diff --git a/src/model_modules/model_class.py b/src/model_modules/model_class.py
index d6dcea179bcfa8a6ec41518db34b186e30d908fc..0064c795e9bba162fafe3e9d5f60a17b95a4d57f 100644
--- a/src/model_modules/model_class.py
+++ b/src/model_modules/model_class.py
@@ -8,11 +8,13 @@ from abc import ABC
from typing import Any, Callable, Dict
import keras
+import logging
from src.model_modules.inception_model import InceptionModelBase
from src.model_modules.flatten import flatten_tail
from src.model_modules.advanced_paddings import PadUtils, Padding2D
+
class AbstractModelClass(ABC):
"""
@@ -32,6 +34,13 @@ class AbstractModelClass(ABC):
self.__loss = None
self.model_name = self.__class__.__name__
self.__custom_objects = {}
+ self.__allowed_compile_options = {'metrics': None,
+ 'loss_weights': None,
+ 'sample_weight_mode': None,
+ 'weighted_metrics': None,
+ 'target_tensors': None
+ }
+ self.__compile_options = self.__allowed_compile_options
def __getattr__(self, name: str) -> Any:
@@ -93,6 +102,26 @@ class AbstractModelClass(ABC):
def custom_objects(self, value) -> None:
self.__custom_objects = value
+ @property
+ def compile_options(self) -> Callable:
+ """
+ The compile options property allows the user to use all keras.compile() arguments which are not already covered
+ by __loss and optimizer
+ :return:
+ """
+ return self.__compile_options
+
+ @compile_options.setter
+ def compile_options(self, value: Dict) -> None:
+ if not isinstance(value, dict):
+ raise TypeError(f"`value' has to be a dictionary. But it is {type(value)}")
+ for new_k, new_v in value.items():
+ if new_k in self.__allowed_compile_options.keys():
+ self.__compile_options[new_k] = new_v
+ else:
+ logging.warning(
+ f"`{new_k}' is not a valid additional compile option. Will be ignored in keras.compile()")
+
def get_settings(self) -> Dict:
"""
Get all class attributes that are not protected in the AbstractModelClass as dictionary.
@@ -106,6 +135,21 @@ class AbstractModelClass(ABC):
def set_loss(self):
pass
+ def set_compile_options(self):
+ """
+ This method only has to be defined in child class, when additional compile options should be used ()
+ (other options than optimizer and loss)
+ Has to be set as dictionary: {'metrics': None,
+ 'loss_weights': None,
+ 'sample_weight_mode': None,
+ 'weighted_metrics': None,
+ 'target_tensors': None
+ }
+
+ :return:
+ """
+ pass
+
def set_custom_objects(self, **kwargs) -> None:
"""
Set custom objects that are not part of keras framework. These custom objects are needed if an already compiled
@@ -157,6 +201,7 @@ class MyLittleModel(AbstractModelClass):
# apply to model
self.set_model()
self.set_loss()
+ self.set_compile_options()
self.set_custom_objects(loss=self.loss)
def set_model(self):
@@ -196,6 +241,9 @@ class MyLittleModel(AbstractModelClass):
self.loss = keras.losses.mean_squared_error
+ def set_compile_options(self):
+ self.compile_options = {"metrics": ["mse", "mae"]}
+
class MyBranchedModel(AbstractModelClass):
@@ -315,6 +363,7 @@ class MyTowerModel(AbstractModelClass):
# apply to model
self.set_model()
self.set_loss()
+ self.set_compile_options()
self.set_custom_objects(loss=self.loss)
def set_model(self):
@@ -392,6 +441,9 @@ class MyTowerModel(AbstractModelClass):
self.loss = [keras.losses.mean_squared_error]
+ def set_compile_options(self):
+ self.compile_options = {"metrics": ["mse"]}
+
class MyPaperModel(AbstractModelClass):
diff --git a/src/run_modules/model_setup.py b/src/run_modules/model_setup.py
index c558b5fc76ff336dc6a792ec0239fa3b64eab466..de2d6a576662702128ae5e486a072904b3c3bf73 100644
--- a/src/run_modules/model_setup.py
+++ b/src/run_modules/model_setup.py
@@ -10,8 +10,8 @@ import tensorflow as tf
from src.model_modules.keras_extensions import HistoryAdvanced, CallbackHandler
# from src.model_modules.model_class import MyBranchedModel as MyModel
-# from src.model_modules.model_class import MyLittleModel as MyModel
-from src.model_modules.model_class import MyTowerModel as MyModel
+from src.model_modules.model_class import MyLittleModel as MyModel
+# from src.model_modules.model_class import MyTowerModel as MyModel
# from src.model_modules.model_class import MyPaperModel as MyModel
from src.run_modules.run_environment import RunEnvironment
@@ -62,7 +62,8 @@ class ModelSetup(RunEnvironment):
def compile_model(self):
optimizer = self.data_store.get("optimizer", self.scope)
loss = self.model.loss
- self.model.compile(optimizer=optimizer, loss=loss, metrics=["mse", "mae"])
+ compile_options = self.model.compile_options
+ self.model.compile(optimizer=optimizer, loss=loss, **compile_options)
self.data_store.set("model", self.model, self.scope)
def _set_callbacks(self):
diff --git a/test/test_model_modules/test_model_class.py b/test/test_model_modules/test_model_class.py
index cee031749b193b91bd1cf16c02acfb3050eaed61..147c92532465574b625907d13c814c5cfcbaeac9 100644
--- a/test/test_model_modules/test_model_class.py
+++ b/test/test_model_modules/test_model_class.py
@@ -40,6 +40,16 @@ class TestAbstractModelClass:
amc.loss = keras.losses.mean_absolute_error
assert amc.loss == keras.losses.mean_absolute_error
+ def test_compile_options_property(self, amc):
+ amc.compile_options = {"metrics": ["mse", "mae"]}
+ assert amc.compile_options == {'loss_weights': None, 'metrics': ['mse', 'mae'], 'sample_weight_mode': None,
+ 'target_tensors': None, 'weighted_metrics': None}
+
+ def test_compile_options_property_type_error(self, amc):
+ with pytest.raises(TypeError) as einfo:
+ amc.compile_options = 'hello world'
+ assert "`value' has to be a dictionary. But it is <class 'str'>" in str(einfo.value)
+
def test_getattr(self, amc):
amc.model = keras.Model()
assert hasattr(amc, "compile") is True