pyfemtet 0.4.20__py3-none-any.whl → 0.4.23__py3-none-any.whl

pyfemtet/opt/_femopt_core.py

@@ -13,6 +13,7 @@ import ctypes
 import numpy as np
 import pandas as pd
 from scipy.stats.qmc import LatinHypercube
+import optuna
 from optuna._hypervolume import WFG
 from dask.distributed import Lock, get_client
 
@@ -336,6 +337,25 @@ class Constraint(Function):
         super().__init__(fun, name, args, kwargs)
 
 
+# from pyfemtet.opt.opt import AbstractOptimizer
+class ParameterConstraint(Constraint):
+    """Variable のみを引数とする constraint 関数"""
+
+    def __init__(self, fun, name, lb, ub, opt: 'AbstractOptimizer'):
+        super().__init__(fun, name, lb, ub, strict=True, args=None, kwargs=None)
+        self.opt = opt
+        self.prm_args = [arg.name for arg in inspect.signature(self.fun).parameters.values()]
+
+    def calc(self, _):
+        # variables 全体から fun に対する引数を作成する
+        kwargs: dict = self.opt.get_parameter('dict')
+        keys = list(kwargs.keys())
+        for k in keys:
+            if k not in self.prm_args:
+                kwargs.pop(k)
+        return float(self.fun(**kwargs))
+
+
 class _HistoryDfCore:
     """Class for managing a DataFrame object in a distributed manner."""
 
@@ -369,7 +389,6 @@ class History:
         prm_names (str): User defined names of parameters.
         obj_names (str): User defined names of objectives.
         cns_names (str): User defined names of constraints.
-        local_data (pd.DataFrame): Local copy (on memory) of optimization history. 
         is_restart (bool): If the optimization process is a continuation of another process or not.
         is_processing (bool): The optimization is running or not.
 
@@ -382,8 +401,9 @@ class History:
     cns_names = []
     is_restart = False
     is_processing = False
-    _future = None
-    _actor_data = None
+    _df = None  # in case without client
+    _future = None  # in case with client
+    _actor_data = None  # in case with client
 
     def __init__(
             self,
@@ -402,9 +422,6 @@ class History:
         self.cns_names = cns_names
         self.additional_metadata = additional_metadata or ''
 
-        # 初期化
-        self.local_data = pd.DataFrame()
-
         # 最適化実行中かどうか
         self.is_processing = client is not None
 
@@ -425,12 +442,11 @@ class History:
             # そうでなければ df を初期化
             else:
                 columns, metadata = self.create_df_columns()
+                df = pd.DataFrame()
                 for c in columns:
-                    self.local_data[c] = None
+                    df[c] = None
                 self.metadata = metadata
-
-            # actor_data の初期化
-            self.actor_data = self.local_data
+                self.set_df(df)
 
             # 一時ファイルに書き込みを試み、UnicodeEncodeError が出ないかチェック
             import tempfile
@@ -446,14 +462,14 @@ class History:
             if not os.path.isfile(self.path):
                 raise FileNotFoundError(self.path)
 
-            # csv の local_data へと、names への読み込み
+            # csv の読み込み
             self.load()
 
     def load(self):
         """Load existing result csv."""
 
         # df を読み込む
-        self.local_data = pd.read_csv(self.path, encoding=self.ENCODING, header=self.HEADER_ROW)
+        df = pd.read_csv(self.path, encoding=self.ENCODING, header=self.HEADER_ROW)
 
         # metadata を読み込む
         with open(self.path, mode='r', encoding=self.ENCODING, newline='\n') as f:
@@ -461,7 +477,7 @@ class History:
             self.metadata = reader.__next__()
 
         # 最適化問題を読み込む
-        columns = self.local_data.columns
+        columns = df.columns
         prm_names = [column for i, column in enumerate(columns) if self.metadata[i] == 'prm']
         obj_names = [column for i, column in enumerate(columns) if self.metadata[i] == 'obj']
         cns_names = [column for i, column in enumerate(columns) if self.metadata[i] == 'cns']
@@ -478,13 +494,19 @@ class History:
             self.obj_names = obj_names
             self.cns_names = cns_names
 
-    @property
-    def actor_data(self):
-        return self._actor_data.get_df().result()
+        self.set_df(df)
 
-    @actor_data.setter
-    def actor_data(self, df):
-        self._actor_data.set_df(df).result()
+    def get_df(self) -> pd.DataFrame:
+        if self._actor_data is None:
+            return self._df
+        else:
+            return self._actor_data.get_df().result()
+
+    def set_df(self, df: pd.DataFrame):
+        if self._actor_data is None:
+            self._df = df
+        else:
+            self._actor_data.set_df(df).result()
 
     def create_df_columns(self):
         """Create columns of history."""
@@ -569,8 +591,8 @@ class History:
         # parameters
         for i, _row in parameters.iterrows():
             row.append(_row['value'])
-            row.append(_row['lb'])
-            row.append(_row['ub'])
+            row.append(_row['lower_bound'])
+            row.append(_row['upper_bound'])
 
         # objectives and their direction
         for (_, obj), obj_value in zip(objectives.items(), obj_values):  # objectives, direction
@@ -594,29 +616,33 @@ class History:
         row.append(datetime.datetime.now())  # time
 
         with Lock('calc-history'):
+
+            df = self.get_df()
+
             # append
-            if len(self.actor_data) == 0:
-                self.local_data = pd.DataFrame([row], columns=self.actor_data.columns)
+            if len(df) == 0:
+                df = pd.DataFrame([row], columns=df.columns)
             else:
-                self.local_data = self.actor_data
-                self.local_data.loc[len(self.local_data)] = row
+                df.loc[len(df)] = row
 
             # calc
-            self.local_data['trial'] = np.arange(len(self.local_data)) + 1  # 1 始まり
-            self._calc_non_domi(objectives)  # update self.local_data
-            self._calc_hypervolume(objectives)  # update self.local_data
-            self.actor_data = self.local_data
+            df['trial'] = np.arange(len(df)) + 1  # 1 始まり
+            self._calc_non_domi(objectives, df)  # update df
+            self._calc_hypervolume(objectives, df)  # update df
+
+            self.set_df(df)
 
             # save file
             if postprocess_args is not None:
-                trial = self.local_data['trial'].values[-1]
+                df = self.get_df()
+                trial = df['trial'].values[-1]
                 client = get_client()  # always returns valid client
                 client.run_on_scheduler(postprocess_func, trial, **postprocess_args)
 
-    def _calc_non_domi(self, objectives):
+    def _calc_non_domi(self, objectives, df):
 
         # 目的関数の履歴を取り出してくる
-        solution_set = self.local_data[self.obj_names]
+        solution_set = df[self.obj_names]
 
         # 最小化問題の座標空間に変換する
         for obj_column, (_, objective) in zip(self.obj_names, objectives.items()):
@@ -628,15 +654,13 @@ class History:
             non_domi.append((row > solution_set).product(axis=1).sum(axis=0) == 0)
 
         # 非劣解の登録
-        self.local_data['non_domi'] = non_domi
-
-    def _calc_hypervolume(self, objectives):
+        df['non_domi'] = non_domi
 
-        # タイピングが面倒
-        df = self.local_data
+    def _calc_hypervolume(self, objectives, df):
 
-        # パレート集合の抽出
+        # filter non-dominated and feasible solutions
         idx = df['non_domi'].values
+        idx = idx * df['feasible'].values  # *= を使うと non_domi 列の値が変わる
         pdf = df[idx]
         pareto_set = pdf[self.obj_names].values
         n = len(pareto_set)  # 集合の要素数
@@ -671,20 +695,29 @@ class History:
             hvs.append(hv)
 
         # 計算結果を履歴の一部に割り当て
-        df.loc[idx, 'hypervolume'] = np.array(hvs)
-
-        # dominated の行に対して、上に見ていって
-        # 最初に見つけた non-domi 行の hypervolume の値を割り当てます
+        # idx: [False, True, False, True, True, False, ...]
+        # hvs: [          1,           2,    3,        ...]
+        # want:[   0,     1,     1,    2,    3,     3, ...]
+        hvs_index = -1
         for i in range(len(df)):
-            if not df.loc[i, 'non_domi']:
-                try:
-                    df.loc[i, 'hypervolume'] = df.loc[:i][df.loc[:i]['non_domi']].iloc[-1]['hypervolume']
-                except IndexError:
-                    df.loc[i, 'hypervolume'] = 0
+
+            # process hvs index
+            if idx[i]:
+                hvs_index += 1
+
+            # get hv
+            if hvs_index < 0:
+                hypervolume = 0.
+            else:
+                hypervolume = hvs[hvs_index]
+
+            df.loc[i, 'hypervolume'] = hypervolume
 
     def save(self, _f=None):
         """Save csv file."""
 
+        df = self.get_df()
+
         if _f is None:
             # save df with columns with prefix
             with open(self.path, 'w', encoding=self.ENCODING) as f:
@@ -692,9 +725,45 @@ class History:
                 writer.writerow(self.metadata)
                 for i in range(self.HEADER_ROW-1):
                     writer.writerow([''] * len(self.metadata))
-                self.actor_data.to_csv(f, index=None, encoding=self.ENCODING, lineterminator='\n')
+                df.to_csv(f, index=None, encoding=self.ENCODING, lineterminator='\n')
         else:  # test
-            self.actor_data.to_csv(_f, index=None, encoding=self.ENCODING, lineterminator='\n')
+            df.to_csv(_f, index=None, encoding=self.ENCODING, lineterminator='\n')
+
+    def create_optuna_study(self):
+        # create study
+        kwargs = dict(
+            # storage='sqlite:///' + os.path.basename(self.path) + '_dummy.db',
+            sampler=None, pruner=None, study_name='dummy',
+            load_if_exists=True,
+        )
+
+        if len(self.obj_names) == 1:
+            kwargs.update(dict(direction='minimize'))
+        else:
+            kwargs.update(dict(directions=['minimize']*len(self.obj_names)))
+
+        study = optuna.create_study(**kwargs)
+
+        # add trial to study
+        df: pd.DataFrame = self.get_df()
+        for i, row in df.iterrows():
+            FD = optuna.distributions.FloatDistribution
+            kwargs = dict(
+                state=optuna.trial.TrialState.COMPLETE,
+                params={k: v for k, v in zip(self.prm_names, row[self.prm_names])},
+                distributions={k: FD(row[f'{k}_lower_bound'], row[f'{k}_upper_bound']) for k in self.prm_names},
+                user_attrs=None,  # TODO: add constraint information by row['feasible']
+            )
+
+            # objective or objectives
+            if len(self.obj_names) == 1:
+                kwargs.update(dict(value=row[self.obj_names].values[0]))
+            else:
+                kwargs.update(dict(values=row[self.obj_names].values))
+            trial = optuna.create_trial(**kwargs)
+            study.add_trial(trial)
+
+        return study
 
 
 class _OptimizationStatusActor:

pyfemtet/opt/femprj_sample/ParametricIF.py

@@ -27,5 +27,3 @@ if __name__ == '__main__':
         n_trials=10,
         n_parallel=1,
     )
-    input()  # required to tests script
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/cad_ex01_NX.py

@@ -133,11 +133,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=20)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/cad_ex01_SW.py

@@ -129,11 +129,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=20)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/gal_ex58_parametric.py

@@ -72,11 +72,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=10)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/gau_ex08_parametric.py

@@ -56,11 +56,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=20)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/her_ex40_parametric.py

@@ -134,11 +134,3 @@ if __name__ == '__main__':
 
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/paswat_ex1_parametric.py

@@ -58,11 +58,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/paswat_ex1_parametric_parallel.py

@@ -59,11 +59,3 @@ if __name__ == '__main__':
     femopt.set_random_seed(42)
     # femopt.optimize(n_trials=15)
     femopt.optimize(n_trials=30, n_parallel=3)  # This line is the only difference with no parallel pattern.
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/wat_ex14_parametric.py

@@ -56,11 +56,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15)
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample/wat_ex14_parametric_parallel.py

@@ -56,11 +56,3 @@ if __name__ == '__main__':
     # Run optimization.
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15, n_parallel=3)  # This line is the only difference with no parallel pattern.
-
-    # Stop script to keep process alive
-    # while you check the result in process monitor.
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/ParametricIF_jp.py

@@ -27,5 +27,3 @@ if __name__ == '__main__':
         n_trials=10,
         n_parallel=1,
     )
-    input('enter to quit')  # required to tests script
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/cad_ex01_NX_jp.py

@@ -127,11 +127,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=20)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/cad_ex01_SW_jp.py

@@ -123,11 +123,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=20)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/gal_ex58_parametric_jp.py

@@ -68,11 +68,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=10)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/gau_ex08_parametric_jp.py

@@ -55,11 +55,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=20)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/her_ex40_parametric_jp.py

@@ -134,11 +134,3 @@ if __name__ == '__main__':
 
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/paswat_ex1_parametric_jp.py

@@ -56,11 +56,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/paswat_ex1_parametric_parallel_jp.py

@@ -57,11 +57,3 @@ if __name__ == '__main__':
     femopt.set_random_seed(42)
     # femopt.optimize(n_trials=15)
     femopt.optimize(n_trials=30, n_parallel=3)  # 並列計算しない場合との差はこの行のみです。
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/wat_ex14_parametric_jp.py

@@ -54,11 +54,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15)
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()

pyfemtet/opt/femprj_sample_jp/wat_ex14_parametric_parallel_jp.py

@@ -54,11 +54,3 @@ if __name__ == '__main__':
     # 最適化を実行
     femopt.set_random_seed(42)
     femopt.optimize(n_trials=15, n_parallel=3)  # この部分のみ変更します
-
-    # プロセスモニタで結果を確認するために
-    # Enter キーが押されるまで処理を停止します。
-    print('================================')
-    print('Finished. Press Enter to quit...')
-    print('================================')
-    input()
-    femopt.terminate_all()