From ff32ac9136aedd5f1ec65308848e8d09f4217b5b Mon Sep 17 00:00:00 2001
From: lukas leufen <l.leufen@fz-juelich.de>
Date: Fri, 19 Jun 2020 16:12:28 +0200
Subject: [PATCH] batch size is now part of the exp setup and not related to
the model class anymore
---
src/model_modules/model_class.py | 5 -----
src/run.py | 3 ++-
src/run_modules/experiment_setup.py | 5 +++--
src/run_modules/model_setup.py | 3 ++-
src/run_modules/training.py | 4 ++--
5 files changed, 9 insertions(+), 11 deletions(-)
diff --git a/src/model_modules/model_class.py b/src/model_modules/model_class.py
index ced01e9a..8a065fa5 100644
--- a/src/model_modules/model_class.py
+++ b/src/model_modules/model_class.py
@@ -119,7 +119,6 @@ from typing import Any, Callable, Dict
import keras
import tensorflow as tf
-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
@@ -356,7 +355,6 @@ class MyLittleModel(AbstractModelClass):
self.dropout_rate = 0.1
self.regularizer = keras.regularizers.l2(0.1)
self.epochs = 20
- self.batch_size = int(256)
self.activation = keras.layers.PReLU
# apply to model
@@ -430,7 +428,6 @@ class MyBranchedModel(AbstractModelClass):
self.dropout_rate = 0.1
self.regularizer = keras.regularizers.l2(0.1)
self.epochs = 20
- self.batch_size = int(256)
self.activation = keras.layers.PReLU
# apply to model
@@ -505,7 +502,6 @@ class MyTowerModel(AbstractModelClass):
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 * 4)
self.activation = keras.layers.PReLU
# apply to model
@@ -619,7 +615,6 @@ class MyPaperModel(AbstractModelClass):
self.lr_decay = src.model_modules.keras_extensions.LearningRateDecay(base_lr=self.initial_lr, drop=.94,
epochs_drop=10)
self.epochs = 150
- self.batch_size = int(256 * 2)
self.activation = keras.layers.ELU
self.padding = "SymPad2D"
diff --git a/src/run.py b/src/run.py
index ac1d27b8..eda0373c 100644
--- a/src/run.py
+++ b/src/run.py
@@ -25,7 +25,8 @@ def run(stations=['DEBW107', 'DEBY081', 'DEBW013', 'DEBW076', 'DEBW087', 'DEBW00
train_min_length=None, val_min_length=None, test_min_length=None,
evaluate_bootstraps=True, number_of_bootstraps=None, create_new_bootstraps=False,
plot_list=None,
- model=None):
+ model=None,
+ batch_size=None):
params = inspect.getfullargspec(ExperimentSetup).args
kwargs = {k: v for k, v in locals().items() if k in params}
diff --git a/src/run_modules/experiment_setup.py b/src/run_modules/experiment_setup.py
index 110e7791..5443e265 100644
--- a/src/run_modules/experiment_setup.py
+++ b/src/run_modules/experiment_setup.py
@@ -233,12 +233,12 @@ class ExperimentSetup(RunEnvironment):
create_new_model: bool = None, bootstrap_path=None, permute_data_on_training: bool = None, transformation=None,
train_min_length=None, val_min_length=None, test_min_length=None, extreme_values: list = None,
extremes_on_right_tail_only: bool = None, evaluate_bootstraps=True, plot_list=None, number_of_bootstraps=None,
- create_new_bootstraps=None, data_path: str = None, login_nodes=None, hpc_hosts=None, model=None):
+ create_new_bootstraps=None, data_path: str = None, login_nodes=None, hpc_hosts=None, model=None, batch_size=None):
# create run framework
super().__init__()
- # experiment setup
+ # experiment setup, hyperparameters
self._set_param("data_path", path_config.prepare_host(data_path=data_path, sampling=sampling))
self._set_param("hostname", path_config.get_host())
self._set_param("hpc_hosts", hpc_hosts, default=DEFAULT_HPC_HOST_LIST + DEFAULT_HPC_LOGIN_LIST)
@@ -257,6 +257,7 @@ class ExperimentSetup(RunEnvironment):
upsampling = self.data_store.get("upsampling", "train")
permute_data = False if permute_data_on_training is None else permute_data_on_training
self._set_param("permute_data", permute_data or upsampling, scope="train")
+ self._set_param("batch_size", batch_size, default=int(256 * 2))
# set experiment name
exp_date = self._get_parser_args(parser_args).get("experiment_date")
diff --git a/src/run_modules/model_setup.py b/src/run_modules/model_setup.py
index 13a13bb7..f9683b95 100644
--- a/src/run_modules/model_setup.py
+++ b/src/run_modules/model_setup.py
@@ -33,6 +33,7 @@ class ModelSetup(RunEnvironment):
* `generator` [train]
* `window_lead_time` [.]
* `window_history_size` [.]
+ * `model_class` [.]
Optional objects
* `lr_decay` [model]
@@ -43,7 +44,7 @@ class ModelSetup(RunEnvironment):
* `hist` [model]
* `callbacks` [model]
* `model_name` [model]
- * all settings from model class like `dropout_rate`, `initial_lr`, `batch_size`, and `optimizer` [model]
+ * all settings from model class like `dropout_rate`, `initial_lr`, and `optimizer` [model]
Creates
* plot of model architecture `<model_name>.pdf`
diff --git a/src/run_modules/training.py b/src/run_modules/training.py
index 8624b515..74e9b799 100644
--- a/src/run_modules/training.py
+++ b/src/run_modules/training.py
@@ -33,7 +33,7 @@ class Training(RunEnvironment):
Required objects [scope] from data store:
* `model` [model]
- * `batch_size` [model]
+ * `batch_size` [.]
* `epochs` [model]
* `callbacks` [model]
* `model_name` [model]
@@ -67,7 +67,7 @@ class Training(RunEnvironment):
self.train_set: Union[Distributor, None] = None
self.val_set: Union[Distributor, None] = None
self.test_set: Union[Distributor, None] = None
- self.batch_size = self.data_store.get("batch_size", "model")
+ self.batch_size = self.data_store.get("batch_size")
self.epochs = self.data_store.get("epochs", "model")
self.callbacks: CallbackHandler = self.data_store.get("callbacks", "model")
self.experiment_name = self.data_store.get("experiment_name")
--
GitLab