MultiOptPy 1.20.6__py3-none-any.whl → 1.20.8__py3-none-any.whl

multioptpy/Potential/keep_potential.py

@@ -1,8 +1,15 @@
-
 from multioptpy.Parameters.parameter import UnitValueLib
 import torch
-  
+
 class StructKeepPotential:
+    """
+    Class for calculating harmonic distance potential between two atoms.
+    
+    Robustness Improvements:
+    - Uses robust norm calculation (sqrt(sum^2 + eps)) to avoid NaN gradients 
+      if atoms perfectly overlap (r=0).
+    - Automatically adapts to the device and dtype of the input geometry.
+    """
     def __init__(self, **kwarg):
         self.config = kwarg
         UVL = UnitValueLib()
@@ -19,14 +26,48 @@ class StructKeepPotential:
         bias_pot_params[0] : keep_pot_spring_const
         bias_pot_params[1] : keep_pot_distance
         """
-        vector = torch.linalg.norm((geom_num_list[self.config["keep_pot_atom_pairs"][0]-1] - geom_num_list[self.config["keep_pot_atom_pairs"][1]-1]), ord=2)
+        # 1. Parameter Retrieval & Device/Dtype handling
+        device = geom_num_list.device if isinstance(geom_num_list, torch.Tensor) else torch.device("cpu")
+        dtype = geom_num_list.dtype
+
         if len(bias_pot_params) == 0:
-            energy = 0.5 * self.config["keep_pot_spring_const"] * (vector - self.config["keep_pot_distance"]/self.bohr2angstroms) ** 2
+            k = self.config["keep_pot_spring_const"]
+            r0_ang = self.config["keep_pot_distance"]
         else:
-            energy = 0.5 * bias_pot_params[0] * (vector - bias_pot_params[1]/self.bohr2angstroms) ** 2
-        return energy #hartree
-    
+            k = bias_pot_params[0]
+            r0_ang = bias_pot_params[1]
+
+        # Convert r0 to Bohr (assuming parameter is in Angstroms) and ensure tensor
+        if not isinstance(r0_ang, torch.Tensor):
+            r0_ang = torch.tensor(r0_ang, device=device, dtype=dtype)
+        r0 = r0_ang / self.bohr2angstroms
+
+        # 2. Vector Calculation
+        # Indices are 1-based in config
+        idx1 = self.config["keep_pot_atom_pairs"][0] - 1
+        idx2 = self.config["keep_pot_atom_pairs"][1] - 1
+        
+        vec = geom_num_list[idx1] - geom_num_list[idx2]
+        
+        # 3. Robust Distance Calculation
+     
+        dist = torch.clamp(torch.sqrt(torch.sum(vec**2)), min=1e-12)
+        
+        # 4. Energy Calculation
+        # E = 0.5 * k * (r - r0)^2
+        energy = 0.5 * k * (dist - r0) ** 2
+        
+        return energy # hartree
+
+
 class StructKeepPotentialv2:
+    """
+    Class for calculating harmonic distance potential between two fragment centers.
+    
+    Robustness Improvements:
+    - Uses robust norm calculation.
+    - vectorized center of mass calculation.
+    """
     def __init__(self, **kwarg):
         self.config = kwarg
         UVL = UnitValueLib()
@@ -34,31 +75,55 @@ class StructKeepPotentialv2:
         self.bohr2angstroms = UVL.bohr2angstroms 
         self.hartree2kjmol = UVL.hartree2kjmol 
         return
+        
     def calc_energy(self, geom_num_list, bias_pot_params=[]):
         """
         # required variables: self.config["keep_pot_v2_spring_const"], 
                              self.config["keep_pot_v2_distance"], 
                              self.config["keep_pot_v2_fragm1"],
                              self.config["keep_pot_v2_fragm2"],
-        bias_pot_params[0] : keep_pot_v2_spring_const
-        bias_pot_params[1] : keep_pot_v2_distance    
         """
-        fragm_1_indices = torch.tensor(self.config["keep_pot_v2_fragm1"]) - 1
-        fragm_2_indices = torch.tensor(self.config["keep_pot_v2_fragm2"]) - 1
+        device = geom_num_list.device if isinstance(geom_num_list, torch.Tensor) else torch.device("cpu")
+        dtype = geom_num_list.dtype
 
-        fragm_1_center = torch.mean(geom_num_list[fragm_1_indices], dim=0)
-        fragm_2_center = torch.mean(geom_num_list[fragm_2_indices], dim=0)
-        
-        distance = torch.linalg.norm(fragm_1_center - fragm_2_center, ord=2)
+        # 1. Parameter Retrieval
         if len(bias_pot_params) == 0:
-            energy = 0.5 * self.config["keep_pot_v2_spring_const"] * (distance - self.config["keep_pot_v2_distance"]/self.bohr2angstroms) ** 2
-        
+            k = self.config["keep_pot_v2_spring_const"]
+            r0_ang = self.config["keep_pot_v2_distance"]
         else:
-            energy = 0.5 * bias_pot_params[0] * (distance - bias_pot_params[1]/self.bohr2angstroms) ** 2
+            k = bias_pot_params[0]
+            r0_ang = bias_pot_params[1]
+
+        if not isinstance(r0_ang, torch.Tensor):
+            r0_ang = torch.tensor(r0_ang, device=device, dtype=dtype)
+        r0 = r0_ang / self.bohr2angstroms
+
+        # 2. Center Calculation (Vectorized)
+        # Convert indices lists to tensors for efficient indexing
+        idx1 = torch.tensor(self.config["keep_pot_v2_fragm1"], device=device, dtype=torch.long) - 1
+        idx2 = torch.tensor(self.config["keep_pot_v2_fragm2"], device=device, dtype=torch.long) - 1
+
+        fragm_1_center = torch.mean(geom_num_list[idx1], dim=0)
+        fragm_2_center = torch.mean(geom_num_list[idx2], dim=0)
         
-        return energy #hartree
-    
+        # 3. Robust Distance Calculation
+        vec = fragm_1_center - fragm_2_center
+        distance = torch.clamp(torch.sqrt(torch.sum(vec**2)), min=1e-12)
+        
+        # 4. Energy
+        energy = 0.5 * k * (distance - r0) ** 2
+        
+        return energy # hartree
+
+
 class StructKeepPotentialAniso:
+    """
+    Class for calculating anisotropic distance potential between fragment centers.
+    
+    Robustness Improvements:
+    - Vectorized centroid calculation (replaced inefficient loops).
+    - Device/Dtype safety for matrix operations.
+    """
     def __init__(self, **kwarg):
         self.config = kwarg
         UVL = UnitValueLib()
@@ -66,34 +131,48 @@ class StructKeepPotentialAniso:
         self.bohr2angstroms = UVL.bohr2angstroms 
         self.hartree2kjmol = UVL.hartree2kjmol 
         return
+        
     def calc_energy(self, geom_num_list, bias_pot_params=[]):
         """
         # required variables:   self.config["aniso_keep_pot_v2_spring_const_mat"]
                                 self.config["aniso_keep_pot_v2_dist"] 
                                 self.config["aniso_keep_pot_v2_fragm1"]
                                 self.config["aniso_keep_pot_v2_fragm2"]
-                             
         """
-        fragm_1_center = torch.tensor([0.0, 0.0, 0.0], dtype=torch.float64, requires_grad=True)
-        for i in self.config["aniso_keep_pot_v2_fragm1"]:
-            fragm_1_center = fragm_1_center + geom_num_list[i-1]
+        device = geom_num_list.device if isinstance(geom_num_list, torch.Tensor) else torch.device("cpu")
+        dtype = geom_num_list.dtype
+
+        # 1. Centroid Calculation (Vectorized)
+        idx1 = torch.tensor(self.config["aniso_keep_pot_v2_fragm1"], device=device, dtype=torch.long) - 1
+        idx2 = torch.tensor(self.config["aniso_keep_pot_v2_fragm2"], device=device, dtype=torch.long) - 1
         
-        fragm_1_center = fragm_1_center / len(self.config["aniso_keep_pot_v2_fragm1"])
+        fragm_1_center = torch.mean(geom_num_list[idx1], dim=0)
+        fragm_2_center = torch.mean(geom_num_list[idx2], dim=0)
         
-        fragm_2_center = torch.tensor([0.0, 0.0, 0.0], dtype=torch.float64, requires_grad=True)
-        for i in self.config["aniso_keep_pot_v2_fragm2"]:
-            fragm_2_center = fragm_2_center + geom_num_list[i-1]
+        # 2. Anisotropic Distance Components
+        # Distance diff vector: |x1-x2|, |y1-y2|, |z1-z2|
+        # Note: Derivative of abs(x) at 0 is 0 in PyTorch subgradient, usually safe enough here.
+        xyz_dist = torch.abs(fragm_1_center - fragm_2_center)
         
-        fragm_2_center = fragm_2_center / len(self.config["aniso_keep_pot_v2_fragm2"])     
-        x_dist = torch.abs(fragm_1_center[0] - fragm_2_center[0])
-        y_dist = torch.abs(fragm_1_center[1] - fragm_2_center[1])
-        z_dist = torch.abs(fragm_1_center[2] - fragm_2_center[2])
-        eq_dist = self.config["aniso_keep_pot_v2_dist"]  / (3 ** 0.5) / self.bohr2angstroms
-        dist_vec = torch.stack([(x_dist - eq_dist) ** 2,(y_dist - eq_dist) ** 2,(z_dist - eq_dist) ** 2])
-        dist_vec = torch.reshape(dist_vec, (3, 1))
-        vec_pot = torch.matmul(torch.tensor(self.config["aniso_keep_pot_v2_spring_const_mat"], dtype=torch.float32), dist_vec)
+        # Equilibrium distance logic (from original code)
+        # Original: eq_dist = dist / sqrt(3)
+        # This implies the target distance is distributed equally among x,y,z components?
+        ref_dist_val = self.config["aniso_keep_pot_v2_dist"]
+        if not isinstance(ref_dist_val, torch.Tensor):
+            ref_dist_val = torch.tensor(ref_dist_val, device=device, dtype=dtype)
+            
+        eq_dist_component = (ref_dist_val / (3 ** 0.5)) / self.bohr2angstroms
         
-        energy = torch.sum(vec_pot)
+        # Squared deviation vector: [(dx - d0)^2, (dy - d0)^2, (dz - d0)^2]
+        dist_vec = (xyz_dist - eq_dist_component) ** 2
+        dist_vec = dist_vec.unsqueeze(1) # Shape (3, 1)
+        
+        # 3. Matrix Multiplication
+        # Ensure spring constant matrix matches device/dtype
+        spring_mat = torch.tensor(self.config["aniso_keep_pot_v2_spring_const_mat"], device=device, dtype=dtype)
         
+        vec_pot = torch.matmul(spring_mat, dist_vec)
+        
+        energy = torch.sum(vec_pot)
         
-        return energy #hartree
+        return energy # hartree

{multioptpy-1.20.6.dist-info → multioptpy-1.20.8.dist-info}/METADATA

@@ -1,21 +1,21 @@
 Metadata-Version: 2.4
 Name: MultiOptPy
-Version: 1.20.6
+Version: 1.20.8
 Summary: Multifunctional geometry optimization tools for quantum chemical calculations.
 Author-email: ss0832 <highlighty876@gmail.com>
 License-Expression: GPL-3.0-or-later
 Requires-Python: >=3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: numpy>=2.2.0
-Requires-Dist: scipy>=1.13.0
-Requires-Dist: matplotlib>=3.10.0
+Requires-Dist: numpy~=2.2.0
+Requires-Dist: scipy~=1.13.0
+Requires-Dist: matplotlib~=3.10.0
 Requires-Dist: torch~=2.6.0
-Requires-Dist: pyscf>=2.9.0
-Requires-Dist: tblite>=0.4.0
+Requires-Dist: pyscf~=2.9.0
+Requires-Dist: tblite~=0.4.0
 Requires-Dist: ase~=3.26.0
 Requires-Dist: fairchem-core~=2.7.0
-Requires-Dist: sympy>=1.13.0
+Requires-Dist: sympy~=1.13.0
 Dynamic: license-file
 
 # MultiOptPy

{multioptpy-1.20.6.dist-info → multioptpy-1.20.8.dist-info}/RECORD

@@ -75,10 +75,11 @@ multioptpy/ModelFunction/opt_mesx.py,sha256=IiUzigFlZqn3ZDjbx4mHTaUEJqUEjXFmsG1l
 multioptpy/ModelFunction/opt_mesx_2.py,sha256=J1lr7LxcgVeAfR0WkvJg4Nt5VTaYPC2ITbP2RL7CwwQ,2530
 multioptpy/ModelFunction/seam_model_function.py,sha256=epOjUcrFpwqLXUYedJeVE5BDl1b0_2vgIFUepL_JjkI,1136
 multioptpy/ModelHessian/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-multioptpy/ModelHessian/approx_hessian.py,sha256=efF8FA7lqBuw8h4DmkfU7T78yZiS5tQ_FhMm0CAseYE,6717
+multioptpy/ModelHessian/approx_hessian.py,sha256=0_fE9-HOCL9l2lV3aAXusOPaYVil0wWWFDPNA9mR23c,6982
 multioptpy/ModelHessian/calc_params.py,sha256=aWkXqqyYpg5_qemflM2oIkFA8g-h3wTiE_C_a5dkLxc,7645
 multioptpy/ModelHessian/fischer.py,sha256=pdXmOMGF4Ie8cA3qN-KuyhJl_cuetvmpoI9t0fE3tNE,11011
 multioptpy/ModelHessian/fischerd3.py,sha256=QWS-vWkHEzPLiMl8bAhd5A9BV1fKj8ym7GVVdZ8fg5M,13450
+multioptpy/ModelHessian/fischerd3old.py,sha256=55JNPGCKZkk5zu3jLxn7hDDOk1_ldOxWCW_Se8i1WUc,17041
 multioptpy/ModelHessian/fischerd4.py,sha256=GuSwCjzMV8g6BHvX0DJjHr_8AfjVdrfNn5DO_BEdDj4,17070
 multioptpy/ModelHessian/gfn0xtb.py,sha256=xUipDBP_o9Y9bJ74RehCY06kroueStc0MsSm1llg638,27051
 multioptpy/ModelHessian/gfnff.py,sha256=INsTLJ2c7wEqk0mxn_lURsQ2JsT8dM6JClR2etEein8,31162
@@ -197,10 +198,10 @@ multioptpy/Potential/electrostatic_potential.py,sha256=13dPbQZnDB_-utsZ2pnMImof9
 multioptpy/Potential/flux_potential.py,sha256=RAZobWEHjRa_S1zSH_RberX5zdZjYspR7rF4uQ7HdXI,1221
 multioptpy/Potential/gaussian_potential.py,sha256=mvuALiAOvtXt5dGGuZq1gJGBhhtkuDNJWyMUW5M5gPY,5636
 multioptpy/Potential/idpp.py,sha256=RlSpElPJ2kyHJRHNsY5zgILaES3d4xAO_50rz0yBE0Q,21690
-multioptpy/Potential/keep_angle_potential.py,sha256=-8P3ZFtzR1axa-j1QHf4MNTZGfrshQqhWzIUShTotho,8791
-multioptpy/Potential/keep_dihedral_angle_potential.py,sha256=HUZL2A8e9zoXDvZDn7rXZZ2K-3VwtZa7s6-dE79CsZA,5951
-multioptpy/Potential/keep_outofplain_angle_potential.py,sha256=CX6R2R_8HcQiq2lONb42yclw5dWGmM3fmQQ45ZQnMAI,3870
-multioptpy/Potential/keep_potential.py,sha256=osK1VI43jHkpTjXslZ8nubeME5LcWTGIbbf91bTTlEw,4827
+multioptpy/Potential/keep_angle_potential.py,sha256=3R2DJkRU88QSxuBCxpiIhI9dN-WYt_lTh1mjldZR3Z0,26229
+multioptpy/Potential/keep_dihedral_angle_potential.py,sha256=vjdPP79eokXS4n7l3x46BCivcqCgGdxI5eBHV1u4mVA,13244
+multioptpy/Potential/keep_outofplain_angle_potential.py,sha256=P1YABikwxwx6v72aaw3s9XwiN2gCYK_sePFnrp7AtHw,12057
+multioptpy/Potential/keep_potential.py,sha256=KHqEA08mk5d1UylhJgYLFOHUQnLh_Z8XPip7nGk8YHM,7413
 multioptpy/Potential/mechano_force_potential.py,sha256=wf18Zl_ogffBD53z3G6u4c4KD3RXmbUkVDQTu6Ny4oY,3394
 multioptpy/Potential/nanoreactor_potential.py,sha256=UIN0dmgyM67TKcCew42rWBWd_cs1XQjxR4WN5gP3K3w,3133
 multioptpy/Potential/potential.py,sha256=WVKrXoWfawsR-gLvDEQvzgzettw6Q5Lz8_344c6b4Hs,52771
@@ -243,9 +244,9 @@ multioptpy/Wrapper/ieip_wrapper.py,sha256=H0Fa7x5mzyFu5rnig7GdcZGX9LFclzRxDuREX5
 multioptpy/Wrapper/md_wrapper.py,sha256=EGQKY8BwBbjWkXxKQGlelzht872OsUp9GmyxPDgWav4,3050
 multioptpy/Wrapper/neb_wrapper.py,sha256=ZOyHhuSDKA5hABnQB7PYPfbQ9OyaAj9KKBeKu_HLiY4,3254
 multioptpy/Wrapper/optimize_wrapper.py,sha256=b5DoiSb5Y4YjzYaSfxNa4A5QMn7_UzAvActPOpL78Ok,2667
-multioptpy-1.20.6.dist-info/licenses/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
-multioptpy-1.20.6.dist-info/METADATA,sha256=wPARwptr8gTPYvTEG3kamkh_rOecyhJbNndyInlmngc,16713
-multioptpy-1.20.6.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
-multioptpy-1.20.6.dist-info/entry_points.txt,sha256=u2Y0tgYMu7UyZumhLCfoME92lCQXhurAAsixL_B3caw,404
-multioptpy-1.20.6.dist-info/top_level.txt,sha256=_MXQNcS1xJbRbGnYUM-F_G-PlLK7wTIkuqx-EvfAhRc,11
-multioptpy-1.20.6.dist-info/RECORD,,
+multioptpy-1.20.8.dist-info/licenses/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
+multioptpy-1.20.8.dist-info/METADATA,sha256=sT6ygFbX30OuibZcZQTN3Ov7-rrF2Kq02QF3A6PxxkE,16713
+multioptpy-1.20.8.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
+multioptpy-1.20.8.dist-info/entry_points.txt,sha256=u2Y0tgYMu7UyZumhLCfoME92lCQXhurAAsixL_B3caw,404
+multioptpy-1.20.8.dist-info/top_level.txt,sha256=_MXQNcS1xJbRbGnYUM-F_G-PlLK7wTIkuqx-EvfAhRc,11
+multioptpy-1.20.8.dist-info/RECORD,,