深度学习——tensorflow2.0搭建分类和回归模型
1.环境安装
步骤1：打开anaconda prompt
步骤2：conda create -n tf2-gpu-py3 python=3.6.7 anaconda
conda activate tf2-gpu-py3
步骤3：python -m pip install --upgrade pip
步骤4：pip install -i https://pypi.tuna.tsinghua.edu.cn/simple tensorflow-gpu==2.0.0-alpha0
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple numpy==1.16.0
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple matplotlib sklearn pandas jupyter
步骤5：https://developer.nvidia.com/cuda-toolkit-archive下载CUDA，完成后点击安装
步骤6：输入cmd，进入终端，输入set path，查看CUDA的目录，如C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
步骤7：https://developer.nvidia.com/rdp/cudnn-archive下载对应CUDA版本的cuDNN，如Download cuDNN v7.6.1 (June 24, 2019), for CUDA 10.0，解压后，复制到C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0目录，下载cuDNN需要登陆账号，可以选择使用微信或QQ登陆
步骤8：添加到系统环境变量path中：
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\libnvvp
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include
步骤9：重启电脑，打开anaconda prompt，conda activate tf2-gpu-py3
步骤10：python
>>> import tensorflow as tf
>>> print(tf.__version__)
2.0.0-alpha0
>>> tf.test.is_gpu_available()
2019-08-26 16:12:05.591836: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2019-08-26 16:12:05.760799: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library nvcuda.dll
2019-08-26 16:12:06.323888: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1467] Found device 0 with properties:
name: GeForce GTX 965M major: 5 minor: 2 memoryClockRate(GHz): 1.15
pciBusID: 0000:01:00.0
totalMemory: 4.00GiB freeMemory: 3.33GiB
2019-08-26 16:12:06.336264: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1546] Adding visible gpu devices: 0
2019-08-26 16:12:17.835996: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1015] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-08-26 16:12:17.842351: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1021] 0
2019-08-26 16:12:17.846232: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1034] 0: N
2019-08-26 16:12:17.853620: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1149] Created TensorFlow device (/device:GPU:0 with 3035 MB memory) -> physical GPU (device: 0, name: GeForce GTX 965M, pci bus id: 0000:01:00.0, compute capability: 5.2)
True
>>> exit()
步骤11：jupyer notebook
ctrl+c
conda deactivate
2.分类模型
import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import time
import tensorflow as tffrom tensorflow import keras
print(tf.__version__)
2.0.0-alpha0
# 步骤1：数据导入
fashion_mnist = keras.datasets.fashion_mnist
(x_train_all, y_train_all), (x_test, y_test) = fashion_mnist.load_data()
x_valid, x_train = x_train_all[:5000], x_train_all[5000:]
y_valid, y_train = y_train_all[:5000], y_train_all[5000:]print(x_valid.shape, y_valid.shape)
print(x_train.shape, y_train.shape)
print(x_test.shape, y_test.shape)
(5000, 28, 28) (5000,) (55000, 28, 28) (55000,) (10000, 28, 28) (10000,)
# 步骤2：数据预处理
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
x_train_scaled = scaler.fit_transform(x_train.astype(np.float32).reshape(-1, 784)).reshape(-1, 28, 28)
x_valid_scaled = scaler.transform(x_valid.astype(np.float32).reshape(-1, 784)).reshape(-1, 28, 28)
x_test_scaled = scaler.transform(x_test.astype(np.float32).reshape(-1, 784)).reshape(-1, 28, 28)
# 步骤3：模型搭建
model = keras.models.Sequential()
model.add(keras.layers.Flatten(input_shape=[28, 28]))
for _ in range(20):
model.add(keras.layers.Dense(100, activation="selu"))
model.add(keras.layers.AlphaDropout(rate=0.5))
model.add(keras.layers.Dense(10, activation="softmax"))
model.compile(loss="sparse_categorical_crossentropy", optimizer = "sgd", metrics = ["accuracy"])
model.summary()
Model: "sequential" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten (Flatten) (None, 784) 0 _________________________________________________________________ dense (Dense) (None, 100) 78500 _________________________________________________________________ dense_1 (Dense) (None, 100) 10100 _________________________________________________________________ dense_2 (Dense) (None, 100) 10100 _________________________________________________________________ dense_3 (Dense) (None, 100) 10100 _________________________________________________________________ dense_4 (Dense) (None, 100) 10100 _________________________________________________________________ dense_5 (Dense) (None, 100) 10100 _________________________________________________________________ dense_6 (Dense) (None, 100) 10100 _________________________________________________________________ dense_7 (Dense) (None, 100) 10100 _________________________________________________________________ dense_8 (Dense) (None, 100) 10100 _________________________________________________________________ dense_9 (Dense) (None, 100) 10100 _________________________________________________________________ dense_10 (Dense) (None, 100) 10100 _________________________________________________________________ dense_11 (Dense) (None, 100) 10100 _________________________________________________________________ dense_12 (Dense) (None, 100) 10100 _________________________________________________________________ dense_13 (Dense) (None, 100) 10100 _________________________________________________________________ dense_14 (Dense) (None, 100) 10100 _________________________________________________________________ dense_15 (Dense) (None, 100) 10100 _________________________________________________________________ dense_16 (Dense) (None, 100) 10100 _________________________________________________________________ dense_17 (Dense) (None, 100) 10100 _________________________________________________________________ dense_18 (Dense) (None, 100) 10100 _________________________________________________________________ dense_19 (Dense) (None, 100) 10100 _________________________________________________________________ alpha_dropout (AlphaDropout) (None, 100) 0 _________________________________________________________________ dense_20 (Dense) (None, 10) 1010 ================================================================= Total params: 271,410 Trainable params: 271,410 Non-trainable params: 0 _________________________________________________________________
# 步骤4：训练和验证模型
logdir = ".\dnn-callbacks"
if not os.path.exists(logdir):
os.mkdir(logdir)
output_model_file = os.path.join(logdir,"fashion_mnist_model.h5")callbacks = [
keras.callbacks.TensorBoard(logdir),
keras.callbacks.ModelCheckpoint(output_model_file,save_best_only = True),
keras.callbacks.EarlyStopping(patience=5, min_delta=1e-3),
]history = model.fit(x_train_scaled, y_train, epochs=10,validation_data=(x_valid_scaled, y_valid),callbacks = callbacks)
Train on 55000 samples, validate on 5000 samples Epoch 1/10 55000/55000 [==============================] - 9s 171us/sample - loss: 1.2599 - accuracy: 0.5799 - val_loss: 0.8497 - val_accuracy: 0.7878 Epoch 2/10 55000/55000 [==============================] - 8s 150us/sample - loss: 0.7253 - accuracy: 0.7650 - val_loss: 0.7718 - val_accuracy: 0.8244 Epoch 3/10 55000/55000 [==============================] - 9s 156us/sample - loss: 0.5915 - accuracy: 0.8106 - val_loss: 0.7111 - val_accuracy: 0.8424 Epoch 4/10 55000/55000 [==============================] - 9s 157us/sample - loss: 0.5271 - accuracy: 0.8299 - val_loss: 0.6823 - val_accuracy: 0.8500 Epoch 5/10 55000/55000 [==============================] - 8s 150us/sample - loss: 0.4831 - accuracy: 0.8416 - val_loss: 0.6299 - val_accuracy: 0.8562 Epoch 6/10 55000/55000 [==============================] - 8s 140us/sample - loss: 0.4503 - accuracy: 0.8527 - val_loss: 0.6326 - val_accuracy: 0.8602 Epoch 7/10 55000/55000 [==============================] - 8s 153us/sample - loss: 0.4278 - accuracy: 0.8588 - val_loss: 0.6135 - val_accuracy: 0.8642 Epoch 8/10 55000/55000 [==============================] - 8s 152us/sample - loss: 0.4086 - accuracy: 0.8654 - val_loss: 0.6081 - val_accuracy: 0.8670 Epoch 9/10 55000/55000 [==============================] - 8s 154us/sample - loss: 0.3898 - accuracy: 0.8704 - val_loss: 0.6020 - val_accuracy: 0.8672 Epoch 10/10 55000/55000 [==============================] - 8s 154us/sample - loss: 0.3757 - accuracy: 0.8741 - val_loss: 0.6161 - val_accuracy: 0.8688
model.evaluate(x_test_scaled, y_test)
10000/10000 [==============================] - 1s 70us/sample - loss: 0.6765 - accuracy: 0.8572
Out[ ]:
[0.6764720353841782, 0.8572]
3.回归模型
import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import time
import tensorflow as tffrom tensorflow import keras
print(tf.__version__)
2.0.0-alpha0
from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split
housing = fetch_california_housing()
x_train_all, x_test, y_train_all, y_test = train_test_split(housing.data, housing.target, random_state = 6)
x_train, x_valid, y_train, y_valid = train_test_split(x_train_all, y_train_all, random_state = 7)
print(x_train.shape, y_train.shape)
print(x_valid.shape, y_valid.shape)
print(x_test.shape, y_test.shape)
(11610, 8) (11610,) (3870, 8) (3870,) (5160, 8) (5160,)
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler()
x_train_scaled = scaler.fit_transform(x_train)
x_valid_scaled = scaler.transform(x_valid)
x_test_scaled = scaler.transform(x_test)
def build_model(hidden_layers = 1, layer_size = 30, learning_rate = 3e-3):
model = keras.models.Sequential()
model.add(keras.layers.Dense(layer_size, activation='relu', input_shape=x_train.shape[1:]))
for _ in range(hidden_layers - 1):
model.add(keras.layers.Dense(layer_size, activation = 'relu'))
model.add(keras.layers.Dense(1))
optimizer = keras.optimizers.SGD(learning_rate)
model.compile(loss = 'mse', optimizer = optimizer)
return modelsklearn_model = keras.wrappers.scikit_learn.KerasRegressor(build_fn = build_model)
callbacks = [keras.callbacks.EarlyStopping(patience=5, min_delta=1e-2)]
history = sklearn_model.fit(x_train_scaled, y_train,validation_data = (x_valid_scaled, y_valid),epochs = 100,callbacks = callbacks)
Train on 11610 samples, validate on 3870 samples Epoch 1/100 11610/11610 [==============================] - 1s 57us/sample - loss: 1.4086 - val_loss: 0.7144 Epoch 2/100 11610/11610 [==============================] - 0s 36us/sample - loss: 0.6111 - val_loss: 0.6281 Epoch 3/100 11610/11610 [==============================] - 0s 40us/sample - loss: 0.5514 - val_loss: 0.5723 Epoch 4/100 11610/11610 [==============================] - 0s 39us/sample - loss: 0.5105 - val_loss: 0.5323 Epoch 5/100 11610/11610 [==============================] - 0s 41us/sample - loss: 0.4841 - val_loss: 0.5068 Epoch 6/100 11610/11610 [==============================] - 0s 39us/sample - loss: 0.4634 - val_loss: 0.4904 Epoch 7/100 11610/11610 [==============================] - 0s 36us/sample - loss: 0.4497 - val_loss: 0.4749 Epoch 8/100 11610/11610 [==============================] - 0s 41us/sample - loss: 0.4395 - val_loss: 0.4632 Epoch 9/100 11610/11610 [==============================] - 1s 47us/sample - loss: 0.4312 - val_loss: 0.4533 Epoch 10/100 11610/11610 [==============================] - 1s 48us/sample - loss: 0.4247 - val_loss: 0.4480 Epoch 11/100 11610/11610 [==============================] - 0s 43us/sample - loss: 0.4194 - val_loss: 0.4430 Epoch 12/100 11610/11610 [==============================] - 0s 33us/sample - loss: 0.4153 - val_loss: 0.4357 Epoch 13/100 11610/11610 [==============================] - 0s 33us/sample - loss: 0.4117 - val_loss: 0.4340 Epoch 14/100 11610/11610 [==============================] - 0s 34us/sample - loss: 0.4077 - val_loss: 0.4294 Epoch 15/100 11610/11610 [==============================] - 0s 36us/sample - loss: 0.4050 - val_loss: 0.4243 Epoch 16/100 11610/11610 [==============================] - 0s 32us/sample - loss: 0.4017 - val_loss: 0.4200 Epoch 17/100 11610/11610 [==============================] - 0s 32us/sample - loss: 0.3990 - val_loss: 0.4181 Epoch 18/100 11610/11610 [==============================] - 0s 33us/sample - loss: 0.3964 - val_loss: 0.4169 Epoch 19/100 11610/11610 [==============================] - 0s 32us/sample - loss: 0.3948 - val_loss: 0.4127 Epoch 20/100 11610/11610 [==============================] - 0s 40us/sample - loss: 0.3930 - val_loss: 0.4087 Epoch 21/100 11610/11610 [==============================] - 1s 48us/sample - loss: 0.3902 - val_loss: 0.4078 Epoch 22/100 11610/11610 [==============================] - 0s 36us/sample - loss: 0.3885 - val_loss: 0.4060 Epoch 23/100 11610/11610 [==============================] - 1s 45us/sample - loss: 0.3869 - val_loss: 0.4028 Epoch 24/100 11610/11610 [==============================] - 0s 40us/sample - loss: 0.3848 - val_loss: 0.4014 Epoch 25/100 11610/11610 [==============================] - 0s 33us/sample - loss: 0.3839 - val_loss: 0.4003 Epoch 26/100 11610/11610 [==============================] - 0s 38us/sample - loss: 0.3825 - val_loss: 0.3983 Epoch 27/100 11610/11610 [==============================] - 0s 38us/sample - loss: 0.3814 - val_loss: 0.3966 Epoch 28/100 11610/11610 [==============================] - 0s 38us/sample - loss: 0.3795 - val_loss: 0.3965 Epoch 29/100 11610/11610 [==============================] - 0s 33us/sample - loss: 0.3781 - val_loss: 0.3932
from scipy.stats import reciprocal
param_distribution = {
"hidden_layers":[1, 2, 3, 4],
"layer_size": np.arange(1, 100),
"learning_rate": reciprocal(1e-4, 1e-2),
}from sklearn.model_selection import RandomizedSearchCV
random_search_cv = RandomizedSearchCV(sklearn_model,
param_distribution,
n_iter = 10,
cv = 3,
n_jobs = 1)
random_search_cv.fit(x_train_scaled, y_train, epochs = 100,validation_data = (x_valid_scaled, y_valid),callbacks = callbacks)
RandomizedSearchCV(cv=3, error_score='raise-deprecating',
estimator=,
iid='warn', n_iter=10, n_jobs=1,
param_distributions={'hidden_layers': [1, 2, 3, 4],
'layer_size': array([ 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17,
18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34,
35, 36...7, 48, 49, 50, 51,
52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68,
69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85,
86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99]),
'learning_rate': },
pre_dispatch='2*n_jobs', random_state=None, refit=True,
return_train_score=False, scoring=None, verbose=0)
print(random_search_cv.best_params_)
print(random_search_cv.best_score_)
print(random_search_cv.best_estimator_)
{'hidden_layers': 3, 'layer_size': 84, 'learning_rate': 0.006306161764204613}
-0.33262315739032833
model = random_search_cv.best_estimator_.model
model.evaluate(x_test_scaled, y_test)
5160/5160 [==============================] - 0s 20us/sample - loss: 0.3405
Out[ ]:
0.3405195293500442