molcraft 0.1.0a5__tar.gz → 0.1.0a7__tar.gz

{molcraft-0.1.0a5 → molcraft-0.1.0a7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: molcraft
-Version: 0.1.0a5
+Version: 0.1.0a7
 Summary: Graph Neural Networks for Molecular Machine Learning
 Author-email: Alexander Kensert <alexander.kensert@gmail.com>
 License: MIT License
@@ -55,7 +55,7 @@ Dynamic: license-file
 - Customizable and serializable **layers** and **models**
 - Customizable **GraphTensor**
 - Fast and efficient featurization of molecular graphs
-- Efficient and easy-to-use input pipelines using TF **records**
+- Fast and efficient input pipelines using TF **records**
 
 ## Examples 
 
@@ -106,20 +106,7 @@ print(pred)
 # featurizers.save_featurizer(featurizer, '/tmp/featurizer.json')
 # models.save_model(model, '/tmp/model.keras')
 
-# featurizers.load_featurizer('/tmp/featurizer.json')
-# models.load_model('/tmp/model.keras')
+# loaded_featurizer = featurizers.load_featurizer('/tmp/featurizer.json')
+# loaded_model = models.load_model('/tmp/model.keras')
 ```
 
-## Installation
-
-Install the pre-release of molcraft via pip:
-
-```bash
-pip install molcraft --pre
-```
-
-with GPU support:
-
-```bash
-pip install molcraft[gpu] --pre
-```

{molcraft-0.1.0a5 → molcraft-0.1.0a7}/README.md

@@ -11,7 +11,7 @@
 - Customizable and serializable **layers** and **models**
 - Customizable **GraphTensor**
 - Fast and efficient featurization of molecular graphs
-- Efficient and easy-to-use input pipelines using TF **records**
+- Fast and efficient input pipelines using TF **records**
 
 ## Examples 
 
@@ -62,20 +62,7 @@ print(pred)
 # featurizers.save_featurizer(featurizer, '/tmp/featurizer.json')
 # models.save_model(model, '/tmp/model.keras')
 
-# featurizers.load_featurizer('/tmp/featurizer.json')
-# models.load_model('/tmp/model.keras')
+# loaded_featurizer = featurizers.load_featurizer('/tmp/featurizer.json')
+# loaded_model = models.load_model('/tmp/model.keras')
 ```
 
-## Installation
-
-Install the pre-release of molcraft via pip:
-
-```bash
-pip install molcraft --pre
-```
-
-with GPU support:
-
-```bash
-pip install molcraft[gpu] --pre
-```

{molcraft-0.1.0a5 → molcraft-0.1.0a7}/molcraft/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '0.1.0a5'
+__version__ = '0.1.0a7'
 
 import os
 os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
@@ -14,4 +14,5 @@ from molcraft import ops
 from molcraft import records 
 from molcraft import tensors
 from molcraft import callbacks
-from molcraft import datasets
+from molcraft import datasets
+from molcraft import losses

molcraft-0.1.0a7/molcraft/callbacks.py

@@ -0,0 +1,93 @@
+import keras
+import warnings
+import numpy as np
+
+
+class TensorBoard(keras.callbacks.TensorBoard):
+
+    def _log_weights(self, epoch):
+        with self._train_writer.as_default():
+            for layer in self.model.layers:
+                for weight in layer.weights:
+                    # Use weight.path istead of weight.name to distinguish
+                    # weights of different layers.
+                    histogram_weight_name = weight.path + "/histogram"
+                    self.summary.histogram(
+                        histogram_weight_name, weight, step=epoch
+                    )
+                    if self.write_images:
+                        image_weight_name = weight.path + "/image"
+                        self._log_weight_as_image(
+                            weight, image_weight_name, epoch
+                        )
+            self._train_writer.flush()
+
+
+class LearningRateDecay(keras.callbacks.LearningRateScheduler):
+
+    def __init__(self, rate: float, delay: int = 0, **kwargs):
+
+        def lr_schedule(epoch: int, lr: float):
+            if epoch < delay:
+                return float(lr)
+            return float(lr * keras.ops.exp(-rate))
+        
+        super().__init__(schedule=lr_schedule, **kwargs)
+
+
+class Rollback(keras.callbacks.Callback):
+
+    def __init__(
+        self, 
+        frequency: int = None,
+        tolerance: float = 0.5, 
+        rollback_optimizer: bool = True,
+    ):
+        super().__init__()
+        self.frequency = frequency or 1_000_000_000
+        self.tolerance = tolerance
+        self.rollback_optimizer = rollback_optimizer
+
+    def on_train_begin(self, logs=None):
+        self.rollback_weights = self._get_model_vars()
+        self.rollback_optimizer_vars = self._get_optimizer_vars()
+        self.rollback_loss = float('inf')
+
+    def on_epoch_end(self, epoch: int, logs: dict = None):
+        current_loss = logs.get('val_loss', logs.get('loss'))
+        deviation = (current_loss - self.rollback_loss) / self.rollback_loss
+
+        if np.isnan(current_loss) or np.isinf(current_loss):
+            self._rollback()
+            print("\nRolling back model, found nan or inf loss.\n")
+            return 
+        
+        if deviation > self.tolerance:
+            self._rollback()
+            print(f"\nRolling back model, found too large deviation: {deviation:.3f}\n")
+        
+        if epoch and epoch % self.frequency == 0:
+            self._rollback()
+            print(f"\nRolling back model, {epoch} % {self.frequency} == 0\n")
+            return 
+        
+        if current_loss < self.rollback_loss:
+            self._save_state(current_loss)
+
+    def _save_state(self, current_loss: float) -> None:
+        self.rollback_loss = current_loss
+        self.rollback_weights = self._get_model_vars()
+        if self.rollback_optimizer:
+            self.rollback_optimizer_vars = self._get_optimizer_vars()
+
+    def _rollback(self) -> None:
+        self.model.set_weights(self.rollback_weights)
+        if self.rollback_optimizer:
+            self.model.optimizer.set_weights(self.rollback_optimizer_vars)
+
+    def _get_optimizer_vars(self):
+        return [v.numpy() for v in self.model.optimizer.variables]
+    
+    def _get_model_vars(self):
+        return self.model.get_weights()
+    

{molcraft-0.1.0a5 → molcraft-0.1.0a7}/molcraft/chem.py

@@ -11,6 +11,7 @@ from rdkit.Chem import rdMolTransforms
 from rdkit.Chem import rdPartialCharges
 from rdkit.Chem import rdMolDescriptors
 from rdkit.Chem import rdForceFieldHelpers
+from rdkit.Chem import rdFingerprintGenerator
 
 
 class Mol(Chem.Mol):
@@ -399,7 +400,6 @@ def embed_conformers(
     mol: Mol, 
     num_conformers: int, 
     method: str = 'ETKDGv3',
-    force: bool = True,
     **kwargs
 ) -> None:
     available_embedding_methods = {
@@ -410,27 +410,39 @@ def embed_conformers(
         'srETKDGv3': rdDistGeom.srETKDGv3(),
         'KDG': rdDistGeom.KDG()
     }
-    default_embedding_method = 'ETKDGv3'
     mol = Mol(mol)
-    params = available_embedding_methods.get(method)
-    if params is None:
-        warn(
-            f"Could not find `method` {method}. "
-            f"Automatically setting method to {default_embedding_method}."
+    embedding_method = available_embedding_methods.get(method)
+    if embedding_method is None:
+        raise ValueError(
+            f'Could not find `method` {method!r}. Specify either of: '
+            '`ETDG`, `ETKDG`, `ETKDGv2`, `ETKDGv3`, `srETKDGv3` or `KDG`.'
         )
-        params = available_embedding_methods[default_embedding_method]
+
     for key, value in kwargs.items():
-        setattr(params, key, value)
+        setattr(embedding_method, key, value)
 
-    success = rdDistGeom.EmbedMultipleConfs(mol, numConfs=num_conformers, params=params)
+    success = rdDistGeom.EmbedMultipleConfs(
+        mol, numConfs=num_conformers, params=embedding_method
+    )
     if not len(success):
-        warning = 'Could not embed conformer(s).'
-        if not force:
-            warn(warning)
+        warn(
+            f'Could not embed conformer(s) for {mol.canonical_smiles!r} using the '
+            'speified method. Giving it another try with more permissive methods.'
+        )
+        max_attempts = (20 * mol.num_atoms) # increasing it from 10xN to 20xN
+        for fallback_method in [method, 'ETDG', 'KDG']:
+            fallback_embedding_method = available_embedding_methods[fallback_method]
+            fallback_embedding_method.useRandomCoords = True
+            fallback_embedding_method.maxAttempts = max_attempts
+            success = rdDistGeom.EmbedMultipleConfs(
+                mol, numConfs=num_conformers, params=fallback_embedding_method
+            )
+            if len(success):
+                break
         else:
-            solution = ' Embedding a conformer (in 3D space) using (x, y) coordinates.'
-            warn(warning + solution)
-            rdDepictor.Compute2DCoords(mol)
+            raise RuntimeError(
+                f'Could not embed conformer(s) for {mol.canonical_smiles!r}. '
+            )
     return mol
 
 def optimize_conformers(
@@ -444,6 +456,11 @@ def optimize_conformers(
     available_force_field_methods = [
         'MMFF', 'MMFF94', 'MMFF94s', 'UFF'
     ]
+    if method not in available_force_field_methods:
+        raise ValueError(
+            f'Could not find `method` {method!r}. Specify either of: '
+            '`UFF`, `MMFF`, `MMFF94` or `MMFF94s`.'
+        )
     mol = Mol(mol)
     try:
         if method.startswith('MMFF'):
@@ -468,7 +485,7 @@ def optimize_conformers(
     except RuntimeError as e:
         warn(
             f'{method} force field minimization raised {e}. '
-            '\nProceeding without force field minimization...'
+            '\nProceeding without force field minimization.'
         )
     return mol
 
@@ -579,8 +596,7 @@ def _calc_mmff_energies(
             energies.append(float('nan'))
     return energies
 
-
-def _split_mol_by_confs(mol: Mol) -> list[Mol]:
+def unpack_conformers(mol: Mol) -> list[Mol]:
     mols = []
     for conf in mol.get_conformers():
         new_mol = Chem.Mol(mol)
@@ -590,11 +606,77 @@ def _split_mol_by_confs(mol: Mol) -> list[Mol]:
         mols.append(new_mol)
     return mols
 
+_fingerprint_types = {
+    'rdkit': rdFingerprintGenerator.GetRDKitFPGenerator,
+    'morgan': rdFingerprintGenerator.GetMorganGenerator,
+    'topological_torsion': rdFingerprintGenerator.GetTopologicalTorsionGenerator,
+    'atom_pair': rdFingerprintGenerator.GetAtomPairGenerator,
+}
+
+def _get_fingerprint(
+    mol: Mol,
+    fp_type: str = 'morgan',
+    binary: bool = True,
+    dtype: str = 'float32',
+    **kwargs,
+) -> np.ndarray:
+    fingerprint: rdFingerprintGenerator.FingerprintGenerator64 = (
+        _fingerprint_types[fp_type](**kwargs)
+    )
+    if not isinstance(mol, Mol):
+        mol = Mol.from_encoding(mol)
+    if binary:
+        fp: np.ndarray = fingerprint.GetFingerprintAsNumPy(mol)
+    else:
+        fp: np.ndarray = fingerprint.GetCountFingerprintAsNumPy(mol)
+    return fp.astype(dtype)
+
+def _rdkit_fingerprint(
+    mol: Chem.Mol, 
+    size: int = 2048, 
+    *,
+    min_path: int = 1, 
+    max_path: int = 7, 
+    binary: bool = True,
+    dtype: str = 'float32',
+) -> np.ndarray:
+    fp_param = {'fpSize': size, 'minPath': min_path, 'maxPath': max_path}
+    return _get_fingerprint(mol, 'rdkit', binary, dtype, **fp_param)
+
+def _morgan_fingerprint(
+    mol: Chem.Mol, 
+    size: int = 2048, 
+    *,
+    radius: int = 3, 
+    binary: bool = True,
+    dtype: str = 'float32',
+) -> np.ndarray:
+    fp_param = {'radius': radius, 'fpSize': size}
+    return _get_fingerprint(mol, 'morgan', binary, dtype, **fp_param)
+
+def _topological_torsion_fingerprint(
+    mol: Chem.Mol, 
+    size: int = 2048, 
+    *,
+    binary: bool = True,
+    dtype: str = 'float32',
+) -> np.ndarray:
+    fp_param = {'fpSize': size}
+    return _get_fingerprint(mol, 'topological_torsion', binary, dtype, **fp_param)
+
+def _atom_pair_fingerprint(
+    mol: Chem.Mol, 
+    size: int = 2048, 
+    *,
+    binary: bool = True,
+    dtype: str = 'float32',
+) -> np.ndarray:
+    fp_param = {'fpSize': size}
+    return _get_fingerprint(mol, 'atom_pair', binary, dtype, **fp_param)
+
 def warn(message: str) -> None:
     warnings.warn(
         message=message,
         category=UserWarning,
         stacklevel=1,
     )
-
-    

{molcraft-0.1.0a5 → molcraft-0.1.0a7}/molcraft/conformers.py

@@ -23,20 +23,17 @@ class ConformerEmbedder(ConformerProcessor):
     def __init__(
         self, 
         method: str = 'ETKDGv3',
-        num_conformers: int = 10, 
-        force: bool = True,
+        num_conformers: int = 5, 
         **kwargs,
     ) -> None:
         self.method = method 
         self.num_conformers = num_conformers 
-        self.force = force
         self.kwargs = kwargs 
 
     def get_config(self) -> dict:
         config = {
             'method': self.method, 
             'num_conformers': self.num_conformers, 
-            'force': self.force, 
         }
         config.update({
             k: v for (k, v) in self.kwargs.items()
@@ -48,7 +45,6 @@ class ConformerEmbedder(ConformerProcessor):
             mol, 
             method=self.method,
             num_conformers=self.num_conformers,
-            force=self.force,
             **self.kwargs,
         )
 

{molcraft-0.1.0a5 → molcraft-0.1.0a7}/molcraft/featurizers.py

@@ -175,7 +175,7 @@ class MolGraphFeaturizer(Featurizer):
             default_bond_features = (
                 bond_features == 'auto' or bond_features == 'default'
             )
-            if default_bond_features or self.radius > 1 or self.self_loops:
+            if default_bond_features or self.radius > 1:
                 vocab = ['zero', 'single', 'double', 'triple', 'aromatic']
                 bond_features = [
                     features.BondType(vocab)
@@ -215,7 +215,7 @@ class MolGraphFeaturizer(Featurizer):
         if mol is None:
             warn(
                 f'Could not obtain `chem.Mol` from {x}. '
-                'Proceeding without it.'
+                'Returning `None` (proceeding without it).'
             )
             return None
         
@@ -254,24 +254,17 @@ class MolGraphFeaturizer(Featurizer):
 
         node = {}
         node['feature'] = atom_feature
-
-        if bond_feature is not None and (self.radius > 1 or self.self_loops):
-            # Append 'zero order' bond feature encoding, which encodes non-bonds. 
-            zero_bond_feature = np.array(
-                [[1., 0., 0., 0., 0.]], dtype=bond_feature.dtype
-            )
-            bond_feature = np.concatenate(
-                [bond_feature, zero_bond_feature], axis=0
-            )    
         
         edge = {}
         if self.radius == 1:
             edge['source'], edge['target'] = mol.adjacency(
                 fill='full', sparse=True, self_loops=self.self_loops, dtype=self.index_dtype
             )
+            if self.self_loops:
+                bond_feature = np.pad(bond_feature, [(0, 1), (0, 0)])
             if bond_feature is not None:
                 bond_indices = []
-                for (atom_i, atom_j) in zip(edge['source'], edge['target']):
+                for atom_i, atom_j in zip(edge['source'], edge['target']):
                     if atom_i == atom_j:
                         bond_indices.append(-1)
                     else:
@@ -279,6 +272,8 @@ class MolGraphFeaturizer(Featurizer):
                             mol.get_bond_between_atoms(atom_i, atom_j).index
                         )
                 edge['feature'] = bond_feature[bond_indices]
+                if self.self_loops:
+                    edge['self_loop'] = (edge['source'] == edge['target'])
         else:
             paths = chem.get_shortest_paths(
                 mol, radius=self.radius, self_loops=self.self_loops
@@ -293,6 +288,12 @@ class MolGraphFeaturizer(Featurizer):
                 [len(path) - 1 for path in paths], dtype=self.index_dtype
             )
             if bond_feature is not None:
+                zero_bond_feature = np.array(
+                    [[1., 0., 0., 0., 0.]], dtype=bond_feature.dtype
+                )
+                bond_feature = np.concatenate(
+                    [bond_feature, zero_bond_feature], axis=0
+                )    
                 edge['feature'] = self._expand_bond_features(
                     mol, paths, bond_feature,
                 )
@@ -511,7 +512,7 @@ class MolGraphFeaturizer3D(MolGraphFeaturizer):
                 steps=[
                     conformers.ConformerEmbedder(
                         method='ETKDGv3', 
-                        num_conformers=10
+                        num_conformers=5
                     ),
                 ]
             )
@@ -588,7 +589,7 @@ class MolGraphFeaturizer3D(MolGraphFeaturizer):
             edge_feature = self.bond_features(mol)
 
         edge = {}
-        mols = chem._split_mol_by_confs(mol)
+        mols = chem.unpack_conformers(mol)
         tensor_list = []
         for i, mol in enumerate(mols):
             node_conformer = copy.deepcopy(node)
@@ -734,6 +735,11 @@ def _add_super_edges(
             ]
         )
 
+    if 'self_loop' in edge:
+        edge['self_loop'] = np.pad(
+            edge['self_loop'], [(0, num_nodes * num_super_nodes * 2)],
+            constant_values=False,
+        )
     if 'length' in edge:
         edge['length'] = np.pad(edge['length'], [(0, 0), (1, 0)])
         zero_array = np.zeros([num_nodes * num_super_nodes * 2], dtype='int32')