在TensorFlow 2.5环境中安装keras_tuner

Question

我正在尝试使用keras_tuner。随机搜索来找到最适合我的模型的参数。我使用以下命令在anaconda命令提示符中安装了keras_tuner:

conda install -c conda-forge keras-tuner

然后导入包如下:导入keras_tuner为kt

但是当我调用kt时。随机搜索，我得到以下错误消息:

tuner_search = kt。RandomSearch (build_modelAttributeError:部分初始化模块'keras_tuner'没有属性'RandomSearch'(很可能是由于循环导入)。

下面是我的代码:

import tensorflow as tf
import keras_tuner as kt
from tensorflow import keras
import os
import cv2
import pandas as pd
import numpy as np
from tensorflow.keras.utils import to_categorical
from sklearn.preprocessing import LabelEncoder
from sklearn.model_selection import train_test_split

f= pd.read_csv('CELLS_ALL.csv')
Labels= f['labels']
path_dir = 'C:\Users\user1\PycharmProjects\imageDataGen\images\'
img_all = []
for i in os.listdir(path_dir):
img_sig = cv2.imread(path_dir+i)
img_sig = cv2.resize(img_sig, (50, 50))
img_all.append(img_sig)
x = np.array(img_all, dtype="float") / 255.0
y = Labels
le = LabelEncoder()
y = le.fit_transform(y)
y = to_categorical(y)
#print(labels)

(trainX, testX, trainY, testY) = train_test_split(x, y, test_size=0.25, random_state=42)
# for cnn images should me of shape (len(training,size,size, channel)
trainX= trainX.reshape(len(trainX),50,50,3)
testX = testX.reshape(len(testX),50,50,3)

def build_model(hp):
model = keras.Sequential([
keras.layers.Conv2D(
filters=hp.Int('conv_1_filter', min_value=128, max_value=256, step=16),
kernel_size=hp.Choice('conv_1_kernel', values=[3, 5]),
activation='relu',
input_shape=(50, 50, 3)
),
keras.layers.Conv2D(
filters=hp.Int('conv_2_filter', min_value=128, max_value=256, step=16),
kernel_size=hp.Choice('conv_2_kernel', values=[3, 5]),
activation='relu'
),
keras.layers.Flatten(),
keras.layers.Dense(
units=hp.Int('dense_1_units', min_value=32, max_value=128, step=16),
activation='relu'
),
keras.layers.Dense(15, activation='softmax')
])
model.compile(optimizer=keras.optimizers.Adam(hp.Choice('learning_rate', values=[1e-2, 1e-3])),
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
return model
tuner_search= kt.RandomSearch(build_model,
objective='val_accuracy',
max_trials=5,directory='tune',project_name="cnn model tunning")

我的问题是如何安装keras_tuner并使用RandomSearch?

Answer 1

很可能你有一个本地文件(当前文件)，它与库(你试图导入的模块)的确切名称相同，因此有循环引用，因为Python认为它是一个模块)。

更改运行代码的文件名(避免命名与库/模块名称完全重叠)，看看它是否有效。