pyfemtet 0.5.4__tar.gz → 0.6.0__tar.gz

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: pyfemtet
-Version: 0.5.4
+Version: 0.6.0
 Summary: Design parameter optimization using Femtet.
 Home-page: https://github.com/pyfemtet/pyfemtet
 License: BSD-3-Clause

pyfemtet-0.6.0/pyfemtet/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.6.0"

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/_message/locales/ja/LC_MESSAGES/messages.po

@@ -80,7 +80,7 @@ msgstr "モデル更新に失敗しました。"
 
 #: pyfemtet/message/messages.py:60
 msgid "It was detected that the configuration of Femtet python macro constants has not been completed. The configuration was done automatically (python -m win32com.client.makepy FemtetMacro). Please restart the program."
-msgstr "Femtet マクロの定数の設定が未完了であることが検出され、設定コマンド（python -m win23com.client.makepy FemtetMacro）が自動実行されました。プログラムを再実行してください。"
+msgstr "Femtet マクロの定数の設定が未完了であることが検出され、設定コマンド（python -m win32com.client.makepy FemtetMacro）が自動実行されました。プログラムを再実行してください。"
 
 #: pyfemtet/message/messages.py:61
 msgid "Failed to connect to Femtet."

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/opt/__init__.py

@@ -9,6 +9,8 @@ from pyfemtet.opt.optimizer import AbstractOptimizer
 
 from pyfemtet.opt._femopt import FEMOpt
 
+from pyfemtet.opt._femopt_core import History
+
 
 __all__ = [
     'FEMOpt',
@@ -21,4 +23,5 @@ __all__ = [
     'OptunaOptimizer',
     'ScipyScalarOptimizer',
     'ScipyOptimizer',
+    'History',
 ]

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/opt/_femopt.py

@@ -12,7 +12,7 @@ from traceback import print_exception
 # 3rd-party
 import numpy as np
 import pandas as pd
-from dask.distributed import LocalCluster, Client, Worker
+from dask.distributed import LocalCluster, Client
 
 # pyfemtet relative
 from pyfemtet.opt.interface import FEMInterface, FemtetInterface
@@ -32,10 +32,13 @@ from pyfemtet._message import Msg, encoding
 from pyfemtet.opt.optimizer.parameter import Parameter, Expression
 from pyfemtet._warning import experimental_feature
 
+from dask import config as cfg
+cfg.set({'distributed.scheduler.worker-ttl': None})
 
-def add_worker(client, worker_name):
+
+def add_worker(client, worker_name, n_workers=1):
     import sys
-    from subprocess import Popen, DEVNULL
+    from subprocess import Popen, DEVNULL, PIPE
 
     current_n_workers = len(client.nthreads().keys())
 
@@ -43,16 +46,16 @@ def add_worker(client, worker_name):
         f'{sys.executable} -m dask worker '
         f'{client.scheduler.address} '
         f'--nthreads 1 '
-        f'--nworkers 1 '
-        f'--name {worker_name} '
-        f'--no-nanny',
+        f'--nworkers {n_workers} '
+        f'--name {worker_name} ',  # A unique name for this worker like ‘worker-1’. If used with -nworkers then the process number will be appended like name-0, name-1, name-2, …
+        # --no-nanny option は --nworkers と併用できない
         shell=True,
         stderr=DEVNULL,
         stdout=DEVNULL,
     )
 
     # worker が増えるまで待つ
-    client.wait_for_workers(n_workers=current_n_workers + 1)
+    client.wait_for_workers(n_workers=current_n_workers + n_workers)
 
 
 class FEMOpt:
@@ -132,6 +135,7 @@ class FEMOpt:
         self.monitor_process_future = None
         self.monitor_server_kwargs = dict()
         self.monitor_process_worker_name = None
+        self._hv_reference = None
 
     # multiprocess 時に pickle できないオブジェクト参照の削除
     def __getstate__(self):
@@ -385,9 +389,9 @@ class FEMOpt:
 
         Warnings:
 
-            When ```strict``` == True and using OptunaOptimizer along with BoTorchSampler,
-            Pyfemtet will solve an optimization subproblem to propose new variables.
-            During this process, the constraint function fun will be executed at each
+            When ```strict``` == True and using OptunaOptimizer along with :class:`PoFBoTorchSampler`,
+            PyFemtet will solve an optimization subproblem to propose new variables.
+            During this process, the constraint function ```fun``` will be executed at each
             iteration of the subproblem, which may include time-consuming operations such
             as retrieving 3D model information via FEMInterface.
             As a result, it may not be feasible to complete the overall optimization within
@@ -406,7 +410,7 @@ class FEMOpt:
 
                 Instead, please do the following.
 
-                    >>> def bottom_area(_, opt):  # 第一引数 Femtet にはアクセスしないようにします。  # doctest: +SKIP
+                    >>> def bottom_area(_, opt):  # Not to access the 1st argument.  # doctest: +SKIP
                     ...     params = opt.get_parameter()  # doctest: +SKIP
                     ...     w, d = params['width'], params['depth']  # doctest: +SKIP
                     ...     return w * d  # doctest: +SKIP
@@ -623,14 +627,21 @@ class FEMOpt:
 
         else:
             # ローカルクラスターを構築
-            logger.info('Launching single machine cluster. This may take tens of seconds.')
+            logger.info('Launching single machine cluster... This may take tens of seconds.')
+
+            # Fixed:
+            #   Nanny の管理機能は必要ないが、Python API では worker_class を Worker にすると
+            #   processes 引数が無視されて Thread worker が立てられる。
+            #   これは CLI の --no-nanny オプションも同様らしい。
+
+            # クラスターの構築
             cluster = LocalCluster(
                 processes=True,
-                n_workers=n_parallel,  # n_parallel = n_parallel - 1 + 1; main 分減らし、monitor 分増やす
+                n_workers=n_parallel,
                 threads_per_worker=1,
-                worker_class=Worker,
             )
             logger.info('LocalCluster launched successfully.')
+
             self.client = Client(cluster, direct_to_workers=False)
             self.scheduler_address = self.client.scheduler.address
             logger.info('Client launched successfully.')
@@ -655,6 +666,7 @@ class FEMOpt:
                         model_name=self.fem.model_name
                     )
                 )
+
                 # Femtet の parametric 設定を目的関数に用いるかどうか
                 if self.fem.parametric_output_indexes_use_as_objective is not None:
                     from pyfemtet.opt.interface._femtet_parametric import add_parametric_results_as_objectives
@@ -665,8 +677,14 @@ class FEMOpt:
                         indexes,
                         directions,
                     )
+
+                # Femtet の confirm_before_exit のセット
+                self.fem.confirm_before_exit = confirm_before_exit
+                self.fem.kwargs['confirm_before_exit'] = confirm_before_exit
+
                 logger.info('Femtet loaded successfully.')
 
+
             # actor の設定
             self.status = OptimizationStatus(_client)
             self.worker_status_list = [OptimizationStatus(_client, name) for name in worker_addresses]  # tqdm 検討
@@ -678,6 +696,7 @@ class FEMOpt:
                 list(self.opt.constraints.keys()),
                 _client,
                 metadata,
+                self._hv_reference
             )
             logger.info('Status Actor initialized successfully.')
 

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/opt/_femopt_core.py

@@ -453,18 +453,12 @@ class History:
         cns_names (List[str], optional): The names of constraints. Defaults to None.
         client (dask.distributed.Client): Dask client.
         additional_metadata (str, optional): metadata of optimization process.
-
-    Raises:
-        FileNotFoundError: If the csv file is not found.
-
-    Attributes:
-        HEADER_ROW (int): Header row number of csv file. Must be grater than 0. Default to 2.
-        ENCODING (str): Encoding of csv file. Default to 'cp932'.
-        prm_names (str): User defined names of parameters.
-        obj_names (str): User defined names of objectives.
-        cns_names (str): User defined names of constraints.
-        is_restart (bool): If the optimization process is a continuation of another process or not.
-        is_processing (bool): The optimization is running or not.
+        hv_reference (str or list[float or np.ndarray, optional):
+            The method to calculate hypervolume or
+            the reference point itself.
+            Valid values are 'dynamic-pareto' or
+            'dynamic-nadir' or 'nadir' or 'pareto'
+            or fixed point (in objective function space).
 
     """
 
@@ -487,7 +481,10 @@ class History:
             cns_names=None,
             client=None,
             additional_metadata=None,
+            hv_reference=None,
     ):
+        # hypervolume 計算メソッド
+        self._hv_reference = hv_reference or 'dynamic-pareto'
 
         # 引数の処理
         self.path = history_path  # .csv
@@ -739,59 +736,98 @@ class History:
 
     def _calc_hypervolume(self, objectives, df):
 
-        # 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
-        m = len(pareto_set.T)  # 目的変数数
-        # 多目的でないと計算できない
-        if m <= 1:
-            return None
-        # 長さが 2 以上でないと計算できない
-        if n <= 1:
-            return None
-        # 最小化問題に convert
-        for i, (_, objective) in enumerate(objectives.items()):
-            for j in range(n):
-                pareto_set[j, i] = objective.convert(pareto_set[j, i])
-                #### reference point の計算[1]
-        # 逆正規化のための範囲計算
-        maximum = pareto_set.max(axis=0)
-        minimum = pareto_set.min(axis=0)
-
-        r = 1.01
-
-        # r を逆正規化
-        reference_point = r * (maximum - minimum) + minimum
-
-        #### hv 履歴の計算
-        hvs = []
-        for i in range(n):
-            hv = compute_hypervolume(pareto_set[:i], reference_point)
-            if np.isnan(hv):
-                hv = 0
-            hvs.append(hv)
-
-        # 計算結果を履歴の一部に割り当て
-        # 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)):
-
-            # process hvs index
-            if idx[i]:
-                hvs_index += 1
-
-            # get hv
-            if hvs_index < 0:
-                hypervolume = 0.
+        if len(objectives) < 2:
+            df.loc[len(df) - 1, 'hypervolume'] = 0.
+            return
+
+        # 最小化問題に変換された objective values を取得
+        raw_objective_values = df[self.obj_names].values
+        objective_values = np.empty_like(raw_objective_values)
+        for n_trial in range(len(raw_objective_values)):
+            for obj_idx, (_, objective) in enumerate(objectives.items()):
+                objective_values[n_trial, obj_idx] = objective.convert(raw_objective_values[n_trial, obj_idx])
+
+        # pareto front を取得
+        def get_pareto(objective_values_, with_partial=False):
+            ret = None
+            if with_partial:
+                ret = []
+
+            pareto_set_ = np.empty((0, len(self.obj_names)))
+            for i in range(len(objective_values_)):
+                target = objective_values_[i]
+                dominated = False
+                # TODO: Array の計算に直して高速化する
+                for j in range(len(objective_values_)):
+                    compare = objective_values_[j]
+                    if all(target > compare):
+                        dominated = True
+                        break
+                if not dominated:
+                    pareto_set_ = np.concatenate([pareto_set_, [target]], axis=0)
+
+                if ret is not None:
+                    ret.append(np.array(pareto_set_))
+
+            if ret is not None:
+                return pareto_set_, ret
             else:
-                hypervolume = hvs[hvs_index]
+                return pareto_set_
+
+        if self._hv_reference == 'dynamic-pareto':
+            pareto_set, pareto_set_list = get_pareto(objective_values, with_partial=True)
+            for i, partial_pareto_set in enumerate(pareto_set_list):
+                ref_point = pareto_set.max(axis=0) + 1e-8
+                hv = compute_hypervolume(partial_pareto_set, ref_point)
+                df.loc[i, 'hypervolume'] = hv
+            return
+
+        elif self._hv_reference == 'dynamic-nadir':
+            _, pareto_set_list = get_pareto(objective_values, with_partial=True)
+            for i, partial_pareto_set in enumerate(pareto_set_list):
+                ref_point = objective_values.max(axis=0) + 1e-8
+                hv = compute_hypervolume(partial_pareto_set, ref_point)
+                df.loc[i, 'hypervolume'] = hv
+            return
+
+        elif self._hv_reference == 'nadir':
+            pareto_set = get_pareto(objective_values)
+            ref_point = objective_values.max(axis=0) + 1e-8
+            hv = compute_hypervolume(pareto_set, ref_point)
+            df.loc[len(df) - 1, 'hypervolume'] = hv
+            return
+
+        elif self._hv_reference == 'pareto':
+            pareto_set = get_pareto(objective_values)
+            ref_point = pareto_set.max(axis=0) + 1e-8
+            hv = compute_hypervolume(pareto_set, ref_point)
+            df.loc[len(df) - 1, 'hypervolume'] = hv
+            return
+
+        elif (
+                isinstance(self._hv_reference, np.ndarray)
+                or isinstance(self._hv_reference, list)
+        ):
+            _buff = np.array(self._hv_reference)
+            assert _buff.shape == (len(self.obj_names),)
+
+            ref_point = np.array(
+                [obj.convert(raw_value) for obj, raw_value in zip(objectives.values(), _buff)]
+            )
 
-            df.loc[i, 'hypervolume'] = hypervolume
+            _buff = get_pareto(objective_values)
+
+            pareto_set = np.empty((0, len(objectives)))
+            for pareto_sol in _buff:
+                if all(pareto_sol < ref_point):
+                    pareto_set = np.concatenate([pareto_set, [pareto_sol]], axis=0)
+
+            hv = compute_hypervolume(pareto_set, ref_point)
+            df.loc[len(df) - 1, 'hypervolume'] = hv
+            return
+
+        else:
+            raise NotImplementedError(f'Invalid Hypervolume reference point calculation method: {self._hv_reference}')
 
     def save(self, _f=None):
         """Save csv file."""

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/opt/interface/_femtet.py

@@ -78,6 +78,10 @@ class FemtetInterface(FEMInterface):
             it will be None and no parametric outputs are used
             as objectives.
 
+        confirm_before_exit (bool):
+            Whether to confirm before (abnormal) termination.
+            Default is True.
+
         **kwargs: Additional arguments from inherited classes.
 
     Warning:
@@ -103,6 +107,7 @@ class FemtetInterface(FEMInterface):
             allow_without_project: bool = False,  # main でのみ True を許容したいので super() の引数にしない。
             open_result_with_gui: bool = True,
             parametric_output_indexes_use_as_objective: dict[int, str or float] = None,
+            confirm_before_exit: bool = True,
             **kwargs  # 継承されたクラスからの引数
     ):
 
@@ -132,6 +137,7 @@ class FemtetInterface(FEMInterface):
         self.parametric_output_indexes_use_as_objective = parametric_output_indexes_use_as_objective
         self._original_autosave_enabled = _get_autosave_enabled()
         _set_autosave_enabled(False)
+        self.confirm_before_exit = confirm_before_exit
 
         # dask サブプロセスのときは femprj を更新し connect_method を new にする
         try:
@@ -169,6 +175,7 @@ class FemtetInterface(FEMInterface):
             open_result_with_gui=self.open_result_with_gui,
             parametric_output_indexes_use_as_objective=self.parametric_output_indexes_use_as_objective,
             save_pdt=self.save_pdt,
+            confirm_before_exit=self.confirm_before_exit,
             **kwargs
         )
 
@@ -241,7 +248,8 @@ class FemtetInterface(FEMInterface):
             print('================')
             print(message)
             print('================')
-            input('Press enter to finish...')
+            if self.confirm_before_exit:
+                input('Press enter to finish...')
             raise RuntimeError(message)
 
         if self.Femtet is None:
@@ -532,7 +540,8 @@ class FemtetInterface(FEMInterface):
                 logger.error(Msg.ERR_CANNOT_ACCESS_API + 'GetVariableNames_py')
                 logger.error(Msg.CERTIFY_MACRO_VERSION)
                 print('================')
-                input(Msg.ENTER_TO_QUIT)
+                if self.confirm_before_exit:
+                    input(Msg.ENTER_TO_QUIT)
                 raise e
                 
             if variable_names is not None:
@@ -543,7 +552,8 @@ class FemtetInterface(FEMInterface):
             logger.error(message)
             logger.error(f'`{param_name}` not in {variable_names}')
             print('================')
-            input(Msg.ENTER_TO_QUIT)
+            if self.confirm_before_exit:
+                input(Msg.ENTER_TO_QUIT)
             raise RuntimeError(message)
         else:
             return None
@@ -736,7 +746,8 @@ class FemtetInterface(FEMInterface):
                 logger.error(Msg.ERR_CANNOT_ACCESS_API + 'Femtet.Exit()')
                 logger.error(Msg.CERTIFY_MACRO_VERSION)
                 print('================')
-                input(Msg.ENTER_TO_QUIT)
+                if self.confirm_before_exit:
+                    input(Msg.ENTER_TO_QUIT)
                 raise e
 
         else:

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/opt/optimizer/__init__.py

@@ -1,12 +1,16 @@
 from pyfemtet.opt.optimizer._base import AbstractOptimizer, logger, OptimizationMethodChecker
-from pyfemtet.opt.optimizer._optuna import OptunaOptimizer
+from pyfemtet.opt.optimizer._optuna._optuna import OptunaOptimizer
 from pyfemtet.opt.optimizer._scipy import ScipyOptimizer
 from pyfemtet.opt.optimizer._scipy_scalar import ScipyScalarOptimizer
 
+from pyfemtet.opt.optimizer._optuna._pof_botorch import PoFBoTorchSampler, PoFConfig
+
 __all__ = [
     'ScipyScalarOptimizer',
     'ScipyOptimizer',
     'OptunaOptimizer',
     'AbstractOptimizer',
     'logger',
+    'PoFBoTorchSampler',
+    'PoFConfig',
 ]

{pyfemtet-0.5.4 → pyfemtet-0.6.0}/pyfemtet/opt/optimizer/_base.py

@@ -142,6 +142,7 @@ class AbstractOptimizer(ABC):
         self.subprocess_idx = None
         self._exception = None
         self.method_checker: OptimizationMethodChecker = OptimizationMethodChecker(self)
+        self._retry_counter = 0
 
     # ===== algorithm specific methods =====
     @abstractmethod
@@ -176,13 +177,11 @@ class AbstractOptimizer(ABC):
 
         # Optimizer の x の更新
         self.set_parameter_values(x)
-
-        logger.info('---------------------')
         logger.info(f'input: {x}')
 
         # FEM の更新
-        logger.debug('fem.update() start')
         try:
+            logger.info(f'Solving FEM...')
             df_to_fem = self.variables.get_variables(
                 format='df',
                 filter_pass_to_fem=True
@@ -193,25 +192,20 @@ class AbstractOptimizer(ABC):
             logger.info(f'{type(e).__name__} : {e}')
             logger.info(Msg.INFO_EXCEPTION_DURING_FEM_ANALYSIS)
             logger.info(x)
-            raise e  # may be just a ModelError, etc.
+            raise e  # may be just a ModelError, etc. Handling them in Concrete classes.
 
         # y, _y, c の更新
-        logger.debug('calculate y start')
         y = [obj.calc(self.fem) for obj in self.objectives.values()]
 
-        logger.debug('calculate _y start')
         _y = [obj.convert(value) for obj, value in zip(self.objectives.values(), y)]
 
-        logger.debug('calculate c start')
         c = [cns.calc(self.fem) for cns in self.constraints.values()]
 
-        logger.debug('history.record start')
-
+        # register to history
         df_to_opt = self.variables.get_variables(
             format='df',
             filter_parameter=True,
         )
-
         self.history.record(
             df_to_opt,
             self.objectives,
@@ -223,9 +217,7 @@ class AbstractOptimizer(ABC):
             postprocess_args=self.fem._create_postprocess_args(),
         )
 
-        logger.debug('history.record end')
-
-        logger.info(f'output: {_y}')
+        logger.info(f'output: {y}')
 
         return np.array(y), np.array(_y), np.array(c)