surface-construct 0.12.5__tar.gz → 0.12.7__tar.gz

{surface_construct-0.12.5/surface_construct.egg-info → surface_construct-0.12.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: surface_construct
-Version: 0.12.5
+Version: 0.12.7
 Summary: Surface construction and surface reaction sampling tools.
 Home-page: https://gitee.com/pjren/surface_construct/
 Author: ren

{surface_construct-0.12.5 → surface_construct-0.12.7}/setup.py

@@ -15,7 +15,7 @@ install_requires = [
 
 setup(
     name='surface_construct',
-    version='0.12.5',
+    version='0.12.7',
     packages=find_packages(),
     url='https://gitee.com/pjren/surface_construct/',
     license='GPL',

{surface_construct-0.12.5 → surface_construct-0.12.7}/surface_construct/sg_sampler.py

@@ -165,11 +165,22 @@ class MaxSigmaSGSampler(SGSamplerBase):
     def _samples(self, size, **kwargs):
         if 'energy' in self.sg_obj.grid_property:
             # 如果已经读入了一些能量，则返回误差最大的点
-            sigma_array = self.sg_obj.grid_property['energy']
+            sigma_array = self.sg_obj.grid_property_sigma['energy']
             sigma0 = sigma_array.max()
-            idx_list = np.argwhere(sigma_array <= sigma0-0.1).flatten().tolist()
-            idx = random.sample(idx_list, size)
-            return idx
+            if len(sigma_array) < size:
+                raise ValueError("Number of points should greater than sample!")
+            elif len(sigma_array) == size:
+                return np.asarray(range(len(sigma_array)))
+            else:
+                delta = 0.01
+                idx_list = []
+                while len(idx_list) < size:
+                    idx_list = np.argwhere(sigma_array > sigma0-delta).flatten().tolist()
+                    # 排除掉已经采过的点
+                    idx_list = [i for i in idx_list if i not in self.sg_obj.calculated_sample]
+                    delta += 0.01
+                idx = random.sample(idx_list, min(size, len(idx_list)))
+                return idx
         else:
             raise "No energy for all population, pls do initial sampling first!"
 
@@ -181,10 +192,19 @@ class MinEnergySGSampler(SGSamplerBase):
     def _samples(self, size, **kwargs):
         if 'energy' in self.sg_obj.grid_property:
             E_array = self.sg_obj.grid_property['energy']
-            # 如果已经读入了一些能量，则返回能量最低的点 (<0.1eV 以内，然后随机选一个)
+            if len(E_array) < size:
+                raise ValueError("Number of points should greater than sample!")
+            elif len(E_array) == size:
+                return np.asarray(range(len(E_array)))
             E0 = E_array.min()
-            idx_list = np.argwhere(E_array <= E0+0.1).flatten().tolist()
-            idx = random.sample(idx_list, size)
+            delta = 0.1
+            idx_list = []
+            while len(idx_list) < size:
+                # 如果已经读入了一些能量，则返回能量最低的点 (<delta eV 以内，然后随机选一个)
+                idx_list = np.argwhere(E_array <= E0+0.1).flatten().tolist()
+                idx_list = [i for i in idx_list if i not in self.sg_obj.calculated_sample]
+                delta += 0.1
+            idx = random.sample(idx_list, min(size, len(idx_list)))
             return idx
         else:
             raise "No energy for all population, pls do initial sampling first!"

{surface_construct-0.12.5 → surface_construct-0.12.7}/surface_construct/structures/surface_grid.py

@@ -770,13 +770,19 @@ class SurfaceGrid:
         :return:
         """
         if self._vector_interval is None:
-            nsample = 100
+            nsample = min(100, len(self.points)-1)  # 如果 points 被过滤了，就需要修改了。
             rng = np.random.default_rng()
-            idx_0 = rng.choice(range(len(self.points)-1), size=nsample)
+            if nsample < 10:
+                comb = itertools.combinations(range(len(self.points)-1), nsample)
+                idx_0 = rng.choice(list(comb))
+            else:
+                idx_0 = rng.choice(range(len(self.points)-1), size=nsample)
             idx_1 = idx_0+1
-            idx = np.asarray([[i,j] for i,j in zip(idx_0, idx_1) if (i not in idx_1 and j not in idx_0)])
+            idx = np.asarray([[i,j] for i,j in zip(idx_0, idx_1)])
             d_grid = np.linalg.norm(self.points[idx[:, 0]] - self.points[idx[:, 1]], axis=1)
-            idx = idx[d_grid<1.2 * self.interval]
+            tol = 1.2 * self.interval
+            if np.any(d_grid<tol):
+                idx = idx[d_grid<tol]
             d_vector = np.linalg.norm(self.vector[idx[:,0]]-self.vector[idx[:,1]], axis=1).mean()
             k = d_vector / self.interval
             self._vector_interval = np.min(k)
@@ -917,11 +923,11 @@ class SurfaceGrid:
         try:
             length_scale_grid = 1.0  # 实空间的length scale, 1 angstrom
             length_scale_vector = length_scale_grid * self.vector_interval
-            length_scale_bounds = (length_scale_vector/2.0, length_scale_vector*2.0)
+            length_scale_bounds = (length_scale_vector/10.0, length_scale_vector*10.0)
             length_scale = length_scale_vector  # 向量空间的 length_scale, isotropic
         except IndexError:
             length_scale = 1.0
-            length_scale_bounds = (1e-5, 1e5)
+            length_scale_bounds = (1e-3, 1e3)
         rbf_kernel = RBF(length_scale=length_scale, length_scale_bounds=length_scale_bounds)
         # noise kernel
         noise_level_dict = {
@@ -935,7 +941,7 @@ class SurfaceGrid:
         noise_level_norm = np.abs(noise_level / scaler.scale_[0])  # 标准化缩放
         white_kernel = WhiteKernel(noise_level=noise_level_norm, noise_level_bounds='fixed')
         # 总的 kernel
-        constant_kernel = ConstantKernel(constant_value=1.0, constant_value_bounds='fixed')
+        constant_kernel = ConstantKernel(constant_value=1.0, constant_value_bounds=(0.1,10))
         kernel = constant_kernel * rbf_kernel + white_kernel
         gp = GaussianProcessRegressor(kernel=kernel, n_restarts_optimizer=9, alpha=noise_level_norm*0.01)
         print(f"Kernel parameters before fit:{kernel})")

{surface_construct-0.12.5 → surface_construct-0.12.7}/surface_construct/tasks/sitesampling.py

@@ -171,6 +171,11 @@ class SurfaceSiteSampleTask(TaskBase):
             self.sg_obj.del_sample(grid_idx)
 
     def run(self):
+        # 在新的采样之前先进行已经采到但是未计算的点
+        uncalc_samples = [idx for idx in self.sg_obj.sample_idx if not self.sg_obj.sample_dct[idx]['calculated']]
+        for grid_idx in uncalc_samples:
+            self.irun(grid_idx, fit=False)
+
         for isampler in self.sampler:
             size = isampler.get("size", 1)
             sg_sampler = isampler.get("surface")

{surface_construct-0.12.5 → surface_construct-0.12.7}/surface_construct/tasks/vipsitetask.py

@@ -1,7 +1,5 @@
-import random
+import os
 
-import ase.data
-import numpy as np
 from matplotlib import pyplot as plt
 from matplotlib.patches import Circle
 
@@ -10,40 +8,43 @@ from .sitesampling import SurfaceSiteSampleTask
 class VIPSiteTask(SurfaceSiteSampleTask):
     def __init__(self, exclude=None, include=None, **kwargs):
         super().__init__(**kwargs)
+        if not os.path.isfile(self.hist_pkl):
+            vip_id = self.sg_obj.vip_id
+            grid_site_type, site_type_dict = self.sg_obj.get_grid_site_type()
+            if exclude is None and include is None:
+                msg = [f"Either exclude or include must be specified"]
+                msg += [f"Possible site types are {len(site_type_dict)} in total:"]
+                msg += [f"{k}: {v}" for k,v in site_type_dict.items() if type(k) in (int,)]
+                raise ValueError('\n'.join(msg))
+            elif exclude is not None and include is not None:
+                raise ValueError("Can not specified exclude and include at the same time")
+            elif exclude is not None and include is None:
+                filtered = [i for i in vip_id if site_type_dict[grid_site_type[i]] not in exclude]
+            else:
+                filtered = [i for i in vip_id if site_type_dict[grid_site_type[i]] in include]
 
-        vip_id = self.sg_obj.vip_id
-        grid_site_type, site_type_dict = self.sg_obj.get_grid_site_type()
-        if exclude is None and include is None:
-            msg = [f"Either exclude or include must be specified"]
-            msg += [f"Possible site types are {len(site_type_dict)} in total:"]
-            msg += [f"{k}: {v}" for k,v in site_type_dict.items() if type(k) in (int,)]
-            raise ValueError('\n'.join(msg))
-        elif exclude is not None and include is not None:
-            raise ValueError("Can not specified exclude and include at the same time")
-        elif exclude is not None and include is None:
-            filtered = [i for i in vip_id if site_type_dict[grid_site_type[i]] not in exclude]
-        else:
-            filtered = [i for i in vip_id if site_type_dict[grid_site_type[i]] in include]
-
-        points = self.sg_obj.points[filtered]
-        vector = self.sg_obj.vector[filtered]
-        raw_vector = self.sg_obj._raw_vector[filtered]
-        self.sg_obj.points = points
-        self.sg_obj.vector = vector
-        self.sg_obj._raw_vector = raw_vector
-        self.sg_obj.vectorgen = None
+            points = self.sg_obj.points[filtered]
+            vector = self.sg_obj.vector[filtered]
+            raw_vector = self.sg_obj._raw_vector[filtered]
+            _ = self.sg_obj.vector_interval  # 此处call 一下，用于后面的 fit 中的参数
+            self.sg_obj.points = points
+            self.sg_obj.vector = vector
+            self.sg_obj._raw_vector = raw_vector
+            self.sg_obj.vectorgen = None
 
-        sampler_size = 0
+        sampler_size = len(self.sg_obj.calculated_sample)
         new_sampler = []
         for isampler in self.sampler:  # 不需要初始采样和key 点采样
             if isampler['surface'] not in ('InitialSGSampler', 'KeyPointSGSampler'):
                 sampler_size += isampler.get('size', 0)
                 new_sampler.append(isampler)
+            else:
+                print("Warning: InitialSGSampler or KeyPointSGSampler should not applied in VIP sampling.")
 
         # 如果 filter 之后的点小于等于 self.sampler['size'], 重新定义sampler
-        if sampler_size >= len(points):
+        if sampler_size >= len(self.sg_obj.points):
             self.sampler = [{
-                'size': len(points),  # 采样大小
+                'size': len(self.sg_obj.points)-len(self.sg_obj.calculated_sample),  # 采样大小
                 'surface': "RandomSGSampler",  # 表面采样方法
             }]
         else:

{surface_construct-0.12.5 → surface_construct-0.12.7/surface_construct.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: surface_construct
-Version: 0.12.5
+Version: 0.12.7
 Summary: Surface construction and surface reaction sampling tools.
 Home-page: https://gitee.com/pjren/surface_construct/
 Author: ren

{surface_construct-0.12.5 → surface_construct-0.12.7}/tests/test_task.py

@@ -3,6 +3,7 @@ import shutil
 from random import randint
 
 import ase.io
+import numpy as np
 import pytest
 from ase import Atom, Atoms
 from ase.constraints import FixAtoms
@@ -175,7 +176,7 @@ class TestTask:
                 'weight': (0.5, 0.5),  # 表面采样方法的权重
             }  # 第二步采样
         ]
-        task_obj = AFMTask(combiner=com_obj, sampler=sampler, optimizer=LBFGS, nz=10) # nz 定义 z方向采多少样
+        task_obj = AFMTask(combiner=com_obj, sampler=sampler, optimizer=None, nz=3) # nz 定义 z方向采多少样
         task_obj.print_task_info()
         task_obj.run()
         print('Done')
@@ -195,12 +196,12 @@ class TestTask:
         ads_grid_comb = AdsGridCombiner(sg_obj, ads_obj)
         sampler =[
             {
+                'surface': "RandomSGSampler",  # 第一步，随机表面采样方法。
                 'size': 2,  # 采样大小
-                'surface': "RandomSGSampler",  # 表面采样方法
             },
             {
-                'size': 4,  # 采样大小
                 'surface': ("MaxDiversitySGSampler", "MinEnergySGSampler", "MaxSigmaSGSampler"),  # 表面采样方法
+                'size': 4,  # 采样大小
                 'weight': (0.4, 0.3, 0.3),  # 表面采样方法的权重
             }  # 第二步采样
         ]
@@ -208,4 +209,8 @@ class TestTask:
         task_obj.print_task_info()
         task_obj.run()
         task_obj.plot()
+        energies = task_obj.sg_obj.grid_property['energy']
+        np.savetxt('energies.csv', np.column_stack((sg_obj.points, energies)),
+                   header='x, y ,z , energy/eV',
+                   delimiter=',', fmt='%f')
         print('Done')