Version 1.0.1

.github/workflows/continuous_integration.yml

@@ -12,22 +12,22 @@ jobs:
   build-linux:
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/checkout@v2
-    - name: Set up Python 3.8
+    - uses: actions/checkout@v3 
+    - name: Set up Python 3.12
       uses: actions/setup-python@v2
       with:
-        python-version: 3.8
+        python-version: 3.12
     - name: Add conda to system path
       run: |
         # $CONDA is an environment variable pointing to the root of the miniconda directory
         echo $CONDA/bin >> $GITHUB_PATH
     - name: Install dependencies
       shell: bash -l {0}
       run: |
-        conda install python=3.8
-        conda install numpy
+        conda install python=3.12
         conda install -c conda-forge gpy
-        conda install pytorch
+        conda install numpy
+        conda install -c conda-forge pytorch=2.1.2
         conda install pandas
         conda install -c conda-forge cclib
         conda install -c conda-forge hyperopt
@@ -48,7 +48,7 @@ jobs:
       matrix:
         cfg:
           - conda-env: docs-cf
-            python-version: 3.8
+            python-version: 3.12
             label: Sphinx
             runs-on: ubuntu-latest
 
@@ -59,13 +59,13 @@ jobs:
     - uses: actions/checkout@v3
 
     - name: Create Environment
-      uses: conda-incubator/setup-miniconda@v2
+      uses: conda-incubator/setup-miniconda@v3
       with:
         activate-environment: test
         environment-file: docs/requirements.yml
         python-version: ${{ matrix.cfg.python-version }}
         auto-activate-base: false
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         use-mamba: true
         add-pip-as-python-dependency: true
@@ -74,4 +74,4 @@ jobs:
     - name: Environment Information
       run: |
         mamba info
-        mamba list --show-channel-urls
+        mamba list

.github/workflows/docs.yml

@@ -20,12 +20,12 @@ jobs:
         uses: actions/checkout@v3
 
       - name: Create Environment
-        uses: conda-incubator/setup-miniconda@v2
+        uses: conda-incubator/setup-miniconda@v3
         with:
           activate-environment: docs-env
           environment-file: docs/requirements.yml
           python-version: ${{ matrix.cfg.python-version }}
-          miniforge-variant: Mambaforge
+          miniforge-variant: Miniforge3
           use-mamba: true
           add-pip-as-python-dependency: true
           channels: conda-forge

docs/requirements.yml

@@ -3,7 +3,7 @@ channels:
     - nodefaults
     - conda-forge
 dependencies:
-    - python=3.8
+    - python=3.12
     - sphinx
     - sphinx_rtd_theme
     - sphinx-book-theme
@@ -13,11 +13,11 @@ dependencies:
     - autodoc-pydantic
 
       # PESLearn depends
+    - GPy >=1.9
     - numpy
     - pydantic >=0.30.1
     - qcelemental >=0.9.0
     - numpy >=1.7
-    - GPy >=1.9
     - scikit-learn >=0.20
     - pandas >=0.24
     - hyperopt >=0.1.1

peslearn/constants.py

@@ -180,3 +180,41 @@ def cart1d_to_distances1d(vec):
     distance_vector = distance_matrix[np.tril_indices(len(distance_matrix),-1)]
     return distance_vector
 """
+
+gradient_nn_convenience_function = """
+# how to use 'gradient_compute()' function
+# --------------------------------------
+# grads = gradient_compute(cartesian_dataset_path)
+# 'cartesian_dataset_path' is a path to a dataset with cartesian coordinates that will be used with model to predict gradients.
+# Cartesian geometries should contain atomic symbol (e.g. H, C, Br, etc.) followed by XYZ cartesian coordinates each separated by spaces.
+# The dataset may contain one or multiple geometries, if given multiple geometries it will return a list of multiple gradients
+# The output of the 'gradient_compute()' function will be the negative derivative of the predicted energy from the model with respect to
+# the cartesian coordinate of the provided geometry. 
+# i.e. ouput = -dE/dq, where output is the returned value from the 'gradient_compute()' function, E is the model predicted energy, and q are the cartesian coordinates
+# Input coordinates need not be in standard order, they will however be transformed into standard order for the output. 
+# Standard order in PES-Learn lists most common atoms first with alphabetical tiebreakers. 
+# e.g. If the input provided for water lists the XYZ coordinates in order of O H1 H2, then the output gradient will be in the order H1 H2 O.
+# The returned gradients will be in units of Hartree/distance where distance is either Bohr or Angstrom. 
+# The distance unit used to construct the model should be the same distance unit used to predict gradients.
+# Outputs are of the form of a list of torch tensors, where each tensor is a predicted gradient in the order provided in the catesian dataset.
+# Using this function for the first time creates a new file containing a function tailored to the given input molecule.
+
+def gradient_compute(cart_dataset_path):
+    grad = []
+    # transform 
+    sorted_atoms, geoms = geometry_transform_helper.load_cartesian_dataset(cart_dataset_path, no_energy=True)
+    for i in range(len(geoms)):
+        geoms[i] = geometry_transform_helper.remove_atom_labels(geoms[i])
+    if not os.path.exists('grad_func.py'):
+        geometry_transform_helper.write_grad_input(sorted_atoms)
+    from grad_func import gradient_prediction
+    for g in range(len(geoms)):
+        grad.append(gradient_prediction(nn, params, model, Xscaler, yscaler, geoms[g]))
+    return grad
+"""
+
+
+
+
+
+

peslearn/datagen/outputfile.py

@@ -24,7 +24,7 @@ def __init__(self, output_path):
             self.output_str = f.read()
 
     def extract_energy_with_regex(self, energy_regex):
-        """
+        r"""
         Finds the energy value (a float) in an output file according to a user supplied
         regular expression identifier.
 
@@ -34,15 +34,15 @@ def extract_energy_with_regex(self, energy_regex):
         FINAL ELECTRONIC ENERGY (Eh):   -2.3564983498
 
         One can obtain this floating point number with the regex identifier:
-        \s*FINAL ELECTRONIC ENERGY \(Eh\):\s+(-\d+\.\d+)
+        r"\s*FINAL ELECTRONIC ENERGY \(Eh\):\s+(-\d+\.\d+)"
 
         Checking ones regular expression is easy with online utilities such as pythex (see pythex.org)
 
         Parameters
         ---------
         energy_regex : str
             A string containing the regex code for capturing an energy floating point number.
-                e.g. "\s*FINAL ELECTRONIC ENERGY \(Eh\):\s+(-\d+\.\d+)"
+                e.g. r"\s*FINAL ELECTRONIC ENERGY \(Eh\):\s+(-\d+\.\d+)"
 
         Returns
         -------
@@ -126,11 +126,11 @@ def extract_from_schema(self, driver):
         """
         if driver == "energy":
             energy = []
-            energy = re.findall("\s\'return_energy\'\:\s+(-\d+\.\d+)", self.output_str)
+            energy = re.findall(r"\s\'return_energy\'\:\s+(-\d+\.\d+)", self.output_str)
             if energy:
                 return energy
             else:
-                success = re.findall("\s\'success\'\:\s+(\S+)\}", self.output_str)
+                success = re.findall(r"\s\'success\'\:\s+(\S+)\}", self.output_str)
                 if success[0] == 'False':
                     energy = 'False'
                     return energy
@@ -140,7 +140,7 @@ def extract_from_schema(self, driver):
                     return energy
 
         if driver == "gradient":
-            gradient = re.findall("\s\'return_gradient\'\:\s+array\(([\s\S]*?)\)\,", self.output_str)
+            gradient = re.findall(r"\s\'return_gradient\'\:\s+array\(([\s\S]*?)\)\,", self.output_str)
             if gradient:
                 import ast
                 gradient = re.sub(r'\s+', "", str(gradient))
@@ -150,7 +150,7 @@ def extract_from_schema(self, driver):
                 gradient = np.asarray(ast.literal_eval(gradient)).astype(np.float64)
                 return gradient
             else:
-                success = re.findall("\s\'success\'\:\s+(\S+)\}", self.output_str)
+                success = re.findall(r"\s\'success\'\:\s+(\S+)\}", self.output_str)
                 if success[0] == 'False':
                     gradient = 'False'
                     return gradient
@@ -160,7 +160,7 @@ def extract_from_schema(self, driver):
 
 
         if driver == "hessian":
-            hessian = re.findall("\s\'return_hessian\'\:\s+array\(([\s\S]*?)\)\,", self.output_str)
+            hessian = re.findall(r"\s\'return_hessian\'\:\s+array\(([\s\S]*?)\)\,", self.output_str)
             if hessian:
                 import ast
                 hessian = re.sub(r'\s+', "", str(hessian))
@@ -170,7 +170,7 @@ def extract_from_schema(self, driver):
                 hessian = np.asarray(ast.literal_eval(hessian)).astype(np.float64)
                 return hessian
             else:
-                success = re.findall("\s\'success\'\:\s+(\S+)\}", self.output_str)
+                success = re.findall(r"\s\'success\'\:\s+(\S+)\}", self.output_str)
                 if success[0] == 'False':
                     hessian = 'False'
                     return hessian
@@ -185,7 +185,7 @@ def extract_from_schema(self, driver):
             return properties
 
     def extract_cartesian_gradient_with_regex(self, header, footer, grad_line_regex):
-        """
+        r"""
         Extracts cartesian gradients according to user supplied regular expressions.
         A bit more tedious to use than the energy regex extractor as the size of the regular expressions may be quite long.
         Requires that the electronic structure theory code prints the cartesian gradient in a logical way.
@@ -214,7 +214,7 @@ def extract_cartesian_gradient_with_regex(self, header, footer, grad_line_regex)
             Atom 1 Cl 0.00000 0.23410 0.32398 
             Atom 2 H  0.02101 0.09233 0.01342     
             Atom 3 N  0.01531 0.04813 0.06118
-            A valid argument for grad_line_regex would be "Atom\s+\d+\s+[A-Z,a-z]+\s+(-?\d+\.\d+)\s+(-?\d+\.\d+)\s+(-?\d+\.\d+)"
+            A valid argument for grad_line_regex would be r"Atom\s+\d+\s+[A-Z,a-z]+\s+(-?\d+\.\d+)\s+(-?\d+\.\d+)\s+(-?\d+\.\d+)"
             This can easily be tested with online utilities such as pythex (see pythex.org)
         Returns
         -------

peslearn/input_processor.py

@@ -17,7 +17,7 @@ class InputProcessor(object):
     """
     def __init__(self, input_string):
         # Remove all comments denoted by '#'
-        self.full_string = re.sub('\s*#.+', '', input_string) 
+        self.full_string = re.sub(r'\s*#.+', '', input_string) 
         if re.search(regex.intcoords_regex, self.full_string):
             self.zmat_string = re.findall(regex.intcoords_regex, self.full_string)[0] 
         self.intcos_ranges = None 
@@ -81,7 +81,7 @@ def get_keywords(self):
                             }
 
         for k in string_keywords:
-            match = re.search(k+"\s*=\s*(.+)", self.full_string)
+            match = re.search(k + r"\s*=\s*(.+)", self.full_string)
             # if the keyword is mentioned
             if match:
                 value = str(match.group(1))
@@ -90,7 +90,7 @@ def get_keywords(self):
                 if k not in regex_keywords:
                     value = value.lower().strip()
                 # if keyword is raw text, add quotes so it is a string
-                if re.match("[a-z\_]+", value):
+                if re.match(r"[a-z\_]+", value):
                     if (r"'" or r'"') not in value:
                         value = "".join((r'"',value,r'"',))
                 try:
@@ -114,19 +114,19 @@ def extract_intcos_ranges(self):
         # for every geometry label look for its range identifer, e.g. R1 = [0.5, 1.2, 25]
         for label in geomlabels:
             # check to make sure parameter isn't defined more than once
-            if len(re.findall("\W" + label+"\s*=\s*", self.full_string)) > 1:
+            if len(re.findall(r"\W" + label + r"\s*=\s*", self.full_string)) > 1:
                 raise Exception("Parameter {} defined more than once.".format(label))
 
             # if geom parameter has a geometry range, save it
-            match = re.search(label+"\s*=\s*(\[.+\])", self.full_string)
+            match = re.search(label + r"\s*=\s*(\[.+\])", self.full_string)
             if match:
                 try:
                     ranges[label] = ast.literal_eval(match.group(1))
                 except: 
                     raise Exception("Something wrong with definition of parameter {} in input. Should be of the form [start, stop, # of points] or a fixed value".format(label))
             # if it has a fixed value, save it
             else:
-                match = re.search(label+"\s*=\s*(-?\d+\.?\d*)", self.full_string)
+                match = re.search(label + r"\s*=\s*(-?\d+\.?\d*)", self.full_string)
                 if not match:
                     raise Exception("\nDefinition of parameter {} not found in geometry input.      \
                                    \nThe definition is either missing or improperly formatted".format(label))

peslearn/lib/A2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2C/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2C2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2C2D/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2C2D2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2C2DE/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2CD/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2CDE/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2B2CDEF/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2BC/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2BCD/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2BCDE/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2BCDEF/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A2BCDEFG/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B2C/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B2C2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B2C2D/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B2CD/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B2CDE/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B3/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B3C/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3B3CD/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3BC/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3BCD/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3BCDE/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A3BCDEF/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4B/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4B2/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4B2C/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4B2CD/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4B3C/__init__.py

@@ -0,0 +1 @@
+from . import output

peslearn/lib/A4BC/__init__.py

@@ -0,0 +1 @@
+from . import output