如何用Keras和几行代码构建大脑

内容概览 设置 介绍 培训,评估和推断 保存和序列化 使用相同的图表层来定义多个模型 所有模型都是可召唤的,就像层一样 操作复杂图形 topologies 具有多个输入和输出的模型 玩具的模型 设置 import numpy as np import tensorflow as tf from tensorflow import keras from keras import layers 介绍 《Keras》 这是创建模型的更灵活的方式。 API. 功能 API 可以处理非线性 topology 模型,共享层,甚至是多个输入或输出。 功能火 keras.Sequential 主要观点是,深度学习模型通常是层的定向非循环图(DAG)。 . 图表的 layers 考虑下面的模型: [Dense (64 单位, relu 激活)] [Dense (64 单位, relu 激活)] [Dense (10 单位, softmax 激活)] (输出: 10 类以上的概率分布的逻辑) 」 这是一个具有三个层的基本图表. 要使用功能 API 构建此模型,请先创建一个输入节点: inputs = keras.Input(shape=(784,)) 数据的形状被设置为 784 维矢量. 批量大小总是被省略,因为只指定了每个样本的形状。 例如,如果您有具有形状的图像输入 你会使用: (32, 32, 3) # Just for demonstration purposes. img_inputs = keras.Input(shape=(32, 32, 3)) 该 返回的内容包含有关形状和 输入数据,你输送到你的模型. 这里是形状: inputs dtype inputs.shape TensorShape([None, 784]) 这里是dtype: inputs.dtype tf.float32 您在层图中创建一个新的节点,将此层称为 对象: inputs dense = layers.Dense(64, activation="relu") x = dense(inputs) “层呼叫”操作就像从“输入”到您创建的这个层的箭头一样。 layer,你得到 就像产量一样。 dense x 让我们在层图中添加一些更多层: x = layers.Dense(64, activation="relu")(x) outputs = layers.Dense(10)(x) 在这一点上,你可以创建一个 通过在层图中指定其输入和输出: Model model = keras.Model(inputs=inputs, outputs=outputs, name="mnist_model") 让我们看看模型总结看起来是什么样子: model.summary() Model: "mnist_model" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= input_1 (InputLayer) [(None, 784)] 0 dense (Dense) (None, 64) 50240 dense_1 (Dense) (None, 64) 4160 dense_2 (Dense) (None, 10) 650 ================================================================= Total params: 55050 (215.04 KB) Trainable params: 55050 (215.04 KB) Non-trainable params: 0 (0.00 Byte) _________________________________________________________________ 您也可以将模型划成图表: keras.utils.plot_model(model, "my_first_model.png") 此外,可选地在绘图图中显示每个层的输入和输出形状: keras.utils.plot_model(model, "my_first_model_with_shape_info.png", show_shapes=True) 这个数字和代码几乎相同,在代码版本中,连接箭头被调用操作取代。 “层图”是深度学习模型的直观心理图像,功能API是创建模型的一种方式,可以密切反映这一点。 培训,评估和推断 培训、评估和推断对使用功能 API 构建的模型的工作方式完全相同 模特。 Sequential 该 课堂提供一个内置的培训循环( 方法)和内置的评估循环(该 请注意,您可以轻松 实施超越监督学习的培训流程(例如: )。 Model fit() evaluate() 定制这些环节 甘地 在这里,加载MNIST图像数据,重塑为矢量,将模型与数据相匹配(同时在验证分割上监控性能),然后在测试数据上评估模型: (x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data() x_train = x_train.reshape(60000, 784).astype("float32") / 255 x_test = x_test.reshape(10000, 784).astype("float32") / 255 model.compile( loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True), optimizer=keras.optimizers.RMSprop(), metrics=[keras.metrics.SparseCategoricalAccuracy()], ) history = model.fit(x_train, y_train, batch_size=64, epochs=2, validation_split=0.2) test_scores = model.evaluate(x_test, y_test, verbose=2) print("Test loss:", test_scores[0]) print("Test accuracy:", test_scores[1]) Epoch 1/2 750/750 [==============================] - 4s 3ms/step - loss: 0.3556 - sparse_categorical_accuracy: 0.8971 - val_loss: 0.1962 - val_sparse_categorical_accuracy: 0.9422 Epoch 2/2 750/750 [==============================] - 2s 2ms/step - loss: 0.1612 - sparse_categorical_accuracy: 0.9527 - val_loss: 0.1461 - val_sparse_categorical_accuracy: 0.9592 313/313 - 0s - loss: 0.1492 - sparse_categorical_accuracy: 0.9556 - 463ms/epoch - 1ms/step Test loss: 0.14915992319583893 Test accuracy: 0.9556000232696533 为了进一步阅读,请参见 指南。 培训与评估 保存和序列化 保存模型和序列化对使用功能 API 构建的模型的工作方式相同。 保存功能模型的标准方法是拨打 将整个模型保存为单个文件. 您可以随后从该文件中重建相同的模型,即使构建该模型的代码不再可用。 Sequential model.save() 这个保存的文件包括: 建筑模型 模型重量值(在训练过程中学到的) 模型培训 config,如果有(如已转到编译) 优化器及其状态,如果有(在您停止训练时重新启动) model.save("path_to_my_model.keras") del model # Recreate the exact same model purely from the file: model = keras.models.load_model("path_to_my_model.keras") 详情阅读模型 指南。 序列化与储蓄 使用相同的图表层来定义多个模型 在功能API中,模型是通过在层图中指定其输入和输出来创建的,这意味着可以使用单个层图来生成多个模型。 在下面的示例中,您使用相同的层堆栈来实例化两个模型:一个 model that turns image inputs into 16-dimensional vectors, and an end-to-end 培训的模型。 encoder autoencoder encoder_input = keras.Input(shape=(28, 28, 1), name="img") x = layers.Conv2D(16, 3, activation="relu")(encoder_input) x = layers.Conv2D(32, 3, activation="relu")(x) x = layers.MaxPooling2D(3)(x) x = layers.Conv2D(32, 3, activation="relu")(x) x = layers.Conv2D(16, 3, activation="relu")(x) encoder_output = layers.GlobalMaxPooling2D()(x) encoder = keras.Model(encoder_input, encoder_output, name="encoder") encoder.summary() x = layers.Reshape((4, 4, 1))(encoder_output) x = layers.Conv2DTranspose(16, 3, activation="relu")(x) x = layers.Conv2DTranspose(32, 3, activation="relu")(x) x = layers.UpSampling2D(3)(x) x = layers.Conv2DTranspose(16, 3, activation="relu")(x) decoder_output = layers.Conv2DTranspose(1, 3, activation="relu")(x) autoencoder = keras.Model(encoder_input, decoder_output, name="autoencoder") autoencoder.summary() Model: "encoder" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= img (InputLayer) [(None, 28, 28, 1)] 0 conv2d (Conv2D) (None, 26, 26, 16) 160 conv2d_1 (Conv2D) (None, 24, 24, 32) 4640 max_pooling2d (MaxPooling2 (None, 8, 8, 32) 0 D) conv2d_2 (Conv2D) (None, 6, 6, 32) 9248 conv2d_3 (Conv2D) (None, 4, 4, 16) 4624 global_max_pooling2d (Glob (None, 16) 0 alMaxPooling2D) ================================================================= Total params: 18672 (72.94 KB) Trainable params: 18672 (72.94 KB) Non-trainable params: 0 (0.00 Byte) _________________________________________________________________ Model: "autoencoder" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= img (InputLayer) [(None, 28, 28, 1)] 0 conv2d (Conv2D) (None, 26, 26, 16) 160 conv2d_1 (Conv2D) (None, 24, 24, 32) 4640 max_pooling2d (MaxPooling2 (None, 8, 8, 32) 0 D) conv2d_2 (Conv2D) (None, 6, 6, 32) 9248 conv2d_3 (Conv2D) (None, 4, 4, 16) 4624 global_max_pooling2d (Glob (None, 16) 0 alMaxPooling2D) reshape (Reshape) (None, 4, 4, 1) 0 conv2d_transpose (Conv2DTr (None, 6, 6, 16) 160 anspose) conv2d_transpose_1 (Conv2D (None, 8, 8, 32) 4640 Transpose) up_sampling2d (UpSampling2 (None, 24, 24, 32) 0 D) conv2d_transpose_2 (Conv2D (None, 26, 26, 16) 4624 Transpose) conv2d_transpose_3 (Conv2D (None, 28, 28, 1) 145 Transpose) ================================================================= Total params: 28241 (110.32 KB) Trainable params: 28241 (110.32 KB) Non-trainable params: 0 (0.00 Byte) _________________________________________________________________ 在这里,解码架构与编码架构是严格对称的,所以输出形状与输入形状相同。 . (28, 28, 1) A 的反面 layer 是 A layer 是一个 拉克 Conv2D Conv2DTranspose MaxPooling2D UpSampling2D 所有模型都是可调用的,就像层一样 你可以对待任何模型,就像它是一个层,通过呼吁它在一个 通过调用一个模型,你不仅重复使用该模型的架构,而且还重复使用其重量。 Input 要在行动中看到这一点,以下是关于自动编码器的另一种例子,它创建了一个编码器模型,一个解码器模型,并将它们链接成两个呼叫来获得自动编码器模型: encoder_input = keras.Input(shape=(28, 28, 1), name="original_img") x = layers.Conv2D(16, 3, activation="relu")(encoder_input) x = layers.Conv2D(32, 3, activation="relu")(x) x = layers.MaxPooling2D(3)(x) x = layers.Conv2D(32, 3, activation="relu")(x) x = layers.Conv2D(16, 3, activation="relu")(x) encoder_output = layers.GlobalMaxPooling2D()(x) encoder = keras.Model(encoder_input, encoder_output, name="encoder") encoder.summary() decoder_input = keras.Input(shape=(16,), name="encoded_img") x = layers.Reshape((4, 4, 1))(decoder_input) x = layers.Conv2DTranspose(16, 3, activation="relu")(x) x = layers.Conv2DTranspose(32, 3, activation="relu")(x) x = layers.UpSampling2D(3)(x) x = layers.Conv2DTranspose(16, 3, activation="relu")(x) decoder_output = layers.Conv2DTranspose(1, 3, activation="relu")(x) decoder = keras.Model(decoder_input, decoder_output, name="decoder") decoder.summary() autoencoder_input = keras.Input(shape=(28, 28, 1), name="img") encoded_img = encoder(autoencoder_input) decoded_img = decoder(encoded_img) autoencoder = keras.Model(autoencoder_input, decoded_img, name="autoencoder") autoencoder.summary() Model: "encoder" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= original_img (InputLayer) [(None, 28, 28, 1)] 0 conv2d_4 (Conv2D) (None, 26, 26, 16) 160 conv2d_5 (Conv2D) (None, 24, 24, 32) 4640 max_pooling2d_1 (MaxPoolin (None, 8, 8, 32) 0 g2D) conv2d_6 (Conv2D) (None, 6, 6, 32) 9248 conv2d_7 (Conv2D) (None, 4, 4, 16) 4624 global_max_pooling2d_1 (Gl (None, 16) 0 obalMaxPooling2D) ================================================================= Total params: 18672 (72.94 KB) Trainable params: 18672 (72.94 KB) Non-trainable params: 0 (0.00 Byte) _________________________________________________________________ Model: "decoder" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= encoded_img (InputLayer) [(None, 16)] 0 reshape_1 (Reshape) (None, 4, 4, 1) 0 conv2d_transpose_4 (Conv2D (None, 6, 6, 16) 160 Transpose) conv2d_transpose_5 (Conv2D (None, 8, 8, 32) 4640 Transpose) up_sampling2d_1 (UpSamplin (None, 24, 24, 32) 0 g2D) conv2d_transpose_6 (Conv2D (None, 26, 26, 16) 4624 Transpose) conv2d_transpose_7 (Conv2D (None, 28, 28, 1) 145 Transpose) ================================================================= Total params: 9569 (37.38 KB) Trainable params: 9569 (37.38 KB) Non-trainable params: 0 (0.00 Byte) _________________________________________________________________ Model: "autoencoder" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= img (InputLayer) [(None, 28, 28, 1)] 0 encoder (Functional) (None, 16) 18672 decoder (Functional) (None, 28, 28, 1) 9569 ================================================================= Total params: 28241 (110.32 KB) Trainable params: 28241 (110.32 KB) Non-trainable params: 0 (0.00 Byte) _________________________________________________________________ 正如你所看到的,模型可以嵌套:模型可以包含子模型(因为模型就像一层)。 例如,以下是如何将一组模型合并为一个模型,以平均他们的预测: 集合 def get_model(): inputs = keras.Input(shape=(128,)) outputs = layers.Dense(1)(inputs) return keras.Model(inputs, outputs) model1 = get_model() model2 = get_model() model3 = get_model() inputs = keras.Input(shape=(128,)) y1 = model1(inputs) y2 = model2(inputs) y3 = model3(inputs) outputs = layers.average([y1, y2, y3]) ensemble_model = keras.Model(inputs=inputs, outputs=outputs) 操作复杂图形 topologies 具有多个输入和输出的模型 功能 API 可轻松操作多个输入和输出。 火了。 Sequential 例如,如果您正在构建一个系统来按优先级排序客户发票,并将其路由到正确的部门,那么该模型将有三个输入: 门票的标题(文本输入), 机票的文本体(文本输入),以及 用户添加的任何标签(分类输入) 该模型将有两个输出: 优先级分为 0 至 1 (尺度 Sigmoid 输出),以及 the department that should handle the ticket (softmax output over the set of departments). 您可以使用功能 API 以几个行构建此模型: num_tags = 12 # Number of unique issue tags num_words = 10000 # Size of vocabulary obtained when preprocessing text data num_departments = 4 # Number of departments for predictions title_input = keras.Input( shape=(None,), name="title" ) # Variable-length sequence of ints body_input = keras.Input(shape=(None,), name="body") # Variable-length sequence of ints tags_input = keras.Input( shape=(num_tags,), name="tags" ) # Binary vectors of size `num_tags` # Embed each word in the title into a 64-dimensional vector title_features = layers.Embedding(num_words, 64)(title_input) # Embed each word in the text into a 64-dimensional vector body_features = layers.Embedding(num_words, 64)(body_input) # Reduce sequence of embedded words in the title into a single 128-dimensional vector title_features = layers.LSTM(128)(title_features) # Reduce sequence of embedded words in the body into a single 32-dimensional vector body_features = layers.LSTM(32)(body_features) # Merge all available features into a single large vector via concatenation x = layers.concatenate([title_features, body_features, tags_input]) # Stick a logistic regression for priority prediction on top of the features priority_pred = layers.Dense(1, name="priority")(x) # Stick a department classifier on top of the features department_pred = layers.Dense(num_departments, name="department")(x) # Instantiate an end-to-end model predicting both priority and department model = keras.Model( inputs=[title_input, body_input, tags_input], outputs=[priority_pred, department_pred], ) 现在就把模型整理一下: keras.utils.plot_model(model, "multi_input_and_output_model.png", show_shapes=True) 在编译此模型时,您可以为每个输出分配不同的损失,甚至可以为每个损失分配不同的重量 - 以调节其对总训练损失的贡献。 model.compile( optimizer=keras.optimizers.RMSprop(1e-3), loss=[ keras.losses.BinaryCrossentropy(from_logits=True), keras.losses.CategoricalCrossentropy(from_logits=True), ], loss_weights=[1.0, 0.2], ) 由于输出层有不同的名称,您还可以用相应的层名称指定损失和损失重量: model.compile( optimizer=keras.optimizers.RMSprop(1e-3), loss={ "priority": keras.losses.BinaryCrossentropy(from_logits=True), "department": keras.losses.CategoricalCrossentropy(from_logits=True), }, loss_weights={"priority": 1.0, "department": 0.2}, ) 通过传递入口和目标的NumPy数组列表来训练模型: # Dummy input data title_data = np.random.randint(num_words, size=(1280, 10)) body_data = np.random.randint(num_words, size=(1280, 100)) tags_data = np.random.randint(2, size=(1280, num_tags)).astype("float32") # Dummy target data priority_targets = np.random.random(size=(1280, 1)) dept_targets = np.random.randint(2, size=(1280, num_departments)) model.fit( {"title": title_data, "body": body_data, "tags": tags_data}, {"priority": priority_targets, "department": dept_targets}, epochs=2, batch_size=32, ) Epoch 1/2 40/40 [==============================] - 8s 112ms/step - loss: 1.2982 - priority_loss: 0.6991 - department_loss: 2.9958 Epoch 2/2 40/40 [==============================] - 3s 64ms/step - loss: 1.3110 - priority_loss: 0.6977 - department_loss: 3.0666 当你与一个合适的呼唤 对象,它应该产生一个类似列表的。 或一大堆字典,如 . Dataset ([title_data, body_data, tags_data], [priority_targets, dept_targets]) ({'title': title_data, 'body': body_data, 'tags': tags_data}, {'priority': priority_targets, 'department': dept_targets}) 对于更详细的解释,请参阅 指南。 培训与评估 玩具的模型 除了具有多个输入和输出的模型外,功能 API 还可以轻松地操纵非线性连接 topologies - 这些模型具有不连续连接的层,而这些层的连接。 火不能处理。 Sequential 一个常见的用例是剩余连接,让我们为 CIFAR10 构建一个玩具 ResNet 模型来证明这一点: inputs = keras.Input(shape=(32, 32, 3), name="img") x = layers.Conv2D(32, 3, activation="relu")(inputs) x = layers.Conv2D(64, 3, activation="relu")(x) block_1_output = layers.MaxPooling2D(3)(x) x = layers.Conv2D(64, 3, activation="relu", padding="same")(block_1_output) x = layers.Conv2D(64, 3, activation="relu", padding="same")(x) block_2_output = layers.add([x, block_1_output]) x = layers.Conv2D(64, 3, activation="relu", padding="same")(block_2_output) x = layers.Conv2D(64, 3, activation="relu", padding="same")(x) block_3_output = layers.add([x, block_2_output]) x = layers.Conv2D(64, 3, activation="relu")(block_3_output) x = layers.GlobalAveragePooling2D()(x) x = layers.Dense(256, activation="relu")(x) x = layers.Dropout(0.5)(x) outputs = layers.Dense(10)(x) model = keras.Model(inputs, outputs, name="toy_resnet") model.summary() Model: "toy_resnet" __________________________________________________________________________________________________ Layer (type) Output Shape Param # Connected to ================================================================================================== img (InputLayer) [(None, 32, 32, 3)] 0 [] conv2d_8 (Conv2D) (None, 30, 30, 32) 896 ['img[0][0]'] conv2d_9 (Conv2D) (None, 28, 28, 64) 18496 ['conv2d_8[0][0]'] max_pooling2d_2 (MaxPoolin (None, 9, 9, 64) 0 ['conv2d_9[0][0]'] g2D) conv2d_10 (Conv2D) (None, 9, 9, 64) 36928 ['max_pooling2d_2[0][0]'] conv2d_11 (Conv2D) (None, 9, 9, 64) 36928 ['conv2d_10[0][0]'] add (Add) (None, 9, 9, 64) 0 ['conv2d_11[0][0]', 'max_pooling2d_2[0][0]'] conv2d_12 (Conv2D) (None, 9, 9, 64) 36928 ['add[0][0]'] conv2d_13 (Conv2D) (None, 9, 9, 64) 36928 ['conv2d_12[0][0]'] add_1 (Add) (None, 9, 9, 64) 0 ['conv2d_13[0][0]', 'add[0][0]'] conv2d_14 (Conv2D) (None, 7, 7, 64) 36928 ['add_1[0][0]'] global_average_pooling2d ( (None, 64) 0 ['conv2d_14[0][0]'] GlobalAveragePooling2D) dense_6 (Dense) (None, 256) 16640 ['global_average_pooling2d[0][ 0]'] dropout (Dropout) (None, 256) 0 ['dense_6[0][0]'] dense_7 (Dense) (None, 10) 2570 ['dropout[0][0]'] ================================================================================================== Total params: 223242 (872.04 KB) Trainable params: 223242 (872.04 KB) Non-trainable params: 0 (0.00 Byte) __________________________________________________________________________________________________ 图为模型: keras.utils.plot_model(model, "mini_resnet.png", show_shapes=True) 现在列车模型: (x_train, y_train), (x_test, y_test) = keras.datasets.cifar10.load_data() x_train = x_train.astype("float32") / 255.0 x_test = x_test.astype("float32") / 255.0 y_train = keras.utils.to_categorical(y_train, 10) y_test = keras.utils.to_categorical(y_test, 10) model.compile( optimizer=keras.optimizers.RMSprop(1e-3), loss=keras.losses.CategoricalCrossentropy(from_logits=True), metrics=["acc"], ) # We restrict the data to the first 1000 samples so as to limit execution time # on Colab. Try to train on the entire dataset until convergence! model.fit(x_train[:1000], y_train[:1000], batch_size=64, epochs=1, validation_split=0.2) 13/13 [==============================] - 4s 39ms/step - loss: 2.3086 - acc: 0.0988 - val_loss: 2.3020 - val_acc: 0.0850 本文最初在 TensorFlow 网站上发表,此文在新标题下出现在这里,并在 CC BY 4.0 版下获得许可。 最初发表在The 本文在新标题下出现,并在 CC BY 4.0 许可证下发布。 腾讯 腾讯