淺談keras的擴(kuò)展性：自定義keras

1. 自定義keras

keras是一種深度學(xué)習(xí)的API，能夠快速實現(xiàn)你的實驗。keras也集成了很多預(yù)訓(xùn)練的模型，可以實現(xiàn)很多常規(guī)的任務(wù)，如圖像分類。TensorFlow 2.0之后tensorflow本身也變的很keras化。

另一方面，keras表現(xiàn)出高度的模塊化和封裝性，所以有的人會覺得keras不易于擴(kuò)展， 比如實現(xiàn)一種新的Loss，新的網(wǎng)絡(luò)層結(jié)構(gòu)；其實可以通過keras的基礎(chǔ)模塊進(jìn)行快速的擴(kuò)展，實現(xiàn)更新的算法。

本文就keras的擴(kuò)展性，總結(jié)了對layer，model和loss的自定義。

2. 自定義keras layers

layers是keras中重要的組成部分，網(wǎng)絡(luò)結(jié)構(gòu)中每一個組成都要以layers來表現(xiàn)。keras提供了很多常規(guī)的layer，如Convolution layers，pooling layers， activation layers， dense layers等， 我們可以通過繼承基礎(chǔ)layers來擴(kuò)展自定義的layers。

2.1 base layer

layer實了輸入tensor和輸出tensor的操作類，以下為base layer的5個方法，自定義layer只要重寫這些方法就可以了。

• init(): 定義自定義layer的一些屬性
• build(self, input_shape)：定義layer需要的權(quán)重weights
• call(self, *args, **kwargs)：layer具體的操作，會在調(diào)用自定義layer自動執(zhí)行
• get_config(self)：layer初始化的配置，是一個字典dictionary。
• compute_output_shape(self,input_shape)：計算輸出tensor的shape

2.2 例子

#?標(biāo)準(zhǔn)化層
class?InstanceNormalize(Layer):
????def?__init__(self,?**kwargs):
????????super(InstanceNormalize,?self).__init__(**kwargs)
????????self.epsilon?=?1e-3
????????????

????def?call(self,?x,?mask=None):
????????mean,?var?=?tf.nn.moments(x,?[1,?2],?keep_dims=True)
????????return?tf.div(tf.subtract(x,?mean),?tf.sqrt(tf.add(var,?self.epsilon)))

?????????????????????????????????????????????????
????def?compute_output_shape(self,input_shape):
????????return?input_shape

#?調(diào)用
inputs?=?keras.Input(shape=(None,?None,?3))
x?=?InstanceNormalize()(inputs)

#?可以通過add_weight()?創(chuàng)建權(quán)重
class?SimpleDense(Layer):

??def?__init__(self,?units=32):
??????super(SimpleDense,?self).__init__()
??????self.units?=?units

??def?build(self,?input_shape):
??????self.w?=?self.add_weight(shape=(input_shape[-1],?self.units),
???????????????????????????????initializer='random_normal',
???????????????????????????????trainable=True)
??????self.b?=?self.add_weight(shape=(self.units,),
???????????????????????????????initializer='random_normal',
???????????????????????????????trainable=True)

??def?call(self,?inputs):
??????return?tf.matmul(inputs,?self.w)?+?self.b

#?調(diào)用
inputs?=?keras.Input(shape=(None,?None,?3))
x?=?SimpleDense(units=64)(inputs)

3. 自定義keras model

我們在定義完網(wǎng)絡(luò)結(jié)構(gòu)時，會把整個工作流放在keras.Model， 進(jìn)行compile(), 然后通過fit()進(jìn)行訓(xùn)練過程。執(zhí)行fit()的時候，執(zhí)行每個batch size data的時候，都會調(diào)用Model中train_step(self, data)

from?keras.models?import?Sequential
from?keras.layers?import?Dense,?Activation
model?=?Sequential()

model.add(Dense(units=64,?input_dim=100))
model.add(Activation("relu"))
model.add(Dense(units=10))
model.add(Activation("softmax"))

model.compile(loss='categorical_crossentropy',?optimizer='sgd',?metrics=['accuracy'])

model.fit(x_train,?y_train,?epochs=5,?batch_size=32)

當(dāng)你需要自己控制訓(xùn)練過程的時候，可以重寫Model的train_step(self, data)方法

class?CustomModel(keras.Model):
????def?train_step(self,?data):
????????#?Unpack?the?data.?Its?structure?depends?on?your?model?and
????????#?on?what?you?pass?to?`fit()`.
????????x,?y?=?data

????????with?tf.GradientTape()?as?tape:
????????????y_pred?=?self(x,?training=True)??#?Forward?pass
????????????#?Compute?the?loss?value
????????????#?(the?loss?function?is?configured?in?`compile()`)
????????????loss?=?self.compiled_loss(y,?y_pred,?regularization_losses=self.losses)

????????#?Compute?gradients
????????trainable_vars?=?self.trainable_variables
????????gradients?=?tape.gradient(loss,?trainable_vars)
????????#?Update?weights
????????self.optimizer.apply_gradients(zip(gradients,?trainable_vars))
????????#?Update?metrics?(includes?the?metric?that?tracks?the?loss)
????????self.compiled_metrics.update_state(y,?y_pred)
????????#?Return?a?dict?mapping?metric?names?to?current?value
????????return?{m.name:?m.result()?for?m?in?self.metrics}

import?numpy?as?np

#?Construct?and?compile?an?instance?of?CustomModel
inputs?=?keras.Input(shape=(32,))
outputs?=?keras.layers.Dense(1)(inputs)
model?=?CustomModel(inputs,?outputs)
model.compile(optimizer="adam",?loss="mse",?metrics=["mae"])

#?Just?use?`fit`?as?usual
x?=?np.random.random((1000,?32))
y?=?np.random.random((1000,?1))
model.fit(x,?y,?epochs=3)

4. 自定義keras loss

keras實現(xiàn)了交叉熵等常見的loss，自定義loss對于使用keras來說是比較常見，實現(xiàn)各種魔改loss，如focal loss。

我們來看看keras源碼中對loss實現(xiàn)

def?categorical_crossentropy(y_true,?y_pred):
????return?K.categorical_crossentropy(y_true,?y_pred)

def?mean_squared_error(y_true,?y_pred):
????return?K.mean(K.square(y_pred?-?y_true),?axis=-1)

可以看出輸入是groud true y_true和預(yù)測值y_pred， 返回為計算loss的函數(shù)。自定義loss可以參照如此模式即可。

def?focal_loss(weights=None,?alpha=0.25,?gamma=2):
????r"""Compute?focal?loss?for?predictions.
????????Multi-labels?Focal?loss?formula:
????????????FL?=?-alpha?*?(z-p)^gamma?*?log(p)?-(1-alpha)?*?p^gamma?*?log(1-p)
?????????????????,which?alpha?=?0.25,?gamma?=?2,?p?=?sigmoid(x),?z?=?target_tensor.
????#?https://github.com/ailias/Focal-Loss-implement-on-Tensorflow/blob/master/focal_loss.py
????Args:
?????prediction_tensor:?A?float?tensor?of?shape?[batch_size,?num_anchors,
????????num_classes]?representing?the?predicted?logits?for?each?class
?????target_tensor:?A?float?tensor?of?shape?[batch_size,?num_anchors,
????????num_classes]?representing?one-hot?encoded?classification?targets
?????weights:?A?float?tensor?of?shape?[batch_size,?num_anchors]
?????alpha:?A?scalar?tensor?for?focal?loss?alpha?hyper-parameter
?????gamma:?A?scalar?tensor?for?focal?loss?gamma?hyper-parameter
????Returns:
????????loss:?A?(scalar)?tensor?representing?the?value?of?the?loss?function
????"""

????def?_custom_loss(y_true,?y_pred):
????????sigmoid_p?=?tf.nn.sigmoid(y_pred)
????????zeros?=?array_ops.zeros_like(sigmoid_p,?dtype=sigmoid_p.dtype)

????????#?For?poitive?prediction,?only?need?consider?front?part?loss,?back?part?is?0;
????????#?target_tensor?>?zeros?<=>?z=1,?so?poitive?coefficient?=?z?-?p.
????????pos_p_sub?=?array_ops.where(y_true?>?zeros,?y_true?-?sigmoid_p,?zeros)

????????#?For?negative?prediction,?only?need?consider?back?part?loss,?front?part?is?0;
????????#?target_tensor?>?zeros?<=>?z=1,?so?negative?coefficient?=?0.
????????neg_p_sub?=?array_ops.where(y_true?>?zeros,?zeros,?sigmoid_p)
????????per_entry_cross_ent?=?-?alpha?*?(pos_p_sub?**?gamma)?*?tf.log(tf.clip_by_value(sigmoid_p,?1e-8,?1.0))?\
??????????????????????????????-?(1?-?alpha)?*?(neg_p_sub?**?gamma)?*?tf.log(
????????????tf.clip_by_value(1.0?-?sigmoid_p,?1e-8,?1.0))
????????return?tf.reduce_sum(per_entry_cross_ent)

????return?_custom_loss

5. 總結(jié)

本文分享了keras的擴(kuò)展功能，擴(kuò)展功能其實也是實現(xiàn)Keras模塊化的一種繼承實現(xiàn)。

總結(jié)如下：

• 繼承Layer實現(xiàn)自定義layer， 記住bulid() call()
• 繼續(xù)Model實現(xiàn)train_step定義訓(xùn)練過程，記住梯度計算tape.gradient(loss, trainable_vars) ，權(quán)重更新optimizer.apply_gradients, 計算evaluate compiled_metrics.update_state(y, y_pred)
• 魔改loss，記住groud true y_true和預(yù)測值y_pred輸入，返回loss function