pytorch/torch/quantization/quantize.py

from __future__ import absolute_import, division, print_function, unicode_literals

import copy
import torch
import torch.nn as nn
import torch.nn._intrinsic as nni
import torch.nn._intrinsic.quantized as nniq
import torch.nn._intrinsic.qat as nniqat
import torch.nn.quantized as nnq
import torch.nn.quantized.dynamic as nnqd
from .QConfig import default_dynamic_qconfig, float16_dynamic_qconfig
import torch.nn.qat as nnqat


DEFAULT_SKIP_LIST = [nn.Dropout, nn.Identity, nn.MaxPool2d, nn.AvgPool2d, nn.AdaptiveAvgPool2d]

def propagate_qconfig_helper(module, qconfig_dict, skip_list=DEFAULT_SKIP_LIST, qconfig_parent=None, prefix=''):
    r"""This is a helper function for `propagate_qconfig`

    Args:
        module: input module
        qconfig_dict: dictionary that maps from name of submodule to quantization
                     configuration
        qconfig_parent: quantization config of parent module, we will fallback to
                       this config when there is no specified config for current
                       module
        prefix: corresponding prefix of the current module, used as key in
                qconfig_dict

    Return:
        None, module is modified inplace with qconfig attached
    """
    if type(module) in skip_list:
        module.qconfig = None
    if not hasattr(module, 'qconfig'):
        module.qconfig = qconfig_parent
        if qconfig_dict:
            if prefix in qconfig_dict:
                module.qconfig = qconfig_dict[prefix]
            elif type(module) in qconfig_dict:
                module.qconfig = qconfig_dict[type(module)]

    # Don't quantize empty Sequential, empty Sequential is same as
    # Identity, but we can't put Sequential into skip list because
    # we also have non-empty Sequential and the qconfig needs to
    # be propagated to child in that case
    # TODO: Add test
    if len(module._modules) == 0 and type(module) == nn.Sequential:
        module.qconfig = None

    for name, child in module.named_children():
        module_prefix = prefix + '.' + name if prefix else name
        propagate_qconfig_helper(child, qconfig_dict, skip_list, module.qconfig, module_prefix)

# TODO(jerryzh): expose skip_list
def propagate_qconfig(module, qconfig_dict=None):
    r"""Propagate qconfig through the module hierarchy and assign `qconfig`
    attribute on each leaf module

    Args:
        module: input module
        qconfig_dict: dictionary that maps from name of submodule to quantization
            configuration, qconfig applies to all submodules of a given
            module unless qconfig for the submodules are specified(when the
            submodule already has qconfig attribute)

    Return:
        None, module is modified inplace with qconfig attached
    """
    if qconfig_dict is None:
        qconfig_dict = {}
    propagate_qconfig_helper(module, qconfig_dict)

def _observer_forward_hook(self, input, output):
    r"""Forward hook that calls observer on the output
    """
    return self.observer(output)

def add_observer(module):
    r"""Add observer for the leaf child of the module.

    This function insert observer module to all leaf child module that
    has a valid qconfig attribute.

    Args:
        module: input module with qconfig attributes for all the leaf modules
        that we want to quantize

    Return:
        None, module is modified inplace with added observer modules and
            forward_hooks
    """
    for child in module.children():
        if type(child) == nnq.FloatFunctional:
            if hasattr(child, 'qconfig') and child.qconfig is not None:
                child.observer = child.qconfig.activation()
        else:
            add_observer(child)

    # Insert observers only for leaf nodes, note that this observer is for
    # the output of the module, for input QuantStub will observe them
    if hasattr(module, 'qconfig') and module.qconfig is not None and \
       len(module._modules) == 0:
        # observer and hook will be gone after we swap the module
        module.add_module('observer', module.qconfig.activation())
        module.register_forward_hook(_observer_forward_hook)

class QuantWrapper(nn.Module):
    r"""A wrapper class that wraps the input module, adds QuantStub and
    DeQuantStub and surround the call to module with call to quant and dequant
    modules.

    This is used by the `quantization` utility functions to add the quant and
    dequant modules, before `convert` function `QuantStub` will just be observer,
    it observes the input tensor, after `convert`, `QuantStub`
    will be swapped to `nnq.Quantize` which does actual quantization. Similarly
    for `DeQuantStub`.
    """
    def __init__(self, module):
        super(QuantWrapper, self).__init__()
        qconfig = module.qconfig if hasattr(module, 'qconfig') else None
        self.add_module('quant', QuantStub(qconfig))
        self.add_module('dequant', DeQuantStub(qconfig))
        self.add_module('module', module)
        self.train(module.training)

    def forward(self, X):
        X = self.quant(X)
        X = self.module(X)
        return self.dequant(X)

def add_quant_dequant(module):
    r"""Wrap the leaf child module in QuantWrapper if it has a valid qconfig
    Note that this function will modify the children of module inplace and it
    can return a new module which wraps the input module as well.

    Args:
        module: input module with qconfig attributes for all the leaf modules
        that we want to quantize

    Return:
        Either the inplace modified module with submodules wrapped in
        `QuantWrapper` based on qconfig or a new `QuantWrapper` module which
        wraps the input module, the latter case only happens when the input
        module is a leaf module and we want to quantize it.
    """
    if len(module._modules) == 0 and hasattr(module, 'qconfig') and module.qconfig:
        return QuantWrapper(module)

    for name, child in module.named_children():
        module._modules[name] = add_quant_dequant(child)
    return module

def prepare(model):
    r"""Prepares the model for calibration or training.

    The model will be attached with observer and quant dequant or fake quant
    modules, and qconfig will be propagated.

    Note that the model will be modified inplace but in case the input model
    is a leaf model, a wrapped model will be returned.

    Args:
        mod: input model
    """
    propagate_qconfig(model)
    add_observer(model)

class QuantStub(nn.Module):
    r"""Quantize stub module, before calibration, this is same as an observer,
    it will be swapped as `nnq.Quantize` in `convert`.

    Args:
        qconfig: quantization configuration for the tensor,
            if qconfig is not provided, we will get qconfig from parent modules
    """
    def __init__(self, qconfig=None):
        super(QuantStub, self).__init__()
        if qconfig:
            self.qconfig = qconfig

    def forward(self, x):
        return x

class DeQuantStub(nn.Module):
    r"""Dequantize stub module, before calibration, this is same as identity,
    this will be swapped as `nnq.DeQuantize` in `convert`.
    """
    def __init__(self, qconfig=None):
        super(DeQuantStub, self).__init__()
        if qconfig:
            self.qconfig = qconfig

    def forward(self, x):
        return x

# Map for swapping float module to quantized ones
DEFAULT_MODULE_MAPPING = {
    nn.Linear: nnq.Linear,
    nn.ReLU: nnq.ReLU,
    nn.Conv2d: nnq.Conv2d,
    QuantStub: nnq.Quantize,
    DeQuantStub: nnq.DeQuantize,
    # Wrapper Modules:
    nnq.FloatFunctional: nnq.QFunctional,
    # Intrinsic modules:
    nni.ConvReLU2d: nniq.ConvReLU2d,
    nni.LinearReLU: nniq.LinearReLU,
    nniqat.ConvReLU2d: nniq.ConvReLU2d,
    nniqat.LinearReLU: nniq.LinearReLU,
    nniqat.ConvBn2d: nnq.Conv2d,
    nniqat.ConvBnReLU2d: nniq.ConvReLU2d,
    # QAT modules:
    nnqat.Linear: nnq.Linear,
    nnqat.Conv2d: nnq.Conv2d,
}

# Map for swapping float module to qat modules
DEFAULT_QAT_MODULE_MAPPING = {
    nn.Linear: nnqat.Linear,
    nn.Conv2d: nnqat.Conv2d,
    # Intrinsic modules:
    nni.ConvBn2d: nniqat.ConvBn2d,
    nni.ConvBnReLU2d: nniqat.ConvBnReLU2d,
    nni.ConvReLU2d: nniqat.ConvReLU2d,
    nni.LinearReLU: nniqat.LinearReLU
}

DEFAULT_DYNAMIC_MODULE_MAPPING = {
    nn.Linear: nnqd.Linear,
    nn.LSTM: nnqd.LSTM,
}

def quantize(model, run_fn, run_args, mapping=DEFAULT_MODULE_MAPPING):
    r"""Converts a float model to quantized model.

    First it will prepare the model for calibration or training, then it calls
    `run_fn` which will run the calibration step or training step,
    after that we will call `convert` which will convert the model to a
    quantized model.

    Args:
        model: input model
        run_fn: a function for evaluating the prepared model, can be a
            function that simply runs the prepared model or a training loop
        run_args: positional arguments for `run_fn`

    Return:
        Quantized model.
    """

    model = copy.deepcopy(model)
    model.eval()
    prepare(model)
    run_fn(model, run_args)
    convert(model, mapping)
    return model

def quantize_dynamic(model, qconfig_dict=None, dtype=torch.qint8, mapping=DEFAULT_DYNAMIC_MODULE_MAPPING):
    r"""Converts a float model to dynamic (i.e. weights-only) quantized model.

    Replaces specified modules with dynamic weight-only quantized versions and output the quantized model.

    For simplest usage provide `dtype` argument that can be float16 or qint8. Weight-only quantization
    by default is performed for layers with large weights size - i.e. Linear and RNN variants.

    Fine grained control is possible with `qconfig_dict` and `mapping` that act similarly to `quantize()`.
    If `qconfig_dict` is provided, the `dtype` argument is ignored.

    Args:
        module: input model
        qconfig_dict: dictionary that maps from name or type of submodule to quantization
            configuration, qconfig applies to all submodules of a given
            module unless qconfig for the submodules are specified (when the
            submodule already has qconfig attribute). Entries in the dictionary
            need to be QConfigDynamic instances.
        mapping: maps type of a submodule to a type of corresponding dynamically quantized version
            with which the submodule needs to be replaced
    """
    if qconfig_dict is None:
        if dtype == torch.qint8:
            qconfig_dict = {
                nn.Linear : default_dynamic_qconfig,
                nn.LSTM : default_dynamic_qconfig,
            }
        elif dtype == torch.float16:
            qconfig_dict = {
                # TODO: uncomment when float16 Linear support is added
                # nn.Linear : default_dynamic_qconfig,
                nn.LSTM : float16_dynamic_qconfig,
            }
        else:
            raise ValueError(
                "Don't know how to quantize with default settings for {}. Provide full qconfig please".format(dtype))

    model = copy.deepcopy(model)
    model.eval()
    propagate_qconfig(model, qconfig_dict)
    convert(model, mapping)
    return model

def prepare_qat(model):
    prepare(model)
    convert(model, DEFAULT_QAT_MODULE_MAPPING)

def quantize_qat(model, run_fn, run_args):
    r"""Do quantization aware training and output a quantized model

    Args:
        model: input model
        run_fn: a function for evaluating the prepared model, can be a
            function that simply runs the prepared model or a training loop
        run_args: positional arguments for `run_fn`

    Return:
        Quantized model.
    """
    model = copy.deepcopy(model)
    model.train()
    prepare_qat(model)
    run_fn(model, run_args)
    convert(model)
    return model

def convert(module, mapping=DEFAULT_MODULE_MAPPING):
    r"""Converts the float module with observers(where we can get quantization
    parameters) to a quantized module.
    Args:
        module: calibrated module with observers
        mapping: a dictionary that maps from float module type to quantized
           module type, can be overwrritten to allow swapping user defined Modules
    """
    reassign = {}
    # TODO(jerryzh): remove after deciding on the impl of intrinsic modules
    SWAPPABLE_MODULES = (nni.ConvBn2d,
                         nni.ConvBnReLU2d,
                         nni.LinearReLU,
                         nni.ConvReLU2d)

    for name, mod in module.named_children():
        if type(mod) not in SWAPPABLE_MODULES:
            convert(mod, mapping)
        reassign[name] = swap_module(mod, mapping)

    for key, value in reassign.items():
        module._modules[key] = value

def swap_module(mod, mapping):
    r"""Swaps the module if it has a quantized counterpart and it has an
    `observer` attached.

    Args:
        mod: input module
        mapping: a dictionary that maps from nn module to nnq module

    Return:
        The corresponding quantized module of `mod`
    """
    new_mod = mod
    if hasattr(mod, 'qconfig') and mod.qconfig is not None:
        if type(mod) in mapping:
            new_mod = mapping[type(mod)].from_float(mod)
    return new_mod

def get_observer_dict(mod, target_dict, prefix=""):
    r"""Traverse the modules and save all observers into dict.
    This is mainly used for quantization accuracy debug
    Args:
        mod: the top module we want to save all observers
        prefix: the prefix for the current module
        target_dict: the dictionary used to save all the observers
    """
    def get_prefix(prefix):
        return prefix if prefix == "" else prefix + '.'

    if hasattr(mod, 'observer'):
        target_dict[get_prefix(prefix) + 'observer'] = mod.observer
    for name, child in mod.named_children():
        module_prefix = get_prefix(prefix) + name if prefix else name
        get_observer_dict(child, target_dict, module_prefix)