post_training_quantization.py 24.5 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 22 23 24 25 26 27
import logging
import numpy as np
from ....executor import global_scope
from .... import io
from .... import core
from .... import framework
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
from .quantization_pass import _op_real_in_out_name
29 30 31 32 33 34 35 36 37 38

__all__ = ['PostTrainingQuantization']

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


class PostTrainingQuantization(object):
    def __init__(self,
                 executor,
39 40 41 42
                 sample_generator,
                 model_dir,
                 model_filename=None,
                 params_filename=None,
43 44 45 46
                 batch_size=10,
                 batch_nums=None,
                 scope=None,
                 algo="KL",
47
                 quantizable_op_type=["conv2d", "depthwise_conv2d", "mul"],
48 49 50
                 is_full_quantize=False,
                 is_use_cache_file=False,
                 cache_dir="./temp_post_training"):
51 52 53 54 55 56 57 58 59
        '''
        The class utilizes post training quantization methon to quantize the 
        fp32 model. It uses calibrate data to calculate the scale factor of 
        quantized variables, and inserts fake quant/dequant op to obtain the 
        quantized model.

        Args:
            executor(fluid.Executor): The executor to load, run and save the 
                quantized model.
60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
            sample_generator(Python Generator): The sample generator provides 
                calibrate data for DataLoader, and it only returns a sample every 
                time.
            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'.
            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.
76 77 78 79 80 81 82
            scope(fluid.Scope, optional): The scope of the program, use it to load 
                and save variables. If scope=None, get scope by global_scope(). 
            algo(str, optional): If algo=KL, use KL-divergenc method to 
                get the more precise scale factor. If algo='direct', use 
                abs_max methon to get the scale factor. Default is KL.
            quantizable_op_type(list[str], optional): List the type of ops 
                that will be quantized. Default is ["conv2d", "depthwise_conv2d", 
83 84
                "mul"].
            is_full_quantized(bool, optional): If set is_full_quantized as True, 
85
                apply quantization to all supported quantizable op type. If set
86 87
                is_full_quantized as False, only apply quantization to the op type 
                according to the input quantizable_op_type.
88 89 90 91 92 93 94
            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.
95 96 97
        Returns:
            None

98 99 100 101 102 103
        Examples:
        .. code-block:: python
            import paddle.fluid as fluid
            from paddle.fluid.contrib.slim.quantization import PostTrainingQuantization
            
            exe = fluid.Executor(fluid.CPUPlace())
104 105 106 107 108 109 110 111 112
            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 
113
            # sample generator must return a sample every time. The reference
114 115 116
            # document: https://www.paddlepaddle.org.cn/documentation/docs/zh
            # /user_guides/howto/prepare_data/use_py_reader.html
            sample_generator = your_sample_generator
117 118 119
            batch_size = 10
            batch_nums = 10
            algo = "KL"
120
            quantizable_op_type = ["conv2d", "depthwise_conv2d", "mul"]
121 122
            ptq = PostTrainingQuantization(
                        executor=exe,
123 124 125 126
                        sample_generator=sample_generator,
                        model_dir=model_dir,
                        model_filename=model_filename,
                        params_filename=params_filename,
127 128 129 130 131 132 133 134
                        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)
        '''
        self._executor = executor
135 136 137 138
        self._sample_generator = sample_generator
        self._model_dir = model_dir
        self._model_filename = model_filename
        self._params_filename = params_filename
139 140 141 142
        self._batch_size = batch_size
        self._batch_nums = batch_nums
        self._scope = global_scope() if scope == None else scope
        self._algo = algo
143 144 145 146
        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)
147 148 149 150 151 152 153 154 155 156 157 158

        supported_quantizable_op_type = \
            QuantizationTransformPass._supported_quantizable_op_type + \
            AddQuantDequantPass._supported_quantizable_op_type
        if is_full_quantize:
            self._quantizable_op_type = supported_quantizable_op_type
        else:
            self._quantizable_op_type = quantizable_op_type
            for op_type in self._quantizable_op_type:
                assert op_type in supported_quantizable_op_type + \
                    AddQuantDequantPass._activation_type, \
                    op_type + " is not supported for quantization."
159 160 161 162 163 164 165

        self._place = self._executor.place
        self._program = None
        self._feed_list = None
        self._fetch_list = None
        self._data_loader = None

166
        self._op_real_in_out_name = _op_real_in_out_name
167
        self._bit_length = 8
168 169
        self._quantized_weight_var_name = set()
        self._quantized_act_var_name = set()
170 171 172 173 174 175 176 177
        self._sampling_data = {}
        self._quantized_var_scale_factor = {}

    def quantize(self):
        '''
        Quantize the fp32 model. Use calibrate data to calculate the scale factor of 
        quantized variables, and inserts fake quant/dequant op to obtain the 
        quantized model.
178 179 180 181

        Args:
            None
        Returns:
182 183
            the program of quantized model.
        '''
184
        self._preprocess()
185 186 187 188 189

        batch_id = 0
        for data in self._data_loader():
            self._executor.run(program=self._program,
                               feed=data,
190 191
                               fetch_list=self._fetch_list,
                               return_numpy=False)
192 193
            self._sample_data(batch_id)

194 195 196 197 198 199 200
            if batch_id % 5 == 0:
                _logger.info("run batch: " + str(batch_id))
            batch_id += 1
            if self._batch_nums and batch_id >= self._batch_nums:
                break
        _logger.info("all run batch: " + str(batch_id))

201
        _logger.info("calculate scale factor ...")
202
        self._calculate_scale_factor()
203 204

        _logger.info("update the program ...")
205 206
        self._update_program()

207
        self._save_output_scale()
208 209 210 211 212 213 214 215
        return self._program

    def save_quantized_model(self, save_model_path):
        '''
        Save the quantized model to the disk.

        Args:
            save_model_path(str): The path to save the quantized model
216
        Returns:
217 218 219 220 221 222 223 224 225
            None
        '''
        io.save_inference_model(
            dirname=save_model_path,
            feeded_var_names=self._feed_list,
            target_vars=self._fetch_list,
            executor=self._executor,
            main_program=self._program)

226
    def _preprocess(self):
227 228 229 230 231 232
        '''
        Load model and set data loader, collect the variable names for sampling, 
        and set activation variables to be persistable.
        '''
        # load model and set data loader
        [self._program, self._feed_list, self._fetch_list] = \
233 234 235 236
            io.load_inference_model(dirname=self._model_dir,
                                    executor=self._executor,
                                    model_filename=self._model_filename,
                                    params_filename=self._params_filename)
237 238 239 240 241
        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)
        self._data_loader.set_sample_generator(
242
            self._sample_generator,
243 244 245 246
            batch_size=self._batch_size,
            drop_last=True,
            places=self._place)

247
        # collect the variable names for sampling
248 249 250 251 252
        persistable_var_names = []
        for var in self._program.list_vars():
            if var.persistable:
                persistable_var_names.append(var.name)

253
        for op in self._program.global_block().ops:
254 255 256
            op_type = op.type
            if op_type in self._quantizable_op_type:
                if op_type in ("conv2d", "depthwise_conv2d"):
257 258 259
                    self._quantized_act_var_name.add(op.input("Input")[0])
                    self._quantized_weight_var_name.add(op.input("Filter")[0])
                    self._quantized_act_var_name.add(op.output("Output")[0])
260
                elif op_type == "mul":
261 262
                    if self._is_input_all_not_persistable(
                            op, persistable_var_names):
263
                        op._set_attr("skip_quant", True)
264
                        _logger.warning("Skip quant a mul op for two "
265 266
                                        "input variables are not persistable")
                    else:
267 268 269
                        self._quantized_act_var_name.add(op.input("X")[0])
                        self._quantized_weight_var_name.add(op.input("Y")[0])
                        self._quantized_act_var_name.add(op.output("Out")[0])
270 271 272 273 274 275 276 277
                else:
                    # process other quantizable op type, the input must all not persistable
                    if self._is_input_all_not_persistable(
                            op, persistable_var_names):
                        input_output_name_list = self._op_real_in_out_name[
                            op_type]
                        for input_name in input_output_name_list[0]:
                            for var_name in op.input(input_name):
278
                                self._quantized_act_var_name.add(var_name)
279 280
                        for output_name in input_output_name_list[1]:
                            for var_name in op.output(output_name):
281
                                self._quantized_act_var_name.add(var_name)
282 283 284

        # set activation variables to be persistable, so can obtain 
        # the tensor data in sample_data
285 286 287 288
        for var in self._program.list_vars():
            if var.name in self._quantized_act_var_name:
                var.persistable = True

289
    def _sample_data(self, iter):
290 291 292 293 294 295 296 297 298
        '''
        Sample the tensor data of quantized variables, 
        applied in every iteration.
        '''
        for var_name in self._quantized_weight_var_name:
            if var_name not in self._sampling_data:
                var_tensor = self._load_var_value(var_name)
                self._sampling_data[var_name] = var_tensor

299 300 301 302 303 304 305 306 307 308 309 310 311 312
        if self._is_use_cache_file:
            for var_name in self._quantized_act_var_name:
                var_tensor = self._load_var_value(var_name)
                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] = []
                var_tensor = self._load_var_value(var_name)
                var_tensor = var_tensor.ravel()
                self._sampling_data[var_name].append(var_tensor)
313 314 315 316 317

    def _calculate_scale_factor(self):
        '''
        Calculate the scale factor of quantized variables.
        '''
318
        # apply channel_wise_abs_max quantization for weights
319 320 321 322 323 324 325 326 327
        for var_name in self._quantized_weight_var_name:
            data = self._sampling_data[var_name]
            scale_factor_per_channel = []
            for i in range(data.shape[0]):
                abs_max_value = np.max(np.abs(data[i]))
                scale_factor_per_channel.append(abs_max_value)
            self._quantized_var_scale_factor[
                var_name] = scale_factor_per_channel

328
        # apply kl quantization for activation
329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
        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)

                if self._algo == "KL":
                    self._quantized_var_scale_factor[var_name] = \
                        self._get_kl_scaling_factor(np.abs(sampling_data))
                else:
                    self._quantized_var_scale_factor[var_name] = \
                        np.max(np.abs(sampling_data))
        else:
            for var_name in self._quantized_act_var_name:
                self._sampling_data[var_name] = np.concatenate(
                    self._sampling_data[var_name])
                if self._algo == "KL":
                    self._quantized_var_scale_factor[var_name] = \
                        self._get_kl_scaling_factor(np.abs(self._sampling_data[var_name]))
                else:
                    self._quantized_var_scale_factor[var_name] = \
                        np.max(np.abs(self._sampling_data[var_name]))
356 357 358 359 360

    def _update_program(self):
        '''
        Insert fake_quantize/fake_dequantize op to the program.
        '''
361
        # reset quantized activation variable
362 363 364 365 366 367 368
        for var in self._program.list_vars():
            if var.name in self._quantized_act_var_name:
                var.persistable = False

        # use QuantizationTransformPass to insert fake_quantize/fake_dequantize op
        graph = IrGraph(core.Graph(self._program.desc), for_test=True)

369 370
        major_quantizable_op_types = []
        for op_type in QuantizationTransformPass._supported_quantizable_op_type:
371
            if op_type in self._quantizable_op_type:
372
                major_quantizable_op_types.append(op_type)
373 374 375 376 377 378 379
        transform_pass = QuantizationTransformPass(
            scope=self._scope,
            place=self._place,
            weight_bits=self._bit_length,
            activation_bits=self._bit_length,
            activation_quantize_type='moving_average_abs_max',
            weight_quantize_type='channel_wise_abs_max',
380
            quantizable_op_type=major_quantizable_op_types)
381 382 383
        transform_pass.apply(graph)

        # use AddQuantDequantPass to insert fake_quant_dequant op
384 385
        minor_quantizable_op_types = []
        for op_type in AddQuantDequantPass._supported_quantizable_op_type:
386
            if op_type in self._quantizable_op_type:
387
                minor_quantizable_op_types.append(op_type)
388 389 390
        add_quant_dequant_pass = AddQuantDequantPass(
            scope=self._scope,
            place=self._place,
391
            quantizable_op_type=minor_quantizable_op_types)
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
        add_quant_dequant_pass.apply(graph)

        # save scale factor to scale var node
        for key, val in self._quantized_var_scale_factor.items():
            self._set_var_node_value(
                key + ".scale", np.array(
                    [val], dtype=np.float32))
            self._set_var_node_value(
                key + ".quant_dequant.scale", np.array(
                    [val], dtype=np.float32))

        # apply QuantizationFreezePass, and obtain the final quant model
        freeze_pass = QuantizationFreezePass(
            scope=self._scope,
            place=self._place,
            weight_bits=self._bit_length,
            activation_bits=self._bit_length,
            weight_quantize_type='channel_wise_abs_max',
410
            quantizable_op_type=major_quantizable_op_types)
411 412 413
        freeze_pass.apply(graph)
        self._program = graph.to_program()

414 415 416 417 418 419 420 421 422
    def _save_output_scale(self):
        '''
        Save output scale to the quantized op.
        '''
        output_scale_name = "output_scale"
        for op in self._program.global_block().ops:
            if op.type in self._quantizable_op_type:
                output_name_list = self._op_real_in_out_name[op.type][1]
                for output_name in output_name_list:
423 424 425 426 427
                    for output_var_name in op.output(output_name):
                        if output_var_name in self._quantized_var_scale_factor:
                            op._set_attr(output_scale_name,
                                         self._quantized_var_scale_factor[
                                             output_var_name])
428

429 430 431 432 433 434 435 436
    def _load_var_value(self, var_name):
        '''
        Load variable value from scope
        '''
        return np.array(self._scope.find_var(var_name).get_tensor())

    def _set_var_node_value(self, var_node_name, np_value):
        '''
437
        Set the value of var node by name, if the node exits,
438 439 440 441 442 443 444 445
        '''
        assert isinstance(np_value, np.ndarray), \
            'The type of value should be numpy array.'
        var_node = self._scope.find_var(var_node_name)
        if var_node != None:
            tensor = var_node.get_tensor()
            tensor.set(np_value, self._place)

446 447 448 449 450 451 452 453 454 455 456 457 458
    def _is_input_all_not_persistable(self, op, persistable_var_names):
        '''
        Analyze the real inputs of the op are all not persistable.
        '''
        is_input_all_not_persistable = True
        input_name_list = self._op_real_in_out_name[op.type][0]
        for input_name in input_name_list:
            for var_name in op.input(input_name):
                if var_name in persistable_var_names:
                    is_input_all_not_persistable = False
                    break
        return is_input_all_not_persistable

459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559
    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:
560 561
                    _logger.error("Fatal error!, idx = " + str(idx) +
                                  " qindex = 0! p_idx = " + str(p_idx))
562 563 564
                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