p47068915
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!7834 rename create_quant_config && add notes 

Merge pull request !7834 from yuchaojie/quant
This commit is contained in:
mindspore-ci-bot 2020-10-28 16:23:44 +08:00 committed by Gitee
parent 9fc0218c56 69d9e76d65
commit 40cfca781a
4 changed files with 65 additions and 43 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
mindspore/compression/quant/qat.py
View File

@ -33,7 +33,7 @@ from ..common import QuantDtype
from .quantizer import Quantizer, OptimizeOption from .quantizer import Quantizer, OptimizeOption
__all__ = ["QuantizationAwareTraining"] __all__ = ["QuantizationAwareTraining", "create_quant_config"]
_ACTIVATION_MAP = {nn.ReLU: quant.ActQuant, _ACTIVATION_MAP = {nn.ReLU: quant.ActQuant,
@ -44,13 +44,12 @@ _ACTIVATION_MAP = {nn.ReLU: quant.ActQuant,
nn.HSwish: quant.HSwishQuant} nn.HSwish: quant.HSwishQuant}
def get_quant_config(quant_observer=(quant.FakeQuantWithMinMaxObserver, quant.FakeQuantWithMinMaxObserver), def create_quant_config(quant_observer=(quant.FakeQuantWithMinMaxObserver, quant.FakeQuantWithMinMaxObserver),
quant_delay=(0, 0), quant_delay=(0, 0),
quant_dtype=(QuantDtype.INT8, QuantDtype.INT8), quant_dtype=(QuantDtype.INT8, QuantDtype.INT8),
per_channel=(False, False), per_channel=(False, False),
symmetric=(False, False), symmetric=(False, False),
narrow_range=(False, False) narrow_range=(False, False)):
):
r""" r"""
Configs the oberser type of weights and data flow with quant params. Configs the oberser type of weights and data flow with quant params.
@ -78,7 +77,7 @@ def get_quant_config(quant_observer=(quant.FakeQuantWithMinMaxObserver, quant.Fa
weight_observer = quant_observer[0].partial_init(quant_delay=quant_delay[0], quant_dtype=quant_dtype[0], weight_observer = quant_observer[0].partial_init(quant_delay=quant_delay[0], quant_dtype=quant_dtype[0],
per_channel=per_channel[0], symmetric=symmetric[0], per_channel=per_channel[0], symmetric=symmetric[0],
narrow_range=narrow_range[0]) narrow_range=narrow_range[0])
act_observer = quant_observer[0].partial_init(quant_delay=quant_delay[-1], quant_dtype=quant_dtype[-1], act_observer = quant_observer[-1].partial_init(quant_delay=quant_delay[-1], quant_dtype=quant_dtype[-1],
per_channel=per_channel[-1], symmetric=symmetric[-1], per_channel=per_channel[-1], symmetric=symmetric[-1],
narrow_range=narrow_range[-1]) narrow_range=narrow_range[-1])
return quant.QuantConfig(weight=weight_observer, activation=act_observer) return quant.QuantConfig(weight=weight_observer, activation=act_observer)
@ -221,7 +220,7 @@ class QuantizationAwareTraining(Quantizer):
self.act_range = Validator.check_bool(narrow_range[-1], "narrow range") self.act_range = Validator.check_bool(narrow_range[-1], "narrow range")
self._convert_method_map = {quant.Conv2dBnAct: self._convert_conv, self._convert_method_map = {quant.Conv2dBnAct: self._convert_conv,
quant.DenseBnAct: self._convert_dense} quant.DenseBnAct: self._convert_dense}
self.quant_config = get_quant_config(quant_delay=quant_delay, self.quant_config = create_quant_config(quant_delay=quant_delay,
quant_dtype=quant_dtype, quant_dtype=quant_dtype,
per_channel=per_channel, per_channel=per_channel,
symmetric=symmetric, symmetric=symmetric,

69
mindspore/nn/layer/quant.py
View File

@ -322,7 +322,7 @@ def _partial_init(cls_or_self, **kwargs):
return r return r
class Observer(Cell): class _Observer(Cell):
""" """
Base class of Observer. Observer is used to calculate the statistics of specific layer. Base class of Observer. Observer is used to calculate the statistics of specific layer.
@ -334,7 +334,7 @@ class Observer(Cell):
""" """
def __init__(self, quant_dtype): def __init__(self, quant_dtype):
super(Observer, self).__init__() super(_Observer, self).__init__()
self.quant_dtype = quant_dtype self.quant_dtype = quant_dtype
def extend_repr(self): def extend_repr(self):
@ -347,7 +347,7 @@ class Observer(Cell):
partial_init = classmethod(_partial_init) partial_init = classmethod(_partial_init)
class UniformQuantObserver(Observer): class UniformQuantObserver(_Observer):
""" """
The base class of Uniform Quantization Observer. The base class of Uniform Quantization Observer.
@ -542,7 +542,8 @@ class Conv2dBnFoldQuant(Cell):
var_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the var_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the
variance vector. Default: 'ones'. variance vector. Default: 'ones'.
fake (bool): Whether Conv2dBnFoldQuant Cell adds FakeQuantWithMinMaxObserver. Default: True. fake (bool): Whether Conv2dBnFoldQuant Cell adds FakeQuantWithMinMaxObserver. Default: True.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
freeze_bn (int): The quantization freeze BatchNormal op is according to the global step. Default: 100000. freeze_bn (int): The quantization freeze BatchNormal op is according to the global step. Default: 100000.
@ -553,7 +554,9 @@ class Conv2dBnFoldQuant(Cell):
Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`. Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`.
Examples: Examples:
>>> conv2d_bnfold = nn.Conv2dBnFoldQuant(1, 6, kernel_size=(2, 2), stride=(1, 1), pad_mode="valid") >>> qconfig = compression.quant.create_quant_config()
>>> conv2d_bnfold = nn.Conv2dBnFoldQuant(1, 6, kernel_size=(2, 2), stride=(1, 1), pad_mode="valid",
>>> quant_config=qconfig)
>>> input = Tensor(np.random.randint(-2, 2, (2, 1, 3, 3)), mindspore.float32) >>> input = Tensor(np.random.randint(-2, 2, (2, 1, 3, 3)), mindspore.float32)
>>> result = conv2d_bnfold(input) >>> result = conv2d_bnfold(input)
>>> result.shape >>> result.shape
@ -725,7 +728,8 @@ class Conv2dBnWithoutFoldQuant(Cell):
weight_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the convolution kernel. weight_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the convolution kernel.
Default: 'normal'. Default: 'normal'.
bias_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the bias vector. Default: 'zeros'. bias_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the bias vector. Default: 'zeros'.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -735,7 +739,9 @@ class Conv2dBnWithoutFoldQuant(Cell):
Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`. Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`.
Examples: Examples:
>>> conv2d_no_bnfold = nn.Conv2dBnWithoutFoldQuant(1, 6, kernel_size=(2, 2), stride=(1, 1), pad_mode="valid") >>> qconfig = compression.quant.create_quant_config()
>>> conv2d_no_bnfold = nn.Conv2dBnWithoutFoldQuant(1, 6, kernel_size=(2, 2), stride=(1, 1), pad_mode="valid",
>>> quant_config=qconfig)
>>> input = Tensor(np.random.randint(-2, 2, (2, 1, 3, 3)), mstype.float32) >>> input = Tensor(np.random.randint(-2, 2, (2, 1, 3, 3)), mstype.float32)
>>> result = conv2d_no_bnfold(input) >>> result = conv2d_no_bnfold(input)
>>> result.shape >>> result.shape
@ -846,7 +852,8 @@ class Conv2dQuant(Cell):
weight_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the convolution kernel. weight_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the convolution kernel.
Default: 'normal'. Default: 'normal'.
bias_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the bias vector. Default: 'zeros'. bias_init (Union[Tensor, str, Initializer, numbers.Number]): Initializer for the bias vector. Default: 'zeros'.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -856,7 +863,9 @@ class Conv2dQuant(Cell):
Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`. Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`.
Examples: Examples:
>>> conv2d_quant = nn.Conv2dQuant(1, 6, kernel_size= (2, 2), stride=(1, 1), pad_mode="valid") >>> qconfig = compression.quant.create_quant_config()
>>> conv2d_quant = nn.Conv2dQuant(1, 6, kernel_size= (2, 2), stride=(1, 1), pad_mode="valid",
>>> quant_config=qconfig)
>>> input = Tensor(np.random.randint(-2, 2, (2, 1, 3, 3)), mindspore.float32) >>> input = Tensor(np.random.randint(-2, 2, (2, 1, 3, 3)), mindspore.float32)
>>> result = conv2d_quant(input) >>> result = conv2d_quant(input)
>>> result.shape >>> result.shape
@ -947,7 +956,8 @@ class DenseQuant(Cell):
has_bias (bool): Specifies whether the layer uses a bias vector. Default: True. has_bias (bool): Specifies whether the layer uses a bias vector. Default: True.
activation (Union[str, Cell, Primitive]): The regularization function applied to the output of the layer, activation (Union[str, Cell, Primitive]): The regularization function applied to the output of the layer,
eg. 'relu'. Default: None. eg. 'relu'. Default: None.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -957,7 +967,8 @@ class DenseQuant(Cell):
Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`. Tensor of shape :math:`(N, C_{out}, H_{out}, W_{out})`.
Examples: Examples:
>>> dense_quant = nn.DenseQuant(3, 6) >>> qconfig = compression.quant.create_quant_config()
>>> dense_quant = nn.DenseQuant(3, 6, quant_config=qconfig)
>>> input = Tensor(np.random.randint(-2, 2, (2, 3)), mindspore.float32) >>> input = Tensor(np.random.randint(-2, 2, (2, 3)), mindspore.float32)
>>> result = dense_quant(input) >>> result = dense_quant(input)
>>> result.shape >>> result.shape
@ -1048,7 +1059,8 @@ class ActQuant(_QuantActivation):
Args: Args:
activation (Cell): Activation cell class. activation (Cell): Activation cell class.
ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999. ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -1058,7 +1070,8 @@ class ActQuant(_QuantActivation):
Tensor, with the same type and shape as the `input`. Tensor, with the same type and shape as the `input`.
Examples: Examples:
>>> act_quant = nn.ActQuant(nn.ReLU()) >>> qconfig = compression.quant.create_quant_config()
>>> act_quant = nn.ActQuant(nn.ReLU(), quant_config=qconfig)
>>> input = Tensor(np.array([[1, 2, -1], [-2, 0, -1]]), mindspore.float32) >>> input = Tensor(np.array([[1, 2, -1], [-2, 0, -1]]), mindspore.float32)
>>> result = act_quant(input) >>> result = act_quant(input)
>>> result >>> result
@ -1096,7 +1109,8 @@ class LeakyReLUQuant(_QuantActivation):
Args: Args:
activation (Cell): Activation cell class. activation (Cell): Activation cell class.
ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999. ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -1106,7 +1120,8 @@ class LeakyReLUQuant(_QuantActivation):
Tensor, with the same type and shape as the `input`. Tensor, with the same type and shape as the `input`.
Examples: Examples:
>>> activation = nn.LeakyReLUQuant(nn.LeakyReLU()) >>> qconfig = compression.quant.create_quant_config()
>>> activation = nn.LeakyReLUQuant(nn.LeakyReLU(), quant_config=qconfig)
>>> input = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32) >>> input = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32)
>>> result = activation(input) >>> result = activation(input)
>>> result >>> result
@ -1153,7 +1168,8 @@ class HSwishQuant(_QuantActivation):
Args: Args:
activation (Cell): Activation cell class. activation (Cell): Activation cell class.
ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999. ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -1163,7 +1179,8 @@ class HSwishQuant(_QuantActivation):
Tensor, with the same type and shape as the `input`. Tensor, with the same type and shape as the `input`.
Examples: Examples:
>>> activation = nn.HSwishQuant(nn.HSwish()) >>> qconfig = compression.quant.create_quant_config()
>>> activation = nn.HSwishQuant(nn.HSwish(), quant_config=qconfig)
>>> input = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32) >>> input = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32)
>>> result = activation(input) >>> result = activation(input)
>>> result >>> result
@ -1210,7 +1227,8 @@ class HSigmoidQuant(_QuantActivation):
Args: Args:
activation (Cell): Activation cell class. activation (Cell): Activation cell class.
ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999. ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -1220,7 +1238,8 @@ class HSigmoidQuant(_QuantActivation):
Tensor, with the same type and shape as the `x`. Tensor, with the same type and shape as the `x`.
Examples: Examples:
>>> activation = nn.HSigmoidQuant(nn.HSigmoid()) >>> qconfig = compression.quant.create_quant_config()
>>> activation = nn.HSigmoidQuant(nn.HSigmoid(), quant_config=qconfig)
>>> input = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32) >>> input = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32)
>>> result = activation(input) >>> result = activation(input)
>>> result >>> result
@ -1266,7 +1285,8 @@ class TensorAddQuant(Cell):
Args: Args:
ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999. ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -1277,7 +1297,8 @@ class TensorAddQuant(Cell):
Tensor, with the same type and shape as the `input_x1`. Tensor, with the same type and shape as the `input_x1`.
Examples: Examples:
>>> add_quant = nn.TensorAddQuant() >>> qconfig = compression.quant.create_quant_config()
>>> add_quant = nn.TensorAddQuant(quant_config=qconfig)
>>> input_x1 = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32) >>> input_x1 = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32)
>>> input_x2 = Tensor(np.ones((2, 3)), mindspore.float32) >>> input_x2 = Tensor(np.ones((2, 3)), mindspore.float32)
>>> result = add_quant(input_x1, input_x2) >>> result = add_quant(input_x1, input_x2)
@ -1311,7 +1332,8 @@ class MulQuant(Cell):
Args: Args:
ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999. ema_decay (float): Exponential Moving Average algorithm parameter. Default: 0.999.
quant_config (QuantConfig): Configs the oberser type of weight and activation. Default: quant_config_default. quant_config (QuantConfig): Configs the oberser types and quant configs of weight and activation. Default:
both set to default FakeQuantWithMinMaxObserver.
quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8. quant_dtype (QuantDtype): Specifies the FakeQuant datatype. Default: QuantDtype.INT8.
Inputs: Inputs:
@ -1322,7 +1344,8 @@ class MulQuant(Cell):
Tensor, with the same type and shape as the `input_x1`. Tensor, with the same type and shape as the `input_x1`.
Examples: Examples:
>>> mul_quant = nn.MulQuant() >>> qconfig = compression.quant.create_quant_config()
>>> mul_quant = nn.MulQuant(quant_config=qconfig)
>>> input_x1 = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32) >>> input_x1 = Tensor(np.array([[1, 2, 1], [-2, 0, -1]]), mindspore.float32)
>>> input_x2 = Tensor(np.ones((2, 3)) * 2, mindspore.float32) >>> input_x2 = Tensor(np.ones((2, 3)) * 2, mindspore.float32)
>>> result = mul_quant(input_x1, input_x2) >>> result = mul_quant(input_x1, input_x2)

4
model_zoo/official/cv/resnet50_quant/models/resnet_quant_manual.py
View File

@ -18,13 +18,13 @@ import mindspore.nn as nn
from mindspore.ops import operations as P from mindspore.ops import operations as P
from mindspore import Tensor from mindspore import Tensor
from mindspore.nn import FakeQuantWithMinMaxObserver, Conv2dBnFoldQuant from mindspore.nn import FakeQuantWithMinMaxObserver, Conv2dBnFoldQuant
from mindspore.compression.quant import qat from mindspore.compression.quant import create_quant_config
_ema_decay = 0.999 _ema_decay = 0.999
_symmetric = True _symmetric = True
_fake = True _fake = True
_per_channel = True _per_channel = True
_quant_config = qat.get_quant_config(per_channel=(_per_channel, False), symmetric=(_symmetric, False)) _quant_config = create_quant_config(per_channel=(_per_channel, False), symmetric=(_symmetric, False))
def _weight_variable(shape, factor=0.01): def _weight_variable(shape, factor=0.01):

4
tests/st/quantization/resnet50_quant/resnet_quant_manual.py
View File

@ -19,13 +19,13 @@ import mindspore.common.initializer as weight_init
from mindspore.ops import operations as P from mindspore.ops import operations as P
from mindspore import Tensor from mindspore import Tensor
from mindspore.nn import FakeQuantWithMinMaxObserver, Conv2dBnFoldQuant from mindspore.nn import FakeQuantWithMinMaxObserver, Conv2dBnFoldQuant
from mindspore.compression.quant import qat from mindspore.compression.quant import create_quant_config
_ema_decay = 0.999 _ema_decay = 0.999
_symmetric = True _symmetric = True
_fake = True _fake = True
_per_channel = True _per_channel = True
_quant_config = qat.get_quant_config(per_channel=(_per_channel, False), symmetric=(_symmetric, False)) _quant_config = create_quant_config(per_channel=(_per_channel, False), symmetric=(_symmetric, False))
def _weight_variable(shape, factor=0.01): def _weight_variable(shape, factor=0.01):