# 導(dǎo)入ReduceLROnPlateau
from keras.callbacks import ReduceLROnPlateau
# 定義ReduceLROnPlateau
reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=2, verbose=1)
# 使用ReduceLROnPlateau
model.fit(X_train, Y_train, callbacks=[reduce_lr])

import numpy as np
import matplotlib.pyplot as plt
import keras
from keras import backend as K
from keras.layers import Flatten,Conv2D,Dropout,Input,Dense,MaxPooling2D
from keras.models import Model
def exponent(global_epoch,
            learning_rate_base,
            decay_rate,
            min_learn_rate=0,
            ):
    learning_rate = learning_rate_base * pow(decay_rate, global_epoch)
    learning_rate = max(learning_rate,min_learn_rate)
    return learning_rate
class ExponentDecayScheduler(keras.callbacks.Callback):
    """
    繼承Callback，實現(xiàn)對學(xué)習(xí)率的調(diào)度
    """
    def __init__(self,
                 learning_rate_base,
                 decay_rate,
                 global_epoch_init=0,
                 min_learn_rate=0,
                 verbose=0):
        super(ExponentDecayScheduler, self).__init__()
        # 基礎(chǔ)的學(xué)習(xí)率
        self.learning_rate_base = learning_rate_base
        # 全局初始化epoch
        self.global_epoch = global_epoch_init
        self.decay_rate = decay_rate
        # 參數(shù)顯示  
        self.verbose = verbose
        # learning_rates用于記錄每次更新后的學(xué)習(xí)率，方便圖形化觀察
        self.min_learn_rate = min_learn_rate
        self.learning_rates = []
    def on_epoch_end(self, epochs ,logs=None):
        self.global_epoch = self.global_epoch + 1
        lr = K.get_value(self.model.optimizer.lr)
        self.learning_rates.append(lr)
	#更新學(xué)習(xí)率
    def on_epoch_begin(self, batch, logs=None):
        lr = exponent(global_epoch=self.global_epoch,
                    learning_rate_base=self.learning_rate_base,
                    decay_rate = self.decay_rate,
                    min_learn_rate = self.min_learn_rate)
        K.set_value(self.model.optimizer.lr, lr)
        if self.verbose > 0:
            print('\nBatch %05d: setting learning '
                  'rate to %s.' % (self.global_epoch + 1, lr))
# 載入Mnist手寫數(shù)據(jù)集
mnist = keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
x_train = np.expand_dims(x_train,-1)
x_test = np.expand_dims(x_test,-1)
#-----------------------------#
#   創(chuàng)建模型
#-----------------------------#
inputs = Input([28,28,1])
x = Conv2D(32, kernel_size= 5,padding = 'same',activation="relu")(inputs)
x = MaxPooling2D(pool_size = 2, strides = 2, padding = 'same',)(x)
x = Conv2D(64, kernel_size= 5,padding = 'same',activation="relu")(x)
x = MaxPooling2D(pool_size = 2, strides = 2, padding = 'same',)(x)
x = Flatten()(x)
x = Dense(1024)(x)
x = Dense(256)(x)
out = Dense(10, activation='softmax')(x)
model = Model(inputs,out)
# 設(shè)定優(yōu)化器，loss，計算準(zhǔn)確率
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
# 設(shè)置訓(xùn)練參數(shù)
epochs = 10
init_epoch = 0
# 每一次訓(xùn)練使用多少個Batch
batch_size = 31
# 最大學(xué)習(xí)率
learning_rate_base = 1e-3
sample_count = len(x_train)
# 學(xué)習(xí)率
exponent_lr = ExponentDecayScheduler(learning_rate_base = learning_rate_base,
                                    global_epoch_init = init_epoch,
                                    decay_rate = 0.9,
                                    min_learn_rate = 1e-6
                                    )
# 利用fit進(jìn)行訓(xùn)練
model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size,
            verbose=1, callbacks=[exponent_lr])
plt.plot(exponent_lr.learning_rates)
plt.xlabel('Step', fontsize=20)
plt.ylabel('lr', fontsize=20)
plt.axis([0, epochs, 0, learning_rate_base*1.1])
plt.xticks(np.arange(0, epochs, 1))
plt.grid()
plt.title('lr decay with exponent', fontsize=20)
plt.show()

import numpy as np
import matplotlib.pyplot as plt
import keras
from keras import backend as K
from keras.layers import Flatten,Conv2D,Dropout,Input,Dense,MaxPooling2D
from keras.models import Model
def cosine_decay_with_warmup(global_step,
                             learning_rate_base,
                             total_steps,
                             warmup_learning_rate=0.0,
                             warmup_steps=0,
                             hold_base_rate_steps=0,
                             min_learn_rate=0,
                             ):
    """
    參數(shù)：
            global_step: 上面定義的Tcur，記錄當(dāng)前執(zhí)行的步數(shù)。
            learning_rate_base：預(yù)先設(shè)置的學(xué)習(xí)率，當(dāng)warm_up階段學(xué)習(xí)率增加到learning_rate_base，就開始學(xué)習(xí)率下降。
            total_steps: 是總的訓(xùn)練的步數(shù)，等于epoch*sample_count/batch_size,(sample_count是樣本總數(shù)，epoch是總的循環(huán)次數(shù))
            warmup_learning_rate: 這是warm up階段線性增長的初始值
            warmup_steps: warm_up總的需要持續(xù)的步數(shù)
            hold_base_rate_steps: 這是可選的參數(shù)，即當(dāng)warm up階段結(jié)束后保持學(xué)習(xí)率不變，知道hold_base_rate_steps結(jié)束后才開始學(xué)習(xí)率下降
    """
    if total_steps < warmup_steps:
        raise ValueError('total_steps must be larger or equal to '
                            'warmup_steps.')
    #這里實現(xiàn)了余弦退火的原理，設(shè)置學(xué)習(xí)率的最小值為0，所以簡化了表達(dá)式
    learning_rate = 0.5 * learning_rate_base * (1 + np.cos(np.pi *
        (global_step - warmup_steps - hold_base_rate_steps) / float(total_steps - warmup_steps - hold_base_rate_steps)))
    #如果hold_base_rate_steps大于0，表明在warm up結(jié)束后學(xué)習(xí)率在一定步數(shù)內(nèi)保持不變
    if hold_base_rate_steps > 0:
        learning_rate = np.where(global_step > warmup_steps + hold_base_rate_steps,
                                    learning_rate, learning_rate_base)
    if warmup_steps > 0:
        if learning_rate_base < warmup_learning_rate:
            raise ValueError('learning_rate_base must be larger or equal to '
                                'warmup_learning_rate.')
        #線性增長的實現(xiàn)
        slope = (learning_rate_base - warmup_learning_rate) / warmup_steps
        warmup_rate = slope * global_step + warmup_learning_rate
        #只有當(dāng)global_step 仍然處于warm up階段才會使用線性增長的學(xué)習(xí)率warmup_rate，否則使用余弦退火的學(xué)習(xí)率learning_rate
        learning_rate = np.where(global_step < warmup_steps, warmup_rate,
                                    learning_rate)
    learning_rate = max(learning_rate,min_learn_rate)
    return learning_rate
class WarmUpCosineDecayScheduler(keras.callbacks.Callback):
    """
    繼承Callback，實現(xiàn)對學(xué)習(xí)率的調(diào)度
    """
    def __init__(self,
                 learning_rate_base,
                 total_steps,
                 global_step_init=0,
                 warmup_learning_rate=0.0,
                 warmup_steps=0,
                 hold_base_rate_steps=0,
                 min_learn_rate=0,
                 verbose=0):
        super(WarmUpCosineDecayScheduler, self).__init__()
        # 基礎(chǔ)的學(xué)習(xí)率
        self.learning_rate_base = learning_rate_base
        # 總共的步數(shù)，訓(xùn)練完所有世代的步數(shù)epochs * sample_count / batch_size
        self.total_steps = total_steps
        # 全局初始化step
        self.global_step = global_step_init
        # 熱調(diào)整參數(shù)
        self.warmup_learning_rate = warmup_learning_rate
        # 熱調(diào)整步長，warmup_epoch * sample_count / batch_size
        self.warmup_steps = warmup_steps
        self.hold_base_rate_steps = hold_base_rate_steps
        # 參數(shù)顯示  
        self.verbose = verbose
        # learning_rates用于記錄每次更新后的學(xué)習(xí)率，方便圖形化觀察
        self.min_learn_rate = min_learn_rate
        self.learning_rates = []
	#更新global_step，并記錄當(dāng)前學(xué)習(xí)率
    def on_batch_end(self, batch, logs=None):
        self.global_step = self.global_step + 1
        lr = K.get_value(self.model.optimizer.lr)
        self.learning_rates.append(lr)
	#更新學(xué)習(xí)率
    def on_batch_begin(self, batch, logs=None):
        lr = cosine_decay_with_warmup(global_step=self.global_step,
                                      learning_rate_base=self.learning_rate_base,
                                      total_steps=self.total_steps,
                                      warmup_learning_rate=self.warmup_learning_rate,
                                      warmup_steps=self.warmup_steps,
                                      hold_base_rate_steps=self.hold_base_rate_steps,
                                      min_learn_rate = self.min_learn_rate)
        K.set_value(self.model.optimizer.lr, lr)
        if self.verbose > 0:
            print('\nBatch %05d: setting learning '
                  'rate to %s.' % (self.global_step + 1, lr))
# 載入Mnist手寫數(shù)據(jù)集
mnist = keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
x_train = np.expand_dims(x_train,-1)
x_test = np.expand_dims(x_test,-1)
#-----------------------------#
#   創(chuàng)建模型
#-----------------------------#
inputs = Input([28,28,1])
x = Conv2D(32, kernel_size= 5,padding = 'same',activation="relu")(inputs)
x = MaxPooling2D(pool_size = 2, strides = 2, padding = 'same',)(x)
x = Conv2D(64, kernel_size= 5,padding = 'same',activation="relu")(x)
x = MaxPooling2D(pool_size = 2, strides = 2, padding = 'same',)(x)
x = Flatten()(x)
x = Dense(1024)(x)
x = Dense(256)(x)
out = Dense(10, activation='softmax')(x)
model = Model(inputs,out)
# 設(shè)定優(yōu)化器，loss，計算準(zhǔn)確率
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
# 設(shè)置訓(xùn)練參數(shù)
epochs = 10
# 預(yù)熱期
warmup_epoch = 3
# 每一次訓(xùn)練使用多少個Batch
batch_size = 16
# 最大學(xué)習(xí)率
learning_rate_base = 1e-3
sample_count = len(x_train)
# 總共的步長
total_steps = int(epochs * sample_count / batch_size)
# 預(yù)熱步長
warmup_steps = int(warmup_epoch * sample_count / batch_size)
# 學(xué)習(xí)率
warm_up_lr = WarmUpCosineDecayScheduler(learning_rate_base=learning_rate_base,
                                            total_steps=total_steps,
                                            warmup_learning_rate=1e-5,
                                            warmup_steps=warmup_steps,
                                            hold_base_rate_steps=5,
                                            min_learn_rate = 1e-6
                                            )
# 利用fit進(jìn)行訓(xùn)練
model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size,
            verbose=1, callbacks=[warm_up_lr])
plt.plot(warm_up_lr.learning_rates)
plt.xlabel('Step', fontsize=20)
plt.ylabel('lr', fontsize=20)
plt.axis([0, total_steps, 0, learning_rate_base*1.1])
plt.xticks(np.arange(0, epochs, 1))
plt.grid()
plt.title('Cosine decay with warmup', fontsize=20)
plt.show()

import numpy as np
import matplotlib.pyplot as plt
import keras
from keras import backend as K
from keras.layers import Flatten,Conv2D,Dropout,Input,Dense,MaxPooling2D
from keras.models import Model
def cosine_decay_with_warmup(global_step,
                             learning_rate_base,
                             total_steps,
                             warmup_learning_rate=0.0,
                             warmup_steps=0,
                             hold_base_rate_steps=0,
                             min_learn_rate=0,
                             ):
    """
    參數(shù)：
            global_step: 上面定義的Tcur，記錄當(dāng)前執(zhí)行的步數(shù)。
            learning_rate_base：預(yù)先設(shè)置的學(xué)習(xí)率，當(dāng)warm_up階段學(xué)習(xí)率增加到learning_rate_base，就開始學(xué)習(xí)率下降。
            total_steps: 是總的訓(xùn)練的步數(shù)，等于epoch*sample_count/batch_size,(sample_count是樣本總數(shù)，epoch是總的循環(huán)次數(shù))
            warmup_learning_rate: 這是warm up階段線性增長的初始值
            warmup_steps: warm_up總的需要持續(xù)的步數(shù)
            hold_base_rate_steps: 這是可選的參數(shù)，即當(dāng)warm up階段結(jié)束后保持學(xué)習(xí)率不變，知道hold_base_rate_steps結(jié)束后才開始學(xué)習(xí)率下降
    """
    if total_steps < warmup_steps:
        raise ValueError('total_steps must be larger or equal to '
                            'warmup_steps.')
    #這里實現(xiàn)了余弦退火的原理，設(shè)置學(xué)習(xí)率的最小值為0，所以簡化了表達(dá)式
    learning_rate = 0.5 * learning_rate_base * (1 + np.cos(np.pi *
        (global_step - warmup_steps - hold_base_rate_steps) / float(total_steps - warmup_steps - hold_base_rate_steps)))
    #如果hold_base_rate_steps大于0，表明在warm up結(jié)束后學(xué)習(xí)率在一定步數(shù)內(nèi)保持不變
    if hold_base_rate_steps > 0:
        learning_rate = np.where(global_step > warmup_steps + hold_base_rate_steps,
                                    learning_rate, learning_rate_base)
    if warmup_steps > 0:
        if learning_rate_base < warmup_learning_rate:
            raise ValueError('learning_rate_base must be larger or equal to '
                                'warmup_learning_rate.')
        #線性增長的實現(xiàn)
        slope = (learning_rate_base - warmup_learning_rate) / warmup_steps
        warmup_rate = slope * global_step + warmup_learning_rate
        #只有當(dāng)global_step 仍然處于warm up階段才會使用線性增長的學(xué)習(xí)率warmup_rate，否則使用余弦退火的學(xué)習(xí)率learning_rate
        learning_rate = np.where(global_step < warmup_steps, warmup_rate,
                                    learning_rate)
    learning_rate = max(learning_rate,min_learn_rate)
    return learning_rate
class WarmUpCosineDecayScheduler(keras.callbacks.Callback):
    """
    繼承Callback，實現(xiàn)對學(xué)習(xí)率的調(diào)度
    """
    def __init__(self,
                 learning_rate_base,
                 total_steps,
                 global_step_init=0,
                 warmup_learning_rate=0.0,
                 warmup_steps=0,
                 hold_base_rate_steps=0,
                 min_learn_rate=0,
                 # interval_epoch代表余弦退火之間的最低點
                 interval_epoch=[0.05, 0.15, 0.30, 0.50],
                 verbose=0):
        super(WarmUpCosineDecayScheduler, self).__init__()
        # 基礎(chǔ)的學(xué)習(xí)率
        self.learning_rate_base = learning_rate_base
        # 熱調(diào)整參數(shù)
        self.warmup_learning_rate = warmup_learning_rate
        # 參數(shù)顯示  
        self.verbose = verbose
        # learning_rates用于記錄每次更新后的學(xué)習(xí)率，方便圖形化觀察
        self.min_learn_rate = min_learn_rate
        self.learning_rates = []
        self.interval_epoch = interval_epoch
        # 貫穿全局的步長
        self.global_step_for_interval = global_step_init
        # 用于上升的總步長
        self.warmup_steps_for_interval = warmup_steps
        # 保持最高峰的總步長
        self.hold_steps_for_interval = hold_base_rate_steps
        # 整個訓(xùn)練的總步長
        self.total_steps_for_interval = total_steps
        self.interval_index = 0
        # 計算出來兩個最低點的間隔
        self.interval_reset = [self.interval_epoch[0]]
        for i in range(len(self.interval_epoch)-1):
            self.interval_reset.append(self.interval_epoch[i+1]-self.interval_epoch[i])
        self.interval_reset.append(1-self.interval_epoch[-1])
	#更新global_step，并記錄當(dāng)前學(xué)習(xí)率
    def on_batch_end(self, batch, logs=None):
        self.global_step = self.global_step + 1
        self.global_step_for_interval = self.global_step_for_interval + 1
        lr = K.get_value(self.model.optimizer.lr)
        self.learning_rates.append(lr)
	#更新學(xué)習(xí)率
    def on_batch_begin(self, batch, logs=None):
        # 每到一次最低點就重新更新參數(shù)
        if self.global_step_for_interval in [0]+[int(i*self.total_steps_for_interval) for i in self.interval_epoch]:
            self.total_steps = self.total_steps_for_interval * self.interval_reset[self.interval_index]
            self.warmup_steps = self.warmup_steps_for_interval * self.interval_reset[self.interval_index]
            self.hold_base_rate_steps = self.hold_steps_for_interval * self.interval_reset[self.interval_index]
            self.global_step = 0
            self.interval_index += 1
        lr = cosine_decay_with_warmup(global_step=self.global_step,
                                      learning_rate_base=self.learning_rate_base,
                                      total_steps=self.total_steps,
                                      warmup_learning_rate=self.warmup_learning_rate,
                                      warmup_steps=self.warmup_steps,
                                      hold_base_rate_steps=self.hold_base_rate_steps,
                                      min_learn_rate = self.min_learn_rate)
        K.set_value(self.model.optimizer.lr, lr)
        if self.verbose > 0:
            print('\nBatch %05d: setting learning '
                  'rate to %s.' % (self.global_step + 1, lr))
# 載入Mnist手寫數(shù)據(jù)集
mnist = keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
x_train = np.expand_dims(x_train,-1)
x_test = np.expand_dims(x_test,-1)
y_train = y_train
#-----------------------------#
#   創(chuàng)建模型
#-----------------------------#
inputs = Input([28,28,1])
x = Conv2D(32, kernel_size= 5,padding = 'same',activation="relu")(inputs)
x = MaxPooling2D(pool_size = 2, strides = 2, padding = 'same',)(x)
x = Conv2D(64, kernel_size= 5,padding = 'same',activation="relu")(x)
x = MaxPooling2D(pool_size = 2, strides = 2, padding = 'same',)(x)
x = Flatten()(x)
x = Dense(1024)(x)
x = Dense(256)(x)
out = Dense(10, activation='softmax')(x)
model = Model(inputs,out)
# 設(shè)定優(yōu)化器，loss，計算準(zhǔn)確率
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
# 設(shè)置訓(xùn)練參數(shù)
epochs = 10
# 預(yù)熱期
warmup_epoch = 2
# 每一次訓(xùn)練使用多少個Batch
batch_size = 256
# 最大學(xué)習(xí)率
learning_rate_base = 1e-3
sample_count = len(x_train)
# 總共的步長
total_steps = int(epochs * sample_count / batch_size)
# 預(yù)熱步長
warmup_steps = int(warmup_epoch * sample_count / batch_size)
# 學(xué)習(xí)率
warm_up_lr = WarmUpCosineDecayScheduler(learning_rate_base=learning_rate_base,
                                            total_steps=total_steps,
                                            warmup_learning_rate=1e-5,
                                            warmup_steps=warmup_steps,
                                            hold_base_rate_steps=5,
                                            min_learn_rate=1e-6
                                            )
# 利用fit進(jìn)行訓(xùn)練
model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size,
            verbose=1, callbacks=[warm_up_lr])
plt.plot(warm_up_lr.learning_rates)
plt.xlabel('Step', fontsize=20)
plt.ylabel('lr', fontsize=20)
plt.axis([0, total_steps, 0, learning_rate_base*1.1])
plt.grid()
plt.title('Cosine decay with warmup', fontsize=20)
plt.show()