight_data[:, i] = \
                np.around(weight_data[:, i] / scale).astype(save_weight_dtype)
        return scales, quantized_weight_data

    def _mul_channel_wise_dequantization(self, quantized_weight_data, scales):
        '''
        For mul, dequantize the weights to fp32.
        '''
        dequantized_weight_data = np.zeros_like(
            quantized_weight_data, dtype=np.float32)
        for i in range(len(scales)):
            dequantized_weight_data[:, i] = \
                (quantized_weight_data[:, i] * scales[i]).astype(np.float32)
        return dequantized_weight_data

1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023
    def _calculate_threshold(self, input, threshold_rate, histogram_bins=5000):
        input_abs = np.abs(input)
        hist, hist_edeges = np.histogram(
            input_abs, bins=histogram_bins, range=(0, np.max(input_abs)))
        hist = hist / float(sum(hist))
        hist_sum = 0
        hist_index = 0
        for i in range(len(hist)):
            hist_sum += hist[i]
            if hist_sum >= 1.0 - threshold_rate:
                hist_index = i + 1
                break
        bin_width = hist_edeges[1] - hist_edeges[0]
        return hist_index * bin_width