From 3b6be777b2eb3ff8f3ba9ad0e5853ca362338db5 Mon Sep 17 00:00:00 2001
From: Dennis Noll <dennis.noll@rwth-aachen.de>
Date: Fri, 31 Jul 2020 14:44:32 +0200
Subject: [PATCH] [keras] custom layers: now uses build function everywhere
---
keras.py | 51 +++++++++++++++++++++++++++++----------------------
1 file changed, 29 insertions(+), 22 deletions(-)
diff --git a/keras.py b/keras.py
index 16c94fe..8b9fcda 100644
--- a/keras.py
+++ b/keras.py
@@ -579,7 +579,7 @@ class CheckpointModel(tf.keras.callbacks.Callback):
return f"{self.savedir}/{self.identifier}-{self.get_index(epoch)}"
def on_epoch_end(self, epoch, logs=None):
- if epoch != 0 and epoch % self.frequency == 0:
+ if epoch % self.frequency == 0:
self.model.save(self.checkpoint_dir(epoch))
@@ -906,21 +906,22 @@ class DenseLayer(tf.keras.layers.Layer):
self.l2 = l2
self.batch_norm = batch_norm
+ def build(self, input_shape):
parts = []
- l2 = tf.keras.regularizers.l2(l2 if l2 else 0.0)
- weights = tf.keras.layers.Dense(nodes, kernel_regularizer=l2)
+ l2 = tf.keras.regularizers.l2(self.l2)
+ weights = tf.keras.layers.Dense(self.nodes, kernel_regularizer=l2)
parts.append(weights)
- if batch_norm:
+ if self.batch_norm:
dropout = 0.0
bn = tf.keras.layers.BatchNormalization()
parts.append(bn)
- act = tf.keras.layers.Activation(activation)
+ act = tf.keras.layers.Activation(self.activation)
parts.append(act)
- if activation == "selu":
+ if self.activation == "selu":
dropout = tf.keras.layers.AlphaDropout(dropout)
else:
dropout = tf.keras.layers.Dropout(dropout)
@@ -987,29 +988,32 @@ class FullyConnected(tf.keras.layers.Layer):
Parameters
----------
- number_layers : int
+ layers : int
The number of layers.
kwargs :
Arguments for DenseLayer.
"""
- def __init__(self, number_layers=0, **kwargs):
+ def __init__(self, layers=0, **kwargs):
super().__init__(name="FullyConnected")
- self.number_layers = number_layers
- layers = []
- for layer in range(self.number_layers):
- layers.append(DenseLayer(**kwargs))
self.layers = layers
+ self.kwargs = kwargs
+
+ def build(self, input_shape):
+ _layers = []
+ for layer in range(self.layers):
+ _layers.append(DenseLayer(**self.kwargs))
+ self._layers = _layers
def call(self, input_tensor, training=False):
x = input_tensor
- for layer in self.layers:
+ for layer in self._layers:
x = layer(x, training=training)
return x
def get_config(self):
- return {"number_layers": self.number_layers}
+ return {"layers": self.layers}
class ResNet(tf.keras.layers.Layer):
@@ -1018,29 +1022,32 @@ class ResNet(tf.keras.layers.Layer):
Parameters
----------
- number_layers : int
+ layers : int
The number of residual blocks.
kwargs :
Arguments for ResNetBlock.
"""
- def __init__(self, number_layers=1, **kwargs):
+ def __init__(self, layers=1, **kwargs):
super().__init__(name="ResNet")
- self.number_layers = number_layers
- layers = []
- for i in range(self.number_layers):
- layers.append(ResNetBlock(**kwargs))
self.layers = layers
+ self.kwargs = kwargs
+
+ def build(self, input_shape):
+ _layers = []
+ for i in range(self.layers):
+ _layers.append(ResNetBlock(**self.kwargs))
+ self._layers = _layers
def call(self, input_tensor, training=False):
x = input_tensor
- for layer in self.layers:
+ for layer in self._layers:
x = layer(x, training=training)
return x
def get_config(self):
- return {"number_layers": self.number_layers}
+ return {"layers": self.layers}
class RemoveLayer(tf.keras.layers.Layer):
--
GitLab