czy36mengfei/tensorflow2_tutorials_chinese

tensorflow2中文教程，持续更新(当前版本:tensorflow2.0)，tag: tensorflow 2.0 tutorials

tensorflow2_tutorials_chinese's Issues

基础CNN网络那节matplotlib显示数字有错

import matplotlib.pyplot as plt
plt.imshow(x_train[0])
plt.show()

似乎是之前下载的数组格式是(28，28，1)，reshape为(28，28)就可以正常显示

tf2.0 initializer怎么使用

用tf.random_normal_initializer()如何初始化变量？tf2.0中没有tf.get_variable()

Tensorflow2教程-基础CNN网络

您好，我在运行标题的程序时在命令行history = model.fit(x_train, y_train, batch_size=64, epochs=5, validation_split=0.1)遇到了这个问题： Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. TF版本是2.0正式版，电脑重启试过还是同样的问题。

我在安装tfds时候除了问题

ImportError Traceback (most recent call last)
in ()
1 from future import absolute_import, division, print_function, unicode_literals
2 # 安装tfds pip install tfds-nightly==1.0.2.dev201904090105
----> 3 import tensorflow_datasets as tfds
4 import tensorflow as tf
5 import tensorflow.keras.layers as layers

~\Anaconda3\lib\site-packages\tensorflow_datasets_init_.py in ()
44 # needs to happen before anything else, since the imports below will try to
45 # import tensorflow, too.
---> 46 from tensorflow_datasets.core import tf_compat
47 tf_compat.ensure_tf_install()
48

~\Anaconda3\lib\site-packages\tensorflow_datasets\core_init_.py in ()
16 """API to define datasets."""
17
---> 18 from tensorflow_datasets.core.dataset_builder import BeamBasedBuilder
19 from tensorflow_datasets.core.dataset_builder import BuilderConfig
20 from tensorflow_datasets.core.dataset_builder import DatasetBuilder

~\Anaconda3\lib\site-packages\tensorflow_datasets\core\dataset_builder.py in ()
31 from tensorflow_datasets.core import api_utils
32 from tensorflow_datasets.core import constants
---> 33 from tensorflow_datasets.core import dataset_utils
34 from tensorflow_datasets.core import download
35 from tensorflow_datasets.core import file_format_adapter

~\Anaconda3\lib\site-packages\tensorflow_datasets\core\dataset_utils.py in ()
27 from tensorflow_datasets.core import api_utils
28 from tensorflow_datasets.core import tf_compat
---> 29 from tensorflow_datasets.core import utils
30
31

~\Anaconda3\lib\site-packages\tensorflow_datasets\core\utils_init_.py in ()
19 from tensorflow_datasets.core.utils.py_utils import *
20 from tensorflow_datasets.core.utils.tf_utils import *
---> 21 from tensorflow_datasets.core.utils.tqdm_utils import *
22 from tensorflow_datasets.core.utils.version import Version
23 # pylint: enable=wildcard-import

~\Anaconda3\lib\site-packages\tensorflow_datasets\core\utils\tqdm_utils.py in ()
23 import contextlib
24
---> 25 from tqdm import auto as tqdm_lib
26
27

ImportError: cannot import name 'auto'

在搞ai项目 wx交流

方便ai技术交流，拉个微信群 希望能多交些搞ai应用和算法的朋友：

batch() and step_per_epoch

train_dataset = tf.data.Dataset.from_tensor_slices((x_train, y_train))
train_dataset = train_dataset.shuffle(buffer_size=1024).batch(64)

val_dataset = tf.data.Dataset.from_tensor_slices((x_val, y_val))
val_dataset = val_dataset.batch(64)

# model.fit(train_dataset, epochs=3)
# steps_per_epoch 每个epoch只训练几步
# validation_steps 每次验证，验证几步
model.fit(train_dataset, epochs=3, steps_per_epoch=100,
         validation_data=val_dataset, validation_steps=3)

你好，请问model.fit 里面的batch_size （epochs/step_per_epoch）是不是会覆盖掉前面的train_dataset.batch(64) ?

tf.keras问题

我将原本在tf.1.12下搭建的模型用tf.2.0编译，修改了几个修改的函数以后，又出现了 unhashable type: 'ListWrapper'的错误，出现在self.keras_model.add_loss(loss) 这一行上，

def compile(self, learning_rate, momentum):
        """Gets the model ready for training. Adds losses, regularization, and
        metrics. Then calls the Keras compile() function.
        """
        # Optimizer object
        optimizer = keras.optimizers.SGD(
            lr=learning_rate, momentum=momentum,
            clipnorm=self.config.GRADIENT_CLIP_NORM, )
        # Add Losses
        # First, clear previously set losses to avoid duplication
        self.keras_model._losses = []
        self.keras_model._per_input_losses = {}
        loss_names = ["loc_loss", "class_loss", "mask_loss"]
        for name in loss_names:
            layer = self.keras_model.get_layer(name)
            if layer.output in self.keras_model.losses:
                continue
            # Mean here because Dataparallel
            loss = tf.reduce_mean(layer.output, keepdims=True)
            self.keras_model.add_loss(tf.abs(loss))

        # Add L2 Reqgularization
        # Skip gamma and beta weights of batch normalization layers.
        reg_losses = [
            keras.regularizers.l2(self.config.WEIGHT_DECAY)(w) / tf.cast(tf.size(w), tf.float32)
            for w in self.keras_model.trainable_weights
            if 'gamma' not in w.name and 'beta' not in w.name]
        #
        self.keras_model.add_loss(tf.add_n(reg_losses))

        # Compile
        self.keras_model.compile(
            optimizer=optimizer,
            loss=[None] * len(self.keras_model.outputs))

        # Add metrics for losses
        for name in loss_names:
            if name in self.keras_model.metrics_names:
                continue
            layer = self.keras_model.get_layer(name)
            self.keras_model.metrics_names.append(name)
            loss = tf.reduce_mean(layer.output, keepdims=True)
            self.keras_model.metrics_tensors.append(loss)

这部分的全部代码是这样的，有人知道怎么修改嘛？

如何控制显存使用

tf1.x里可以使用
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
session = tf.Session(config=config)
来控制显存的使用

但是在tf2.0里，session已经被去除，请问如何控制显存的使用呢？

谢谢！

教程很好，可以出一些实战项目的教程嘛

tf2.0 在模型训练的时候还是用的静态图吧？

导入的tf2.0 print() 网络最后的输出 好像具体的数值还是不能出来 并不是显示的“ops.EagerTensor”

新增内容请求

能否增加一下tf.data 和 tf.feature_column的使用

请问，tf2如何实现共享变量？

tf1可以通过设用tf.variable_scope设置前缀，然后通过tf.get_variable获得相同name和维度的变量，从而实现权重共享。tf2如何实现呢？

022cnn的reshape

cnn中为什么有这个步骤，第一维的负一是什么意思
x_train = x_train.reshape((-1,28,28,1))
x_test = x_test.reshape((-1,28,28,1))

save model

how can i save the encoder and decoder model respectively for /024-AutoEncoder/cnn_vae.ipynb?

thank you very much

为什么我的tensorflow2.0不支持python的if 还有while 这些语句，即使我加了@tf.function

使用tensorflow2.0中的keras，如何嵌入预训练的词向量

transform

谁有transform的数据集？

tf2.0训练速度慢于1.x

Hi, @czy36mengfei 请问你有和tf1.0对比过训练的速度吗? 我这边看是不管用keras的optimizer来训练还是用GradientTape + tf.function，2.0的速度都显著慢于1.x，大概慢了2.5倍

CNN+RNN

咨询一下可否二次分发，修改

注明出处和作者。

tensorflow