From 5de64e132b3c6de853ae69a731e1c8be9338aacd Mon Sep 17 00:00:00 2001
From: lukas leufen <l.leufen@fz-juelich.de>
Date: Fri, 31 Jan 2020 12:05:20 +0100
Subject: [PATCH] created new module keras_extensions to collect all callback
extensions in one place
---
src/helpers.py | 58 +---------
src/model_modules/keras_extensions.py | 108 ++++++++++++++++++
src/model_modules/model_class.py | 6 +-
src/plotting/training_monitoring.py | 3 +-
src/run_modules/model_setup.py | 66 +----------
src/run_modules/training.py | 26 ++---
test/test_helpers.py | 40 +------
.../test_keras_extensions.py | 42 +++++++
test/test_modules/test_training.py | 3 +-
.../test_plotting/test_training_monitoring.py | 2 +-
10 files changed, 178 insertions(+), 176 deletions(-)
create mode 100644 src/model_modules/keras_extensions.py
create mode 100644 test/test_model_modules/test_keras_extensions.py
diff --git a/src/helpers.py b/src/helpers.py
index f119f140..a4ce625c 100644
--- a/src/helpers.py
+++ b/src/helpers.py
@@ -11,12 +11,10 @@ import time
import socket
import datetime as dt
-import keras
import keras.backend as K
-import numpy as np
import xarray as xr
-from typing import Union, Dict, Callable
+from typing import Dict, Callable
def to_list(arg):
@@ -45,60 +43,6 @@ def l_p_loss(power: int):
return loss
-class LearningRateDecay(keras.callbacks.History):
- """
- Decay learning rate during model training. Start with a base learning rate and lower this rate after every
- n(=epochs_drop) epochs by drop value (0, 1], drop value = 1 means no decay in learning rate.
- """
-
- def __init__(self, base_lr: float = 0.01, drop: float = 0.96, epochs_drop: int = 8):
- super().__init__()
- self.lr = {'lr': []}
- self.base_lr = self.check_param(base_lr, 'base_lr')
- self.drop = self.check_param(drop, 'drop')
- self.epochs_drop = self.check_param(epochs_drop, 'epochs_drop', upper=None)
- self.epoch = []
- self.history = {}
-
- @staticmethod
- def check_param(value: float, name: str, lower: Union[float, None] = 0, upper: Union[float, None] = 1):
- """
- Check if given value is in interval. The left (lower) endpoint is open, right (upper) endpoint is closed. To
- only one side of the interval, set the other endpoint to None. If both ends are set to None, just return the
- value without any check.
- :param value: value to check
- :param name: name of the variable to display in error message
- :param lower: left (lower) endpoint of interval, opened
- :param upper: right (upper) endpoint of interval, closed
- :return: unchanged value or raise ValueError
- """
- if lower is None:
- lower = -np.inf
- if upper is None:
- upper = np.inf
- if lower < value <= upper:
- return value
- else:
- raise ValueError(f"{name} is out of allowed range ({lower}, {upper}{')' if upper == np.inf else ']'}: "
- f"{name}={value}")
-
- def on_train_begin(self, logs=None):
- pass
-
- def on_epoch_begin(self, epoch: int, logs=None):
- """
- Lower learning rate every epochs_drop epochs by factor drop.
- :param epoch: current epoch
- :param logs: ?
- :return: update keras learning rate
- """
- current_lr = self.base_lr * math.pow(self.drop, math.floor(epoch / self.epochs_drop))
- K.set_value(self.model.optimizer.lr, current_lr)
- self.lr['lr'].append(current_lr)
- logging.info(f"Set learning rate to {current_lr}")
- return K.get_value(self.model.optimizer.lr)
-
-
class TimeTracking(object):
"""
Track time to measure execution time. Time tracking automatically starts on initialisation and ends by calling stop
diff --git a/src/model_modules/keras_extensions.py b/src/model_modules/keras_extensions.py
new file mode 100644
index 00000000..f84f33d1
--- /dev/null
+++ b/src/model_modules/keras_extensions.py
@@ -0,0 +1,108 @@
+__author__ = 'Lukas Leufen, Felix Kleinert'
+__date__ = '2020-01-31'
+
+import logging
+import math
+import pickle
+from typing import Union
+
+import numpy as np
+from keras import backend as K
+from keras.callbacks import History, ModelCheckpoint
+
+
+class HistoryAdvanced(History):
+
+ def __init__(self, old_epoch=None, old_history=None):
+ self.epoch = old_epoch or []
+ self.history = old_history or {}
+ super().__init__()
+
+ def on_train_begin(self, logs=None):
+ pass
+
+
+class LearningRateDecay(History):
+ """
+ Decay learning rate during model training. Start with a base learning rate and lower this rate after every
+ n(=epochs_drop) epochs by drop value (0, 1], drop value = 1 means no decay in learning rate.
+ """
+
+ def __init__(self, base_lr: float = 0.01, drop: float = 0.96, epochs_drop: int = 8):
+ super().__init__()
+ self.lr = {'lr': []}
+ self.base_lr = self.check_param(base_lr, 'base_lr')
+ self.drop = self.check_param(drop, 'drop')
+ self.epochs_drop = self.check_param(epochs_drop, 'epochs_drop', upper=None)
+ self.epoch = []
+ self.history = {}
+
+ @staticmethod
+ def check_param(value: float, name: str, lower: Union[float, None] = 0, upper: Union[float, None] = 1):
+ """
+ Check if given value is in interval. The left (lower) endpoint is open, right (upper) endpoint is closed. To
+ only one side of the interval, set the other endpoint to None. If both ends are set to None, just return the
+ value without any check.
+ :param value: value to check
+ :param name: name of the variable to display in error message
+ :param lower: left (lower) endpoint of interval, opened
+ :param upper: right (upper) endpoint of interval, closed
+ :return: unchanged value or raise ValueError
+ """
+ if lower is None:
+ lower = -np.inf
+ if upper is None:
+ upper = np.inf
+ if lower < value <= upper:
+ return value
+ else:
+ raise ValueError(f"{name} is out of allowed range ({lower}, {upper}{')' if upper == np.inf else ']'}: "
+ f"{name}={value}")
+
+ def on_train_begin(self, logs=None):
+ pass
+
+ def on_epoch_begin(self, epoch: int, logs=None):
+ """
+ Lower learning rate every epochs_drop epochs by factor drop.
+ :param epoch: current epoch
+ :param logs: ?
+ :return: update keras learning rate
+ """
+ current_lr = self.base_lr * math.pow(self.drop, math.floor(epoch / self.epochs_drop))
+ K.set_value(self.model.optimizer.lr, current_lr)
+ self.lr['lr'].append(current_lr)
+ logging.info(f"Set learning rate to {current_lr}")
+ return K.get_value(self.model.optimizer.lr)
+
+
+class ModelCheckpointAdvanced(ModelCheckpoint):
+ """
+ IMPORTANT: Always add the model checkpoint advanced as last callback to properly update all tracked callbacks, e.g.
+ fit_generator(callbacks=[..., <last_here>])
+ """
+ def __init__(self, *args, **kwargs):
+ self.callbacks = kwargs.pop("callbacks")
+ super().__init__(*args, **kwargs)
+
+ def update_best(self, hist):
+ self.best = hist.history.get(self.monitor)[-1]
+
+ def update_callbacks(self, callbacks):
+ self.callbacks = callbacks
+
+ def on_epoch_end(self, epoch, logs=None):
+ super().on_epoch_end(epoch, logs)
+
+ for callback in self.callbacks:
+ file_path = callback["path"]
+ if self.epochs_since_last_save == 0 and epoch != 0:
+ if self.save_best_only:
+ current = logs.get(self.monitor)
+ if current == self.best:
+ print('\nEpoch %05d: save to %s' % (epoch + 1, file_path))
+ with open(file_path, "wb") as f:
+ pickle.dump(callback["callback"], f)
+ else:
+ with open(file_path, "wb") as f:
+ pickle.dump(callback["callback"], f)
\ No newline at end of file
diff --git a/src/model_modules/model_class.py b/src/model_modules/model_class.py
index 5e9931d7..02f43dc1 100644
--- a/src/model_modules/model_class.py
+++ b/src/model_modules/model_class.py
@@ -1,3 +1,5 @@
+import src.model_modules.keras_extensions
+
__author__ = "Lukas Leufen"
__date__ = '2019-12-12'
@@ -109,7 +111,7 @@ class MyLittleModel(AbstractModelClass):
self.regularizer = keras.regularizers.l2(0.1)
self.initial_lr = 1e-2
self.optimizer = keras.optimizers.SGD(lr=self.initial_lr, momentum=0.9)
- self.lr_decay = helpers.LearningRateDecay(base_lr=self.initial_lr, drop=.94, epochs_drop=10)
+ self.lr_decay = src.model_modules.keras_extensions.LearningRateDecay(base_lr=self.initial_lr, drop=.94, epochs_drop=10)
self.epochs = 20
self.batch_size = int(256)
self.activation = keras.layers.PReLU
@@ -189,7 +191,7 @@ class MyBranchedModel(AbstractModelClass):
self.regularizer = keras.regularizers.l2(0.1)
self.initial_lr = 1e-2
self.optimizer = keras.optimizers.SGD(lr=self.initial_lr, momentum=0.9)
- self.lr_decay = helpers.LearningRateDecay(base_lr=self.initial_lr, drop=.94, epochs_drop=10)
+ self.lr_decay = src.model_modules.keras_extensions.LearningRateDecay(base_lr=self.initial_lr, drop=.94, epochs_drop=10)
self.epochs = 20
self.batch_size = int(256)
self.activation = keras.layers.PReLU
diff --git a/src/plotting/training_monitoring.py b/src/plotting/training_monitoring.py
index 87e4071a..339ba637 100644
--- a/src/plotting/training_monitoring.py
+++ b/src/plotting/training_monitoring.py
@@ -9,8 +9,7 @@ import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
-from src.helpers import LearningRateDecay
-
+from src.model_modules.keras_extensions import LearningRateDecay
matplotlib.use('Agg')
history_object = Union[Dict, keras.callbacks.History]
diff --git a/src/run_modules/model_setup.py b/src/run_modules/model_setup.py
index a4d89f65..2512947b 100644
--- a/src/run_modules/model_setup.py
+++ b/src/run_modules/model_setup.py
@@ -4,16 +4,13 @@ __date__ = '2019-12-02'
import keras
from keras import losses
-from keras.callbacks import ModelCheckpoint, History
-from keras.regularizers import l2
-from keras.optimizers import SGD
import tensorflow as tf
import logging
import os
-import pickle
from src.run_modules.run_environment import RunEnvironment
-from src.helpers import l_p_loss, LearningRateDecay
+from src.helpers import l_p_loss
+from src.model_modules.keras_extensions import HistoryAdvanced, ModelCheckpointAdvanced
from src.model_modules.inception_model import InceptionModelBase
from src.model_modules.flatten import flatten_tail
# from src.model_modules.model_class import MyBranchedModel as MyModel
@@ -75,8 +72,8 @@ class ModelSetup(RunEnvironment):
self.data_store.set("hist", hist, scope="general.model")
callbacks = [{"callback": lr, "path": self.callbacks_name % "lr"},
{"callback": hist, "path": self.callbacks_name % "hist"}]
- checkpoint = ModelCheckpointAdvanced2(filepath=self.checkpoint_name, verbose=1, monitor='val_loss',
- save_best_only=True, mode='auto', callbacks=callbacks)
+ checkpoint = ModelCheckpointAdvanced(filepath=self.checkpoint_name, verbose=1, monitor='val_loss',
+ save_best_only=True, mode='auto', callbacks=callbacks)
self.data_store.set("checkpoint", checkpoint, self.scope)
def load_weights(self):
@@ -104,61 +101,6 @@ class ModelSetup(RunEnvironment):
keras.utils.plot_model(self.model, to_file=file_name, show_shapes=True, show_layer_names=True)
-class HistoryAdvanced(History):
-
- def __init__(self, old_epoch=None, old_history=None):
- self.epoch = old_epoch or []
- self.history = old_history or {}
- super().__init__()
-
- def on_train_begin(self, logs=None):
- pass
-
-
-class ModelCheckpointAdvanced(ModelCheckpoint):
-
- def __init__(self, *args, **kwargs):
- self.callbacks_to_save = kwargs.pop("callbacks_to_save")
- self.callbacks_filepath = kwargs.pop("callbacks_filepath")
- super().__init__(*args, **kwargs)
-
- def on_epoch_end(self, epoch, logs=None):
- super().on_epoch_end(epoch, logs)
-
- file_path = self.callbacks_filepath
- if self.epochs_since_last_save == 0 and epoch != 0:
- if self.save_best_only:
- current = logs.get(self.monitor)
- if current == self.best:
- with open(file_path, "wb") as f:
- pickle.dump(self.callbacks_to_save, f)
- else:
- with open(file_path, "wb") as f:
- pickle.dump(self.callbacks_to_save, f)
-
-
-class ModelCheckpointAdvanced2(ModelCheckpoint):
-
- def __init__(self, *args, **kwargs):
- self.callbacks = kwargs.pop("callbacks")
- super().__init__(*args, **kwargs)
-
- def on_epoch_end(self, epoch, logs=None):
- super().on_epoch_end(epoch, logs)
-
- for callback in self.callbacks:
- file_path = callback["path"]
- if self.epochs_since_last_save == 0 and epoch != 0:
- if self.save_best_only:
- current = logs.get(self.monitor)
- if current == self.best:
- with open(file_path, "wb") as f:
- pickle.dump(callback["callback"], f)
- else:
- with open(file_path, "wb") as f:
- pickle.dump(callback["callback"], f)
-
-
def my_loss():
loss = l_p_loss(4)
keras_loss = losses.mean_squared_error
diff --git a/src/run_modules/training.py b/src/run_modules/training.py
index e9c7487b..7dfa06ff 100644
--- a/src/run_modules/training.py
+++ b/src/run_modules/training.py
@@ -10,8 +10,7 @@ import pickle
from src.run_modules.run_environment import RunEnvironment
from src.data_handling.data_distributor import Distributor
from src.plotting.training_monitoring import PlotModelHistory, PlotModelLearningRate
-from src.helpers import LearningRateDecay
-from src.run_modules.model_setup import ModelCheckpointAdvanced2
+from src.model_modules.keras_extensions import LearningRateDecay, ModelCheckpointAdvanced
class Training(RunEnvironment):
@@ -24,7 +23,7 @@ class Training(RunEnvironment):
self.test_set = None
self.batch_size = self.data_store.get("batch_size", "general.model")
self.epochs = self.data_store.get("epochs", "general.model")
- self.checkpoint: ModelCheckpointAdvanced2 = self.data_store.get("checkpoint", "general.model")
+ self.checkpoint: ModelCheckpointAdvanced = self.data_store.get("checkpoint", "general.model")
# self.callbacks = self.data_store.get("callbacks", "general.model")
self.lr_sc = self.data_store.get("lr_decay", "general.model")
self.hist = self.data_store.get("hist", "general.model")
@@ -88,7 +87,7 @@ class Training(RunEnvironment):
validation_data=self.val_set.distribute_on_batches(),
validation_steps=len(self.val_set),
# callbacks=self.callbacks)
- callbacks=[self.checkpoint, self.lr_sc, self.hist])
+ callbacks=[self.lr_sc, self.hist, self.checkpoint])
else:
lr_filepath = self.checkpoint.callbacks[0]["path"] # TODO: stopped here. why does training start 1 epoch too early or doesn't it?
hist_filepath = self.checkpoint.callbacks[1]["path"]
@@ -100,15 +99,16 @@ class Training(RunEnvironment):
initial_epoch = max(hist_callbacks.epoch) + 1
callbacks = [{"callback": self.lr_sc, "path": lr_filepath},
{"callback": self.hist, "path": hist_filepath}]
- self.checkpoint.callbacks = callbacks
- history = self.model.fit_generator(generator=self.train_set.distribute_on_batches(),
- steps_per_epoch=len(self.train_set),
- epochs=self.epochs,
- verbose=2,
- validation_data=self.val_set.distribute_on_batches(),
- validation_steps=len(self.val_set),
- callbacks=[self.checkpoint, self.lr_sc, self.hist],
- initial_epoch=initial_epoch)
+ self.checkpoint.update_callbacks(callbacks)
+ self.checkpoint.update_best(hist_callbacks)
+ self.hist = self.model.fit_generator(generator=self.train_set.distribute_on_batches(),
+ steps_per_epoch=len(self.train_set),
+ epochs=self.epochs,
+ verbose=2,
+ validation_data=self.val_set.distribute_on_batches(),
+ validation_steps=len(self.val_set),
+ callbacks=[self.lr_sc, self.hist, self.checkpoint],
+ initial_epoch=initial_epoch)
history = self.hist
self.save_callbacks(history)
self.load_best_model(self.checkpoint.filepath)
diff --git a/test/test_helpers.py b/test/test_helpers.py
index e98a46fa..c909960b 100644
--- a/test/test_helpers.py
+++ b/test/test_helpers.py
@@ -7,6 +7,8 @@ import numpy as np
import mock
import platform
+from src.model_modules.keras_extensions import LearningRateDecay
+
class TestToList:
@@ -44,44 +46,6 @@ class TestLoss:
assert hist.history['loss'][0] == 2.25
-class TestLearningRateDecay:
-
- def test_init(self):
- lr_decay = LearningRateDecay()
- assert lr_decay.lr == {'lr': []}
- assert lr_decay.base_lr == 0.01
- assert lr_decay.drop == 0.96
- assert lr_decay.epochs_drop == 8
-
- def test_check_param(self):
- lr_decay = object.__new__(LearningRateDecay)
- assert lr_decay.check_param(1, "tester") == 1
- assert lr_decay.check_param(0.5, "tester") == 0.5
- with pytest.raises(ValueError) as e:
- lr_decay.check_param(0, "tester")
- assert "tester is out of allowed range (0, 1]: tester=0" in e.value.args[0]
- with pytest.raises(ValueError) as e:
- lr_decay.check_param(1.5, "tester")
- assert "tester is out of allowed range (0, 1]: tester=1.5" in e.value.args[0]
- assert lr_decay.check_param(1.5, "tester", upper=None) == 1.5
- with pytest.raises(ValueError) as e:
- lr_decay.check_param(0, "tester", upper=None)
- assert "tester is out of allowed range (0, inf): tester=0" in e.value.args[0]
- assert lr_decay.check_param(0.5, "tester", lower=None) == 0.5
- with pytest.raises(ValueError) as e:
- lr_decay.check_param(0.5, "tester", lower=None, upper=0.2)
- assert "tester is out of allowed range (-inf, 0.2]: tester=0.5" in e.value.args[0]
- assert lr_decay.check_param(10, "tester", upper=None, lower=None)
-
- def test_on_epoch_begin(self):
- lr_decay = LearningRateDecay(base_lr=0.02, drop=0.95, epochs_drop=2)
- model = keras.Sequential()
- model.add(keras.layers.Dense(1, input_dim=1))
- model.compile(optimizer=keras.optimizers.Adam(), loss=l_p_loss(2))
- model.fit(np.array([1, 0, 2, 0.5]), np.array([1, 1, 0, 0.5]), epochs=5, callbacks=[lr_decay])
- assert lr_decay.lr['lr'] == [0.02, 0.02, 0.02 * 0.95, 0.02 * 0.95, 0.02 * 0.95 * 0.95]
-
-
class TestTimeTracking:
def test_init(self):
diff --git a/test/test_model_modules/test_keras_extensions.py b/test/test_model_modules/test_keras_extensions.py
new file mode 100644
index 00000000..7bf5cf51
--- /dev/null
+++ b/test/test_model_modules/test_keras_extensions.py
@@ -0,0 +1,42 @@
+import pytest
+from src.model_modules.keras_extensions import *
+import keras
+import numpy as np
+
+
+class TestLearningRateDecay:
+
+ def test_init(self):
+ lr_decay = LearningRateDecay()
+ assert lr_decay.lr == {'lr': []}
+ assert lr_decay.base_lr == 0.01
+ assert lr_decay.drop == 0.96
+ assert lr_decay.epochs_drop == 8
+
+ def test_check_param(self):
+ lr_decay = object.__new__(LearningRateDecay)
+ assert lr_decay.check_param(1, "tester") == 1
+ assert lr_decay.check_param(0.5, "tester") == 0.5
+ with pytest.raises(ValueError) as e:
+ lr_decay.check_param(0, "tester")
+ assert "tester is out of allowed range (0, 1]: tester=0" in e.value.args[0]
+ with pytest.raises(ValueError) as e:
+ lr_decay.check_param(1.5, "tester")
+ assert "tester is out of allowed range (0, 1]: tester=1.5" in e.value.args[0]
+ assert lr_decay.check_param(1.5, "tester", upper=None) == 1.5
+ with pytest.raises(ValueError) as e:
+ lr_decay.check_param(0, "tester", upper=None)
+ assert "tester is out of allowed range (0, inf): tester=0" in e.value.args[0]
+ assert lr_decay.check_param(0.5, "tester", lower=None) == 0.5
+ with pytest.raises(ValueError) as e:
+ lr_decay.check_param(0.5, "tester", lower=None, upper=0.2)
+ assert "tester is out of allowed range (-inf, 0.2]: tester=0.5" in e.value.args[0]
+ assert lr_decay.check_param(10, "tester", upper=None, lower=None)
+
+ def test_on_epoch_begin(self):
+ lr_decay = LearningRateDecay(base_lr=0.02, drop=0.95, epochs_drop=2)
+ model = keras.Sequential()
+ model.add(keras.layers.Dense(1, input_dim=1))
+ model.compile(optimizer=keras.optimizers.Adam(), loss=l_p_loss(2))
+ model.fit(np.array([1, 0, 2, 0.5]), np.array([1, 1, 0, 0.5]), epochs=5, callbacks=[lr_decay])
+ assert lr_decay.lr['lr'] == [0.02, 0.02, 0.02 * 0.95, 0.02 * 0.95, 0.02 * 0.95 * 0.95]
diff --git a/test/test_modules/test_training.py b/test/test_modules/test_training.py
index accb32e5..4631fe5a 100644
--- a/test/test_modules/test_training.py
+++ b/test/test_modules/test_training.py
@@ -14,7 +14,8 @@ from src.run_modules.training import Training
from src.run_modules.run_environment import RunEnvironment
from src.data_handling.data_distributor import Distributor
from src.data_handling.data_generator import DataGenerator
-from src.helpers import LearningRateDecay, PyTestRegex
+from src.helpers import PyTestRegex
+from src.model_modules.keras_extensions import LearningRateDecay
def my_test_model(activation, window_history_size, channels, dropout_rate, add_minor_branch=False):
diff --git a/test/test_plotting/test_training_monitoring.py b/test/test_plotting/test_training_monitoring.py
index e2d17057..358a19ad 100644
--- a/test/test_plotting/test_training_monitoring.py
+++ b/test/test_plotting/test_training_monitoring.py
@@ -3,7 +3,7 @@ import pytest
import os
from src.plotting.training_monitoring import PlotModelLearningRate, PlotModelHistory
-from src.helpers import LearningRateDecay
+from src.model_modules.keras_extensions import LearningRateDecay
@pytest.fixture
--
GitLab