小编Nah*_*hua的帖子

背景：

由于Keras运行在TensorFlow之上，因此对TensorFlow进行了标记，这更是一个普遍的深度学习问题。

我一直在研究Kaggle Digit Recognizer问题，并使用Keras训练了该任务的CNN模型。下面的模型具有我用于本次比赛的原始CNN结构，并且效果还不错。

def build_model1():
    model = models.Sequential()

    model.add(layers.Conv2D(32, (3, 3), padding="Same" activation="relu", input_shape=[28, 28, 1]))
    model.add(layers.MaxPooling2D((2, 2)))
    model.add(layers.Dropout(0.25))

    model.add(layers.Conv2D(64, (3, 3), padding="Same", activation="relu"))
    model.add(layers.MaxPooling2D((2, 2)))
    model.add(layers.Dropout(0.25))

    model.add(layers.Conv2D(64, (3, 3), padding="Same", activation="relu"))
    model.add(layers.MaxPooling2D((2, 2)))
    model.add(layers.Dropout(0.25))

    model.add(layers.Flatten())
    model.add(layers.Dense(64, activation="relu"))
    model.add(layers.Dropout(0.5))
    model.add(layers.Dense(10, activation="softmax"))

    return model

然后，我在Kaggle上阅读了一些其他笔记本，并借用了另一个CNN结构（复制如下），该结构比上面的结构好得多，因为它具有更好的准确性，更低的错误率，并且在过度拟合训练数据之前花费了更多的时间。

def build_model2():
    model = models.Sequential()

    model.add(layers.Conv2D(32, (5, 5),padding ='Same', activation='relu', input_shape = (28, 28, 1)))
    model.add(layers.Conv2D(32, (5, 5),padding = 'Same', activation ='relu'))
    model.add(layers.MaxPool2D((2, 2)))
    model.add(layers.Dropout(0.25))

    model.add(layers.Conv2D(64,(3, 3),padding = 'Same', activation ='relu'))
    model.add(layers.Conv2D(64, (3, …

Edit 1: Another person reported similar issue with asdf.

Edit 2: Another reported similar issue.

Edit 3: I uninstalled rvm, installed rvm, installed ruby and rails, and attempted some activities.

Edit 4: Another person mentioned that it's caused by the line I put into .zshrc (or .bash_profile/.bashrc): RUBYOPT: "-W:no-deprecated -W:no-experimental".

Problem: I have two projects, one using ruby-2.7.0, the other using ruby-2.6.5.

Every time I installed over two versions of ruby, only …