quanter.py

# 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.

import copy
import paddle
import paddle.fluid as fluid
from paddle.fluid.framework import IrGraph
from paddle.fluid.contrib.slim.quantization import QuantizationTransformPass
from paddle.fluid.contrib.slim.quantization import QuantizationFreezePass
from paddle.fluid.contrib.slim.quantization import ConvertToInt8Pass
from paddle.fluid.contrib.slim.quantization import TransformForMobilePass
from paddle.fluid.contrib.slim.quantization import PostTrainingQuantization
from paddle.fluid.contrib.slim.quantization import AddQuantDequantPass
from paddle.fluid import core

WEIGHT_QUANTIZATION_TYPES = [
    'abs_max', 'channel_wise_abs_max', 'range_abs_max',
    'moving_average_abs_max'
]
ACTIVATION_QUANTIZATION_TYPES = [
    'abs_max', 'range_abs_max', 'moving_average_abs_max'
]
VALID_DTYPES = ['int8']
TRANSFORM_PASS_OP_TYPES = ['conv2d', 'depthwise_conv2d', 'mul']
QUANT_DEQUANT_PASS_OP_TYPES = ['elementwise_add', 'pool2d']

_quant_config_default = {
    # weight quantize type, default is 'abs_max'
    'weight_quantize_type': 'abs_max',
    # activation quantize type, default is 'abs_max'
    'activation_quantize_type': 'abs_max',
    # weight quantize bit num, default is 8
    'weight_bits': 8,
    # activation quantize bit num, default is 8
    'activation_bits': 8,
    # ops of name_scope in not_quant_pattern list, will not be quantized
    'not_quant_pattern': ['skip_quant'],
    # ops of type in quantize_op_types, will be quantized
    'quantize_op_types':
    ['conv2d', 'depthwise_conv2d', 'mul', 'elementwise_add', 'pool2d'],
    # data type after quantization, such as 'uint8', 'int8', etc. default is 'int8'
    'dtype': 'int8',
    # window size for 'range_abs_max' quantization. defaulf is 10000
    'window_size': 10000,
    # The decay coefficient of moving average, default is 0.9
    'moving_rate': 0.9,
    # if set quant_weight_only True, then only quantize parameters of layers which need to be quantized,
    # and activations will not be quantized.
    'quant_weight_only': False
}


def _parse_configs(user_config):
    """
    check user configs is valid, and set default value if user not config.
    Args:
        user_config(dict):the config of user.
    Return:
        configs(dict): final configs will be used.
    """

    configs = copy.deepcopy(_quant_config_default)
    configs.update(user_config)

    # check configs is valid
    assert configs['weight_quantize_type'] in WEIGHT_QUANTIZATION_TYPES, \
        "Unknown weight_quantize_type: '%s'. It can only be " + " ".join(WEIGHT_QUANTIZATION_TYPES)

    assert configs['activation_quantize_type'] in ACTIVATION_QUANTIZATION_TYPES, \
        "Unknown activation_quantize_type: '%s'. It can only be " + " ".join(ACTIVATION_QUANTIZATION_TYPES)

    assert isinstance(configs['weight_bits'], int), \
        "weight_bits must be int value."

    assert (configs['weight_bits'] >= 1 and configs['weight_bits'] <= 16), \
        "weight_bits should be between 1 and 16."

    assert isinstance(configs['activation_bits'], int), \
        "activation_bits must be int value."

    assert (configs['activation_bits'] >= 1 and configs['activation_bits'] <= 16), \
        "activation_bits should be between 1 and 16."

    assert isinstance(configs['not_quant_pattern'], list), \
        "not_quant_pattern must be a list"

    assert isinstance(configs['quantize_op_types'], list), \
        "quantize_op_types must be a list"

    for op_type in configs['quantize_op_types']:
        assert (op_type in QUANT_DEQUANT_PASS_OP_TYPES) or (
            op_type in TRANSFORM_PASS_OP_TYPES), "{} is not support, \
                    now support op types are {}".format(
                op_type, TRANSFORM_PASS_OP_TYPES + QUANT_DEQUANT_PASS_OP_TYPES)

    assert isinstance(configs['dtype'], str), \
        "dtype must be a str."

    assert (configs['dtype'] in VALID_DTYPES), \
        "dtype can only be " + " ".join(VALID_DTYPES)

    assert isinstance(configs['window_size'], int), \
        "window_size must be int value, window size for 'range_abs_max' quantization, default is 10000."

    assert isinstance(configs['moving_rate'], float), \
        "moving_rate must be float value, The decay coefficient of moving average, default is 0.9."

    assert isinstance(configs['quant_weight_only'], bool), \
        "quant_weight_only must be bool value, if set quant_weight_only True, " \
        "then only quantize parameters of layers which need to be quantized, " \
        " and activations will not be quantized."

    return configs


def quant_aware(program, place, config, scope=None, for_test=False):
    """
    add trainable quantization ops in program.
    Args:
        program(fluid.Program): program
        scope(fluid.Scope): the scope to store var, it's should be the value of program's scope, usually it's fluid.global_scope().
        place(fluid.CPUPlace or fluid.CUDAPlace): place
        config(dict): configs for quantization, default values are in quant_config_default dict.
        for_test: if program is test program, for_test should be set True, else False.
    Return:
        fluid.Program: user can finetune this quantization program to enhance the accuracy.
    """

    scope = fluid.global_scope() if not scope else scope
    assert isinstance(config, dict), "config must be dict"

    assert 'weight_quantize_type' in config.keys(
    ), 'weight_quantize_type must be configured'
    assert 'activation_quantize_type' in config.keys(
    ), 'activation_quantize_type must be configured'

    config = _parse_configs(config)
    main_graph = IrGraph(core.Graph(program.desc), for_test=for_test)

    transform_pass_ops = []
    quant_dequant_ops = []
    for op_type in config['quantize_op_types']:
        if op_type in TRANSFORM_PASS_OP_TYPES:
            transform_pass_ops.append(op_type)
        elif op_type in QUANT_DEQUANT_PASS_OP_TYPES:
            quant_dequant_ops.append(op_type)
    if len(transform_pass_ops) > 0:
        transform_pass = QuantizationTransformPass(
            scope=scope,
            place=place,
            weight_bits=config['weight_bits'],
            activation_bits=config['activation_bits'],
            activation_quantize_type=config['activation_quantize_type'],
            weight_quantize_type=config['weight_quantize_type'],
            window_size=config['window_size'],
            moving_rate=config['moving_rate'],
            quantizable_op_type=transform_pass_ops,
            skip_pattern=config['not_quant_pattern'])

        transform_pass.apply(main_graph)

    if len(quant_dequant_ops) > 0:
        quant_dequant_pass = AddQuantDequantPass(
            scope=scope,
            place=place,
            moving_rate=config['moving_rate'],
            quant_bits=config['activation_bits'],
            skip_pattern=config['not_quant_pattern'],
            quantizable_op_type=quant_dequant_ops)
        quant_dequant_pass.apply(main_graph)

    if for_test:
        quant_program = main_graph.to_program()
    else:
        quant_program = fluid.CompiledProgram(main_graph.graph)
    return quant_program


def quant_post(executor,
               model_dir,
               quantize_model_path,
               sample_generator,
               model_filename=None,
               params_filename=None,
               batch_size=16,
               batch_nums=None,
               scope=None,
               algo='KL',
               quantizable_op_type=["conv2d", "depthwise_conv2d", "mul"]):
    """
    The function utilizes post training quantization method 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.
        model_dir(str): The path of fp32 model that will be quantized, and 
            the model and params that saved by fluid.io.save_inference_model 
            are under the path.
        quantize_model_path(str): The path to save quantized model using api
            fluid.io.save_inference_model.
        sample_generator(Python Generator): The sample generator provides 
            calibrate data for DataLoader, and it only returns a sample every time.
        model_filename(str, optional): The name of model file to load the inference 
                program. If parameters were saved in separate files, 
                set it as 'None'. Default is 'None'.
        params_filename(str, optional): The name of params file to load all parameters.
                When all parameters were saved in a single file, set it 
                as 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 16.
        batch_nums(int, optional): If set batch_nums, the number of calibrate 
                        data is 'batch_size*batch_nums'. If batch_nums is None, use all data
                        generated by sample_generator  as calibrate data.
        scope(fluid.Scope, optional): The scope to run program, use it to load 
                        and save variables. If scope is None, will use fluid.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 method to get the scale factor. Default is 'KL'.
        quantizable_op_type(list[str], optional): The list of op types
                        that will be quantized. Default is ["conv2d", "depthwise_conv2d", 
                        "mul"].
    Returns:
        None
    """
    post_training_quantization = PostTrainingQuantization(
        executor=executor,
        sample_generator=sample_generator,
        model_dir=model_dir,
        model_filename=model_filename,
        params_filename=params_filename,
        batch_size=batch_size,
        batch_nums=batch_nums,
        scope=scope,
        algo=algo,
        quantizable_op_type=quantizable_op_type,
        is_full_quantize=False)
    post_training_quantization.quantize()
    post_training_quantization.save_quantized_model(quantize_model_path)


def convert(program, place, config, scope=None, save_int8=False):
    """
    add quantization ops in program. the program returned is not trainable.
    Args:
        program(fluid.Program): program
        scope(fluid.Scope): the scope to store var, when is None will use fluid.global_scope()
        place(fluid.CPUPlace or fluid.CUDAPlace): place
        config(dict): configs for quantization, default values are in quant_config_default dict.
        save_int8: is export int8 freezed program.
    Return:
        fluid.Program: freezed program which can be used for inference.
                       parameters is float32 type, but it's value in int8 range.
        fluid.Program: freezed int8 program which can be used for inference.
                       if save_int8 is False, this value is None.
    """
    scope = fluid.global_scope() if not scope else scope
    test_graph = IrGraph(core.Graph(program.desc), for_test=True)

    # Freeze the graph after training by adjusting the quantize
    # operators' order for the inference.
    freeze_pass = QuantizationFreezePass(
        scope=scope,
        place=place,
        weight_quantize_type=config['weight_quantize_type'])
    freeze_pass.apply(test_graph)
    freezed_program = test_graph.to_program()

    if save_int8:
        convert_int8_pass = ConvertToInt8Pass(
            scope=fluid.global_scope(), place=place)
        convert_int8_pass.apply(test_graph)
        freezed_program_int8 = test_graph.to_program()
        return freezed_program, freezed_program_int8
    else:
        return freezed_program