keras model의 파라미터 초기화에 대하여

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
from LossHistory import *
from data_load import load_mnist
from NNClass.nnclass import *
from NNClass.neural_nework_class import *
import sys
import numpy as np
from tensorflow.keras.models import Model
import matplotlib.pyplot as plt
import time

def Mymodel_get(save_path=None, file1=None, file2=None, mini_batch_size=128, epochs=1,X_train=None, y_train=None,
                X_test=None, y_test=None):
    save_path = save_path.format(file1, file2)
    model = Mymodel()
    model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])
    loss = Loss_and_Accuracy(model=model, save_path=save_path, step_per=True,
                             train_data=X_train, train_label=y_train, test_data=X_test,test_label=y_test)
    #loss = LossHistory(model=model, val_data=X_test, val_label=y_test, result_save_path=save_path,step_per_val=True)
    st = time.time()
    model.fit(x=X_train, y=y_train, batch_size=mini_batch_size, epochs=epochs,
                     shuffle=False, verbose=1, validation_data=(X_test, y_test), callbacks=[loss])
    print(time.time() - st ,"times")
    loss.save_result()

import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

from LossHistory import *

from data_load import load_mnist

from NNClass.nnclass import *

from NNClass.neural_nework_class import *

import sys

import numpy as np

from tensorflow.keras.models import Model

import matplotlib.pyplot as plt

import time

def Mymodel_get(save_path=None, file1=None, file2=None, mini_batch_size=128, epochs=1,X_train=None, y_train=None,

X_test=None, y_test=None):

save_path = save_path.format(file1, file2)

model = Mymodel()

model.compile(optimizer='sgd', loss='categorical_crossentropy', metrics=['accuracy'])

loss = Loss_and_Accuracy(model=model, save_path=save_path, step_per=True,

train_data=X_train, train_label=y_train, test_data=X_test,test_label=y_test)

#loss = LossHistory(model=model, val_data=X_test, val_label=y_test, result_save_path=save_path,step_per_val=True)

st = time.time()

model.fit(x=X_train, y=y_train, batch_size=mini_batch_size, epochs=epochs,

shuffle=False, verbose=1, validation_data=(X_test, y_test), callbacks=[loss])

print(time.time() - st ,"times")

loss.save_result()

import tensorflow as tf
#tf.random.set_seed(41)
from tensorflow.keras import Sequential, layers
from tensorflow.keras.models import Model

class Mymodel(Model):
    def __init__(self):
        super(Mymodel, self).__init__()
        self.dense1 = layers.Dense(128)
        self.activation1 = layers.ReLU()
        self.dense2 = layers.Dense(10, activation=tf.nn.softmax)

    def call(self, inputs):
        x = self.dense1(inputs)
        x = self.activation1(x)
        return self.dense2(x)

import tensorflow as tf

#tf.random.set_seed(41)

from tensorflow.keras import Sequential, layers

from tensorflow.keras.models import Model

class Mymodel(Model):

def __init__(self):

super(Mymodel, self).__init__()

self.dense1 = layers.Dense(128)

self.activation1 = layers.ReLU()

self.dense2 = layers.Dense(10, activation=tf.nn.softmax)

def call(self, inputs):

x = self.dense1(inputs)

x = self.activation1(x)

return self.dense2(x)

아직 정확히 작동방식을 몰라 두 가지 조건에 대해서 실험 Model.fit(shuffle = True) , class Mymodel 위쪽 import 영역 tf.random.seed(41) Model.fit(shuffle = True)…

머신러닝 - Keras2021년 5월 4일2021년 5월 5일

케라스에서 epoch안의 step에서 train data와 test data는 어떻게 계산되는가?

keras : 60,000개의 학습데이터를 섞지 않고 128개의 batch 크기, 2 epoch 을 step 당 계산 python : python도 마찬가지로 같은 구조의 같은 파라미터 초기화,…

머신러닝 - Keras2019년 8월 27일2019년 8월 27일

Keras 따라하기 Early Stopping 사용법

regularization 기법 중 하나인 조기종료(Early stopping)를 Keras를 통해 구현해본다. keras 의 callbacks 에 Earlystopping 기능을 지원한다. 설명을 보면 Stop training when a monitored quantity has…

머신러닝 - Keras2019년 8월 26일2019년 8월 26일

Keras 따라하기 모델저장과 복원

Keras 를 사용하여 모델저장과 복원하는 방법을 배워본다. 모델의 신속성을 위해 샘플은 처음 1,000개만 사용한다.

from __future__ import absolute_import, division, print_function, unicode_literals
import os
import tensorflow as tf
from tensorflow import keras

(train_images, train_labels),(test_images, test_labels) = tf.keras.datasets.mnist.load_data()

train_labels = train_labels[:1000]
test_labels = test_labels[:1000]

train_images = train_images[:1000].reshape(-1,28 * 28) / 255.
test_images = test_images[:1000].reshape(-1, 28 * 28) / 255.

from __future__ import absolute_import, division, print_function, unicode_literals

import os

import tensorflow as tf

from tensorflow import keras

(train_images, train_labels),(test_images, test_labels) = tf.keras.datasets.mnist.load_data()

train_labels = train_labels[:1000]

test_labels = test_labels[:1000]

train_images = train_images[:1000].reshape(-1,28 * 28) / 255.

test_images = test_images[:1000].reshape(-1, 28 * 28) / 255.

가중치를 저장하고 불러오는 예제를 위해 모델을 정의한다.

def create_model():
    model = tf.keras.models.Sequential([
        keras.layers.Dense(512, activation=tf.keras.activations.relu, input_shape=(784,)),
        keras.layers.Dropout(0.2),
        keras.layers.Dense(10, activation=tf.keras.activations.softmax)
    ])
    
    model.compile(optimizer=tf.keras.optimizers.Adam(),
                 loss=tf.keras.losses.sparse_categorical_crossentropy,
                 metrics=['accuracy'])
    return model

model = create_model()
model.summary()

def create_model():

model = tf.keras.models.Sequential([

keras.layers.Dense(512, activation=tf.keras.activations.relu, input_shape=(784,)),

keras.layers.Dropout(0.2),

keras.layers.Dense(10, activation=tf.keras.activations.softmax)

])

model.compile(optimizer=tf.keras.optimizers.Adam(),

loss=tf.keras.losses.sparse_categorical_crossentropy,

metrics=['accuracy'])

return model

model = create_model()

model.summary()

…