pyfemtet 1.2.2__tar.gz → 1.3.0__tar.gz

{pyfemtet-1.2.2 → pyfemtet-1.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyfemtet
-Version: 1.2.2
+Version: 1.3.0
 Summary: Design parameter optimization using Femtet.
 License-Expression: MIT
 License-File: LICENSE
@@ -25,7 +25,7 @@ Requires-Dist: fire (>=0.7)
 Requires-Dist: frozendict (>=2.4.6)
 Requires-Dist: numpy (>=2.0.0,<3)
 Requires-Dist: openpyxl (>=3.1.2,<4)
-Requires-Dist: optuna (>=3.4.0,<5.0.0)
+Requires-Dist: optuna (>=4.5.0,<4.6.0)
 Requires-Dist: optuna-integration (>=3.6.0,<5.0.0)
 Requires-Dist: pandas (>=2.2.3,<3)
 Requires-Dist: plotly (>=5.22.0,<6)

{pyfemtet-1.2.2 → pyfemtet-1.3.0}/pyfemtet/_util/dask_util.py

@@ -23,6 +23,7 @@ remove_all_output(get_dask_logger())
 warnings.filterwarnings('ignore', category=RuntimeWarning, message="Couldn't detect a suitable IP address")
 
 cfg.set({'distributed.scheduler.worker-ttl': None})
+cfg.set({"distributed.scheduler.locks.lease-timeout": "inf"})
 
 logger = get_module_logger('opt.dask', False)
 
@@ -62,9 +63,7 @@ def Lock(name, client=None):
 
     if client is not None:
         # import inspect
-        # logger.debug(f'{name}, {[stack.function for stack in inspect.stack()[1:]]}')
-        with cfg.set({"distributed.scheduler.locks.lease-timeout": "inf"}):
-            _lock = _DaskLock(name)
+        _lock = _DaskLock(name)
 
     else:
         if name in _lock_pool:

{pyfemtet-1.2.2 → pyfemtet-1.3.0}/pyfemtet/opt/history/_history.py

@@ -9,8 +9,7 @@ import math
 import json
 import datetime
 import dataclasses
-from time import sleep
-from contextlib import nullcontext
+from time import sleep, time
 
 import numpy as np
 import pandas as pd
@@ -51,6 +50,16 @@ MAIN_FILTER: dict = {
 
 
 logger = get_module_logger('opt.history', False)
+logger_dask = get_module_logger('opt.dask', False)
+
+
+def _assert_locked_with_timeout(lock, assertion_message=None, timeout=10.):
+    start = time()
+    while not lock.locked():
+        sleep(0.5)
+        if time() - start > timeout:
+            assert False, assertion_message or "Lock is not acquired."
+        logger_dask.debug("Lock is not acquired. Retry to check locked.")
 
 
 class MetaColumnNames(StrEnum):
@@ -153,13 +162,6 @@ class DataFrameWrapper:
     def lock(self):
         return Lock(self._lock_name)
 
-    @property
-    def lock_if_not_locked(self):
-        if self.lock.locked():
-            return nullcontext()
-        else:
-            return self.lock
-
     def get_df(self, equality_filters: dict = None) -> pd.DataFrame:
         """
 
@@ -235,7 +237,7 @@ class DataFrameWrapper:
         # partial_df を get_df した時点のものから
         # 変わっていたらエラーになる
         if equality_filters is not None:
-            assert self.lock.locked(), 'set_df() with equality_filters must be called with locking.'
+            _assert_locked_with_timeout(self.lock, 'set_df() with equality_filters must be called with locking.')
             partial_df = df
             df = self.get_df()
             apply_partial_df(df, partial_df, equality_filters)
@@ -796,7 +798,7 @@ class EntireDependentValuesCalculator:
         self.entire_df: pd.DataFrame = entire_df
         self.partial_df: pd.DataFrame = get_partial_df(entire_df, equality_filters)
 
-        assert self.records.df_wrapper.lock.locked()
+        _assert_locked_with_timeout(self.records.df_wrapper.lock)
 
         # get column names
         obj_names = self.records.column_manager.get_obj_names()
@@ -826,7 +828,7 @@ class EntireDependentValuesCalculator:
 
     def update_optimality(self):
 
-        assert self.records.df_wrapper.lock.locked()
+        _assert_locked_with_timeout(self.records.df_wrapper.lock)
 
         # calc optimality
         optimality = calc_optimality(
@@ -839,13 +841,13 @@ class EntireDependentValuesCalculator:
 
     def update_hypervolume(self):
 
-        assert self.records.df_wrapper.lock.locked()
+        _assert_locked_with_timeout(self.records.df_wrapper.lock)
 
         # calc hypervolume
         hv_values = calc_hypervolume(
             self.partial_y_internal,
             self.partial_feasibility,
-            ref_point='nadir-up-to-the-point',
+            ref_point='optuna-nadir',
         )
 
         # update
@@ -853,7 +855,7 @@ class EntireDependentValuesCalculator:
 
     def update_trial_number(self):
 
-        assert self.records.df_wrapper.lock.locked()
+        _assert_locked_with_timeout(self.records.df_wrapper.lock)
 
         # calc trial
         trial_number = 1 + np.arange(len(self.partial_df)).astype(int)
@@ -1139,45 +1141,45 @@ class Records:
 
     def update_entire_dependent_values(self, processing_df: pd.DataFrame):
 
-        with self.df_wrapper.lock_if_not_locked:
+        _assert_locked_with_timeout(self.df_wrapper.lock)
 
-            # check trial_id is filled
-            trial_processed = False
-            if processing_df['trial_id'].notna().all():
-                id_to_n: dict = {tid: i + 1 for i, tid
-                                 in enumerate(processing_df['trial_id'].unique())}
-                processing_df['trial'] = processing_df['trial_id'].map(id_to_n)
-                trial_processed = True
+        # check trial_id is filled
+        trial_processed = False
+        if processing_df['trial_id'].notna().all():
+            id_to_n: dict = {tid: i + 1 for i, tid
+                             in enumerate(processing_df['trial_id'].unique())}
+            processing_df['trial'] = processing_df['trial_id'].map(id_to_n)
+            trial_processed = True
 
-            # update main fidelity
-            equality_filters = MAIN_FILTER
+        # update main fidelity
+        equality_filters = MAIN_FILTER
+        mgr = EntireDependentValuesCalculator(
+            self,
+            equality_filters,
+            processing_df,
+        )
+        mgr.update_optimality()
+        mgr.update_hypervolume()
+        if not trial_processed:
+            mgr.update_trial_number()  # per_fidelity
+        pdf = mgr.partial_df
+        apply_partial_df(df=processing_df, partial_df=pdf, equality_filters=equality_filters)
+
+        # update sub fidelity
+        sub_fidelity_names: list = np.unique(processing_df['sub_fidelity_name']).tolist()
+        if MAIN_FIDELITY_NAME in sub_fidelity_names:
+            sub_fidelity_names.remove(MAIN_FIDELITY_NAME)
+        for sub_fidelity_name in sub_fidelity_names:
+            equality_filters = {'sub_fidelity_name': sub_fidelity_name}
             mgr = EntireDependentValuesCalculator(
                 self,
                 equality_filters,
-                processing_df,
+                processing_df
             )
-            mgr.update_optimality()
-            mgr.update_hypervolume()
             if not trial_processed:
                 mgr.update_trial_number()  # per_fidelity
-            pdf = mgr.partial_df
-            apply_partial_df(df=processing_df, partial_df=pdf, equality_filters=equality_filters)
-
-            # update sub fidelity
-            sub_fidelity_names: list = np.unique(processing_df['sub_fidelity_name']).tolist()
-            if MAIN_FIDELITY_NAME in sub_fidelity_names:
-                sub_fidelity_names.remove(MAIN_FIDELITY_NAME)
-            for sub_fidelity_name in sub_fidelity_names:
-                equality_filters = {'sub_fidelity_name': sub_fidelity_name}
-                mgr = EntireDependentValuesCalculator(
-                    self,
-                    equality_filters,
-                    processing_df
-                )
-                if not trial_processed:
-                    mgr.update_trial_number()  # per_fidelity
-            pdf = mgr.partial_df
-            apply_partial_df(df=processing_df, partial_df=pdf, equality_filters=equality_filters)
+        pdf = mgr.partial_df
+        apply_partial_df(df=processing_df, partial_df=pdf, equality_filters=equality_filters)
 
 
 class History:

{pyfemtet-1.2.2 → pyfemtet-1.3.0}/pyfemtet/opt/history/_hypervolume.py

@@ -12,7 +12,13 @@ else:
     from optuna._hypervolume import wfg
     compute_hypervolume = wfg.compute_hypervolume
 
-from ._optimality import *
+import datetime
+from optuna.visualization._hypervolume_history import _get_hypervolume_history_info, _get_hypervolume_history_plot
+
+try:
+    from ._optimality import *
+except ImportError:
+    from _optimality import *
 
 
 __all__ = [
@@ -23,7 +29,7 @@ __all__ = [
 def calc_hypervolume(
         y_internal: np.ndarray,
         feasibility: np.ndarray,
-        ref_point: str | np.ndarray = 'nadir-up-to-the-point'
+        ref_point: str | np.ndarray = 'optuna-nadir'
 ) -> np.ndarray:
     """
 
@@ -57,6 +63,8 @@ def calc_hypervolume(
         elif ref_point.lower() == 'worst':
             hv = calc_hypervolume_worst(y_internal, feasibility)
             hv_values = hv * np.ones(len(y_internal)).astype(float)
+        elif ref_point.lower() == 'optuna-nadir':
+            hv_values = calc_hypervolume_by_optuna(y_internal, feasibility)
         else:
             raise NotImplementedError
 
@@ -80,12 +88,13 @@ def get_pareto_set(
     return non_dominated_solutions
 
 
-def calc_hypervolume_nadir(y, feasibility) -> float:
+def calc_hypervolume_nadir(y, feasibility, ref_point=None) -> float:
     """Use Nadir point as the ref_point.
 
     Args:
         y: (n, m) shaped 2d-array. float.
         feasibility (np.ndarray): n shaped 1d-array. bool.
+        ref_point: (m) shaped array. float.
 
     Returns: float.
 
@@ -95,8 +104,10 @@ def calc_hypervolume_nadir(y, feasibility) -> float:
     if len(pareto_set) == 0:
         return 0.
 
-    nadir_point = pareto_set.max(axis=0)
-    ref_point = nadir_point + 1e-8
+    if ref_point is None:
+        nadir_point = pareto_set.max(axis=0)
+        ref_point = nadir_point + 1e-8
+
     hv = compute_hypervolume(pareto_set, ref_point)
 
     return hv
@@ -116,10 +127,26 @@ def calc_hypervolume_nadir_up_to_the_point(y, feasibility) -> np.ndarray:
     out = []
 
     assert len(y) == len(feasibility)
+
+    nadir_points = []
     for n in range(len(y)):
         y_up = y[:n]
         f_up = feasibility[:n]
-        out.append(calc_hypervolume_nadir(y_up, f_up))
+        pareto_set = get_pareto_set(y_up, f_up)
+        if len(pareto_set) == 0:
+            continue
+        nadir_points.append(pareto_set.max(axis=0))
+
+    if len(nadir_points) == 0:
+        return np.zeros(len(y))
+
+    nadir_point = np.array(nadir_points).max(axis=0)
+    ref_point = nadir_point + 1e-8
+
+    for n in range(len(y)):
+        y_up = y[:n]
+        f_up = feasibility[:n]
+        out.append(calc_hypervolume_nadir(y_up, f_up, ref_point=ref_point))
 
     return np.array(out).astype(float)
 
@@ -167,3 +194,34 @@ def calc_hypervolume_fixed_point(y, feasibility, ref_point) -> float:
     hv = compute_hypervolume(feasible_solutions, ref_point)
 
     return hv
+
+
+def calc_hypervolume_by_optuna(y, f):
+    study = optuna.create_study(
+        directions=['minimize'] * y.shape[-1]
+    )
+    number = 0
+    for y_ in y:
+        system_attrs = {}
+        if np.isnan(y_).any():
+            system_attrs.update({"constraints": [1.]})
+            y_ = y[0]
+        trial = optuna.trial.FrozenTrial(
+            number=number,
+            value=None,
+            values=y_,
+            state=optuna.trial.TrialState.COMPLETE,
+            datetime_start=datetime.datetime.now(),
+            datetime_complete=datetime.datetime.now(),
+            params={},
+            distributions={},
+            user_attrs={},
+            system_attrs=system_attrs,
+            intermediate_values={},
+            trial_id=number,
+        )
+        study.add_trial(trial)
+        number += 1
+    ref_point = get_pareto_set(y, f).max(axis=0) + 1e-8
+    info = _get_hypervolume_history_info(study, reference_point=ref_point)
+    return np.array(info.values)

{pyfemtet-1.2.2 → pyfemtet-1.3.0}/pyfemtet/opt/interface/_femtet_interface/femtet_interface.py

@@ -354,7 +354,7 @@ class FemtetInterface(COMInterface):
 
         """
         # noinspection PyGlobalUndefined
-        global constants
+        global constants, Dispatch
 
         if connect_method == "new":
             self._connect_new_femtet()
@@ -379,8 +379,9 @@ class FemtetInterface(COMInterface):
 
             import win32com.client
             importlib.reload(win32com.client)
-            from win32com.client import Dispatch, constants
+            Dispatch = win32com.client.Dispatch
             Dispatch('FemtetMacro.Femtet')
+            constants = win32com.client.constants
 
             if not hasattr(constants, "STATIC_C"):
                 message = _(

{pyfemtet-1.2.2 → pyfemtet-1.3.0}/pyproject.toml

@@ -22,7 +22,7 @@ dependencies = [
     "pandas>=2.2.3,<3",
     "openpyxl>=3.1.2,<4",
     "scipy>=1.11.4,<2",
-    "optuna>=3.4.0, <5.0.0",
+    "optuna~=4.5.0",
     "optuna-integration>=3.6.0, <5.0.0",
     "botorch>=0.12.0, <0.15.0",
     "torch", # botorch specifies torch's version
@@ -33,7 +33,7 @@ dependencies = [
 
 #[project.optional-dependencies]
 #brep = ["brepmatching>=0.1.8,<0.2"]
-version = "1.2.2"
+version = "1.3.0"
 
 [project.urls]
 Repository = "https://github.com/pyfemtet/pyfemtet"