Hi, I wanna to extract the weights from VGG model in the keras library for further other usage. And I read the weights from the following script

from keras.applications.vgg16 import VGG16
model = VGG16()
weights = []
for layer in model.layers:
    weights.append(layer.get_weights())  # list of numpy array

And the shape of weights in CONV1 is tuple(3,3,3,64). Could you help me check the order of this tuple? Is it (width of filter, height of filter, channel of filter, number of kernel) or ( height of filter, width of filter, channel of filter, number of kernel)?

Thanks

没人管了吗？左侧目录失效，点不了，看起来好难受啊

keras.io/zh打卡子目录时，左栏目录有问题

作为参考文献的引用格式怎么写？

原文“Sequential顺序建模指引”小节中的“示例”标题下，”examples”链接失效
原链接为：https://github.com/keras-team/keras/tree/master/examples
按照上下文的意思，examples指向的链接似乎应当是：https://github.com/keras-team/keras/tree/master/keras/datasets

keras.backend.image_data_format()

原作者注释写成'channels_first'
不过现在 都是last了，

在英文文档中，侧边栏有一个 wy-side-scroll ，覆盖了父元素的 overflow-y: hidden; 属性为 scroll，而中文文档没有这个 nav，导致侧边栏无法滚动

输入：x_train=np.random.random((32,30,300)) y_train=np.random.randint(2,size=(32,30,1))
模型A：inputs=Input(shape=(None,300)) lstm=LSTM(40,return_sequences=True)(inputs) outer=TimeDistributed(Dense(1,activation='sigmoid'))(lstm)
模型B：‘inputs=Input(shape=(30,300))
lstm=LSTM(40,return_sequences=True)(inputs)
outer=TimeDistributed(Dense(1,activation='sigmoid'))(lstm)’
二者都不会报错，请问准确率各方面有什么区别吗？

We need to follow the English document recently, would you like to contribute?

Currently, these files need to be updates:

Contributions are welcome.

If you wanna contribute, please reply to this issue directly, we will mark you in this table. Thanks.

这些地方显示不正确也没人管。

关于backend里的函数翻译很不严谨，返回类型tf.variable、tf.tensor、keras tensor混乱不清。感觉就是程序源码注释里的翻译；eval()应该是计算，翻译成估计

如图所示， 在ubuntu下采用火狐浏览器， 打开https://keras.io/zh/总是出现左边的排版错误， 其中： 快速开始， 模型， layers， 数据预处理等4章无法打开， 这个问题已经持续了有一周了。

目前的 keras 已经被集成到 tensorflow2 ，是否有意向对文档进行更新？

The version released by github.io is inconsistent with the version compiled by the current project, and the page jump style is different.

The home version of the deploy version is different from other theme styles. For example, click on the quick start, the directory bar can not view all the directories, but I compiled the version in the repository without this problem.

please deploy the latest version to github.io!

I'm Chinese and would like to contribute to the keras docs-zh translation.
Is there any ways that we can help and share the task? Thanks.

Feedback: The link to an edit button in the upper right corner of the website is wrong, please update. Maybe you can link it to this project.

仓库中关于example的资料已经没有了。
跳转后显示Page not found

https://keras.io/zh/ 左上角输入中文搜索失败，英文OK，验证多浏览器均有此问题。

The introduction of Convolutional Layers Conv1D:

参数:
padding: "valid", "causal" 或 "same" 之一 (大小写敏感) "valid" 表示「不填充」。 "same" 表示填充输入以使输出具有与原始输入相同的长度。 "causal" 表示因果（膨胀）卷积， 例如，output[t] 不依赖于 input[t+1:]， 在模型不应违反时间顺序的时间数据建模时非常有用。 在模型不应违反时间顺序的时间数据建模时非常有用。 详见 WaveNet: A Generative Model for Raw Audio, section 2.1。

The sentence of “在模型不应违反时间顺序的时间数据建模时非常有用。” is repeated.

有PDF版本的吗？或者可以生成PDF？