update to v0.2.1

Author: qzhu2017

docs/conf.py

@@ -23,9 +23,9 @@
 author = 'Qiang Zhu, Howard Yanxon, David Zagaceta'
 
 # The short X.Y version
-version = '0.2.0'
+version = '0.2.1'
 # The full version, including alpha/beta/rc tags
-release = 'Feb 2022'
+release = 'May 2022'
 
 
 # -- General configuration ---------------------------------------------------

docs/index.rst

@@ -2,7 +2,7 @@ PyXtal_FF
 ======================================
 The aim of PyXtalFF project is to provide an automated computational infrastructure to train the interatomic potentials for inorganic periodic systems from high-end quantum mechanical calculations.
 
-The current version is ``0.2.0`` at `GitHub <https://github.com/qzhu2017/pyxtal_ff>`_. 
+The current version is ``0.2.1`` at `GitHub <https://github.com/qzhu2017/pyxtal_ff>`_. 
 
 Expect updates upon request by `Qiang Zhu\'s group <https://qzhu2017.github.io>`_ at University of Nevada Las Vegas.
 

pyxtal_ff/models/neuralnetwork.py

@@ -616,31 +616,32 @@ def calculate_properties(self, descriptor, bforce=True, bstress=False):
             scale = (1 - 0) / (_drange[:, 1] - _drange[:, 0])
 
             e = np.where(np.array(descriptor['elements'])==element)[0]
-            ee0 = np.where(descriptor['seq'][:,0]==e[0])[0][0]
-            ee1 = np.where(descriptor['seq'][:,0]==e[-1])[0][-1]
+            if len(e) > 0:
+                ee0 = np.where(descriptor['seq'][:,0]==e[0])[0][0]
+                ee1 = np.where(descriptor['seq'][:,0]==e[-1])[0][-1]
 
-            i_size = list(descriptor['elements']).count(element) #number of atoms in type i
-            m_size = ee1-ee0+1# number of pairs
+                i_size = list(descriptor['elements']).count(element) #number of atoms in type i
+                m_size = ee1-ee0+1# number of pairs
 
-            d['x'][element] = torch.zeros([i_size, no_of_descriptors])
-            d['dxdr'][element] = torch.zeros([m_size, no_of_descriptors, 3], dtype=torch.float64) ####
-            d['rdxdr'][element] = torch.zeros([i_size, no_of_descriptors, 6])
+                d['x'][element] = torch.zeros([i_size, no_of_descriptors])
+                d['dxdr'][element] = torch.zeros([m_size, no_of_descriptors, 3], dtype=torch.float64) ####
+                d['rdxdr'][element] = torch.zeros([i_size, no_of_descriptors, 6])
 
-            if e.size > 0:
-                des = 0 + np.einsum('j,ij->ij', scale, (descriptor['x'][e[0]:e[-1]+1] - np.expand_dims(_drange[:, 0], 0)))
-                desp = np.einsum('j,ijk->ijk', scale, descriptor['dxdr'][ee0:ee1+1])
-                dess = np.einsum('j,ijk->ijk', scale, descriptor['rdxdr'][e[0]:e[-1]+1])
-                
-                d['x'][element] += torch.from_numpy(des)
-                d['seq'][element] = deepcopy(descriptor['seq'][ee0:ee1+1])
+                if e.size > 0:
+                    des = 0 + np.einsum('j,ij->ij', scale, (descriptor['x'][e[0]:e[-1]+1] - np.expand_dims(_drange[:, 0], 0)))
+                    desp = np.einsum('j,ijk->ijk', scale, descriptor['dxdr'][ee0:ee1+1])
+                    dess = np.einsum('j,ijk->ijk', scale, descriptor['rdxdr'][e[0]:e[-1]+1])
+                    
+                    d['x'][element] += torch.from_numpy(des)
+                    d['seq'][element] = deepcopy(descriptor['seq'][ee0:ee1+1])
 
-                if self.unit == 'eV':  
-                    d['dxdr'][element] += torch.from_numpy(desp)
-                    d['rdxdr'][element] += torch.from_numpy(dess)
-                else:
-                    d['dxdr'][element] += torch.from_numpy(0.529177 * desp)
-                    d['rdxdr'][element] += torch.from_numpy(0.529177**3*desp)
-            d['seq'][element][:, 0] -= min(d['seq'][element][:, 0]) #adjust the initial position
+                    if self.unit == 'eV':  
+                        d['dxdr'][element] += torch.from_numpy(desp)
+                        d['rdxdr'][element] += torch.from_numpy(dess)
+                    else:
+                        d['dxdr'][element] += torch.from_numpy(0.529177 * desp)
+                        d['rdxdr'][element] += torch.from_numpy(0.529177**3*desp)
+                d['seq'][element][:, 0] -= min(d['seq'][element][:, 0]) #adjust the initial position
 
         x = d['x']
         if bforce:

setup.py

@@ -39,7 +39,7 @@
         'scipy>=1.6.1',
         'matplotlib>=2.0.0',
         'ase>=3.20.1',
-        'torch>=1.1.0',
+        'torch>=1.10.2',
         'phonopy==2.3.2', #tmp fix
         'spglib>=1.12.1',
         'monty>=3.0.2',