chore: dependency updates, performance optimizations, and bug fixes

.gitignore

@@ -10,3 +10,4 @@ tmp/
 **/checkpoints/
 *.png
 **/.DS_Store
+*.egg-info

molscribe/augment.py

@@ -1,12 +1,11 @@
 import albumentations as A
 from albumentations.augmentations.geometric.functional import safe_rotate_enlarged_img_size, _maybe_process_in_chunks, \
-                                                              keypoint_rotate
+    keypoint_rotate
 import cv2
 import math
 import random
 import numpy as np
 
-
 def safe_rotate(
     img: np.ndarray,
     angle: int = 0,

molscribe/chemistry.py

@@ -4,10 +4,11 @@
 import multiprocessing
 import itertools
 
-import rdkit
+import rdkit.RDLogger as RDLogger
 import rdkit.Chem as Chem
+import rdkit.Chem.AllChem as AllChem
 
-rdkit.RDLogger.DisableLog('rdApp.*')
+RDLogger.DisableLog('rdApp.*')
 
 from SmilesPE.pretokenizer import atomwise_tokenizer
 
@@ -366,7 +367,11 @@ def get_smiles_from_symbol(symbol, mol, atom, bonds):
 
     total_bonds = int(sum([bond.GetBondTypeAsDouble() for bond in bonds]))
     formula_list = _expand_carbon(_parse_formula(symbol))
-    smiles, bonds_left, num_trails, success = _condensed_formula_list_to_smiles(formula_list, total_bonds, None)
+    # smiles, bonds_left, num_trails, success = _condensed_formula_list_to_smiles(formula_list, total_bonds, None)
+    if len(bonds) != 2:
+        smiles, bonds_left, num_trails, success = _condensed_formula_list_to_smiles(formula_list, total_bonds, None)
+    else:
+        smiles, bonds_left, num_trails, success = _condensed_formula_list_to_smiles(formula_list, 1, None)
     if success:
         return smiles
     return None
@@ -448,39 +453,61 @@ def _need_expand(mol, mappings):
                     atom.SetIsotope(0)
                     continue
 
-                # remove bonds connected to abbreviation/condensed formula
-                adjacent_indices = [bond.GetOtherAtomIdx(i) for bond in bonds]
-                for adjacent_idx in adjacent_indices:
-                    mol_w.RemoveBond(i, adjacent_idx)
-
-                adjacent_atoms = [mol_w.GetAtomWithIdx(adjacent_idx) for adjacent_idx in adjacent_indices]
-                for adjacent_atom, bond in zip(adjacent_atoms, bonds):
-                    adjacent_atom.SetNumRadicalElectrons(int(bond.GetBondTypeAsDouble()))
-
-                # get indices of atoms of main body that connect to substituent
-                bonding_atoms_w = adjacent_indices
-                # assume indices are concated after combine mol_w and mol_r
-                bonding_atoms_r = [mol_w.GetNumAtoms()]
-                for atm in mol_r.GetAtoms():
-                    if atm.GetNumRadicalElectrons() and atm.GetIdx() > 0:
-                        bonding_atoms_r.append(mol_w.GetNumAtoms() + atm.GetIdx())
-
-                # combine main body and substituent into a single molecule object
-                combo = Chem.CombineMols(mol_w, mol_r)
-
-                # connect substituent to main body with bonds
-                mol_w = Chem.RWMol(combo)
-                # if len(bonding_atoms_r) == 1:  # substituent uses one atom to bond to main body
-                for atm in bonding_atoms_w:
-                    bond_order = mol_w.GetAtomWithIdx(atm).GetNumRadicalElectrons()
-                    mol_w.AddBond(atm, bonding_atoms_r[0], order=BOND_TYPES[bond_order])
-
-                # reset radical electrons
-                for atm in bonding_atoms_w:
-                    mol_w.GetAtomWithIdx(atm).SetNumRadicalElectrons(0)
-                for atm in bonding_atoms_r:
-                    mol_w.GetAtomWithIdx(atm).SetNumRadicalElectrons(0)
-                atoms_to_remove.append(i)
+                if "(" in symbol and len(bonds) == 2:
+                    # Get connection information for the current atom
+                    connected_info = [(neighbor.GetIdx(), mol_w.GetBondBetweenAtoms(i, neighbor.GetIdx()).GetBondType())
+                                      for neighbor in atom.GetNeighbors()]
+
+                    # Create a new molecule by combining the current molecule with the expanded fragment
+                    combined = Chem.RWMol(AllChem.CombineMols(mol_w, mol_r))
+
+                    # Connect the first atom of the expanded fragment to the first neighbor
+                    combined.AddBond(connected_info[0][0], mol_w.GetNumAtoms(), connected_info[0][1])
+
+                    # Connect the last atom of the expanded fragment to the second neighbor
+                    combined.AddBond(connected_info[1][0], mol_w.GetNumAtoms() + mol_r.GetNumAtoms() - 1,
+                                     connected_info[1][1])
+
+                    # Update the working molecule
+                    mol_w = combined
+
+                    # Mark the original atom for removal
+                    atoms_to_remove.append(i)
+
+                else:
+                    # remove bonds connected to abbreviation/condensed formula
+                    adjacent_indices = [bond.GetOtherAtomIdx(i) for bond in bonds]
+                    for adjacent_idx in adjacent_indices:
+                        mol_w.RemoveBond(i, adjacent_idx)
+
+                    adjacent_atoms = [mol_w.GetAtomWithIdx(adjacent_idx) for adjacent_idx in adjacent_indices]
+                    for adjacent_atom, bond in zip(adjacent_atoms, bonds):
+                        adjacent_atom.SetNumRadicalElectrons(int(bond.GetBondTypeAsDouble()))
+
+                    # get indices of atoms of main body that connect to substituent
+                    bonding_atoms_w = adjacent_indices
+                    # assume indices are concated after combine mol_w and mol_r
+                    bonding_atoms_r = [mol_w.GetNumAtoms()]
+                    for atm in mol_r.GetAtoms():
+                        if atm.GetNumRadicalElectrons() and atm.GetIdx() > 0:
+                            bonding_atoms_r.append(mol_w.GetNumAtoms() + atm.GetIdx())
+
+                    # combine main body and substituent into a single molecule object
+                    combo = Chem.CombineMols(mol_w, mol_r)
+
+                    # connect substituent to main body with bonds
+                    mol_w = Chem.RWMol(combo)
+                    # if len(bonding_atoms_r) == 1:  # substituent uses one atom to bond to main body
+                    for atm in bonding_atoms_w:
+                        bond_order = mol_w.GetAtomWithIdx(atm).GetNumRadicalElectrons()
+                        mol_w.AddBond(atm, bonding_atoms_r[0], order=BOND_TYPES[bond_order])
+
+                    # reset radical electrons
+                    for atm in bonding_atoms_w:
+                        mol_w.GetAtomWithIdx(atm).SetNumRadicalElectrons(0)
+                    for atm in bonding_atoms_r:
+                        mol_w.GetAtomWithIdx(atm).SetNumRadicalElectrons(0)
+                    atoms_to_remove.append(i)
 
         # Remove atom in the end, otherwise the id will change
         # Reverse the order and remove atoms with larger id first
@@ -551,10 +578,12 @@ def _convert_graph_to_smiles(coords, symbols, edges, image=None, debug=False):
             ratio = width / height
             coords = [[x * ratio * 10, y * 10] for x, y in coords]
         mol = _verify_chirality(mol, coords, symbols, edges, debug)
-        # molblock is obtained before expanding func groups, otherwise the expanded group won't have coordinates.
-        # TODO: make sure molblock has the abbreviation information
-        pred_molblock = Chem.MolToMolBlock(mol)
-        pred_smiles, mol = _expand_functional_group(mol, {}, debug)
+
+        # First expand functional groups
+        pred_smiles, expanded_mol = _expand_functional_group(mol, {}, debug)
+
+        # Generate molblock from the expanded molecule
+        pred_molblock = Chem.MolToMolBlock(expanded_mol)
         success = True
     except Exception as e:
         if debug:

molscribe/constants.py

@@ -4,7 +4,8 @@
 ORGANIC_SET = {'B', 'C', 'N', 'O', 'P', 'S', 'F', 'Cl', 'Br', 'I'}
 
 RGROUP_SYMBOLS = ['R', 'R1', 'R2', 'R3', 'R4', 'R5', 'R6', 'R7', 'R8', 'R9', 'R10', 'R11', 'R12',
-                  'Ra', 'Rb', 'Rc', 'Rd', 'X', 'Y', 'Z', 'Q', 'A', 'E', 'Ar']
+                  'Ra', 'Rb', 'Rc', 'Rd', 'Rf', 'X', 'Y', 'Z', 'Q', 'A', 'E', 'Ar', 'Ar1', 'Ar2', 'Ari', 'Ar3', 'Ar4','Ar5','Ar6','Ar7',"R'",
+                  '1*', '2*','3*', '4*','5*', '6*','7*', '8*','9*', '10*','11*', '12*','[a*]', '[b*]','[c*]', '[d*]',"EWG",'Nu']
 
 PLACEHOLDER_ATOMS = ["Lv", "Lu", "Nd", "Yb", "At", "Fm", "Er"]
 
@@ -21,8 +22,9 @@ def __init__(self, abbrvs, smarts, smiles, probability):
 
 SUBSTITUTIONS: List[Substitution] = [
     Substitution(['NO2', 'O2N'], '[N+](=O)[O-]', "[N+](=O)[O-]", 0.5),
+    Substitution(['OCOCH3'], '[#8]-[#6](=[#8])-[#6]', "[O]C(=O)C]", 0.5),
     Substitution(['CHO', 'OHC'], '[CH1](=O)', "[CH1](=O)", 0.5),
-    Substitution(['CO2Et', 'COOEt'], 'C(=O)[OH0;D2][CH2;D2][CH3]', "[C](=O)OCC", 0.5),
+    Substitution(['CO2Et', 'COOEt', 'EtO2C'], 'C(=O)[OH0;D2][CH2;D2][CH3]', "[C](=O)OCC", 0.5),
 
     Substitution(['OAc'], '[OH0;X2]C(=O)[CH3]', "[O]C(=O)C", 0.7),
     Substitution(['NHAc'], '[NH1;D2]C(=O)[CH3]', "[NH]C(=O)C", 0.7),

molscribe/interface.py

@@ -1,4 +1,5 @@
 import argparse
+import os
 from typing import List
 
 import cv2
@@ -30,7 +31,7 @@ def __init__(self, model_path, device=None, num_workers=1):
         MolScribe Interface
         :param model_path: path of the model checkpoint.
         :param device: torch device, defaults to be CPU.
-        :param multiprocessing_enabled: uses multiprocessing to parallelize parts of the inference when enabled, defaults to False.
+        :param num_workers: number of workers for parallel processing, defaults to 1.
         """
         model_states = torch.load(model_path, map_location=torch.device('cpu'))
         args = self._get_args(model_states['args'])
@@ -41,6 +42,9 @@ def __init__(self, model_path, device=None, num_workers=1):
         self.encoder, self.decoder = self._get_model(args, self.tokenizer, self.device, model_states)
         self.transform = get_transforms(args.input_size, augment=False)
         self.num_workers = num_workers
+        # MPS-specific optimizations
+        self.is_mps = str(device).startswith('mps')
+        self.optimal_batch_size = 32 if self.is_mps else 4
 
     def _get_args(self, args_states=None):
         parser = argparse.ArgumentParser()
@@ -90,7 +94,13 @@ def _get_model(self, args, tokenizer, device, states):
         decoder.eval()
         return encoder, decoder
 
-    def predict_images(self, input_images: List, return_atoms_bonds=False, return_confidence=False, batch_size=16):
+    def predict_images(self, input_images: List, return_atoms_bonds=False, return_confidence=False, batch_size=None):
+        """
+         Optimized version of predict_images with better MPS performance
+         """
+        if batch_size is None:
+            batch_size = self.optimal_batch_size
+
         device = self.device
         predictions = []
         self.decoder.compute_confidence = return_confidence
@@ -104,7 +114,9 @@ def predict_images(self, input_images: List, return_atoms_bonds=False, return_co
                 batch_predictions = self.decoder.decode(features, hiddens)
             predictions += batch_predictions
 
-        smiles = [pred['chartok_coords']['smiles'] for pred in predictions]
+        return self.convert_graph_to_output(predictions, input_images, return_confidence, return_atoms_bonds)
+
+    def convert_graph_to_output(self, predictions, input_images, return_confidence=False, return_atoms_bonds=False):
         node_coords = [pred['chartok_coords']['coords'] for pred in predictions]
         node_symbols = [pred['chartok_coords']['symbols'] for pred in predictions]
         edges = [pred['edges'] for pred in predictions]

requirements.txt

@@ -1,14 +1,15 @@
-torch
+torch>=2.5.0,<=2.7.1
 torchvision
 numpy>=1.19.5,<2.0
-pandas>=1.2.4
+pandas>=2.2.3
 matplotlib>=3.5.3
-opencv-python==4.5.5.64
-transformers>=4.5.1
+opencv-python>=4.10.0.84
 huggingface-hub>=0.11.0
 tensorboardX
 SmilesPE==0.0.3
 OpenNMT-py==2.2.0
-rdkit>=2022.3.3
-albumentations @ git+https://github.com/albumentations-team/albumentations@37e714fd2e326f6f88778e425f98c2de8c8d5372
-timm @ git+https://github.com/rwightman/pytorch-image-models.git@54a6cca27a9a3e092a07457f5d56709da56e3cf5
+rdkit~=2025.3.6
+albumentations==1.1.0
+timm>=0.4.12,<=0.5.4
+# Only this version of transformers are compatible with timm
+transformers>=4.47.0,<=4.52.4

setup.py

@@ -1,8 +1,12 @@
 from distutils.core import setup
-from pathlib import Path
+
+def read_requirements():
+    """Read the requirements.txt file and return a list of dependencies."""
+    with open("requirements.txt", "r", encoding="utf-8") as fh:
+        return fh.read().splitlines()
 
 setup(name='MolScribe',
-      version='1.1.1',
+      version='1.2.1',
       description='MolScribe',
       author='Yujie Qian',
       author_email='yujieq@csail.mit.edu',
@@ -12,15 +16,5 @@
       package_data={'molscribe': ['vocab/*']},
       python_requires='>=3.7',
       setup_requires=['numpy'],
-      install_requires=[
-        "numpy>=1.19.5,<2.0",
-        "torch>=1.11.0",
-        "pandas",
-        "matplotlib",
-        "opencv-python>=4.5.5.64",
-        "SmilesPE==0.0.3",
-        "OpenNMT-py==2.2.0",
-        "rdkit>=2022.3.3",
-        "albumentations==1.1.0",
-        "timm==0.4.12"
-      ])
+      install_requires=read_requirements(),
+      )