Python 3.10 support (#242)

* Python310 support (#241)

* updating README and docs [skip ci]

* chore: attempt 3.10 support

* ci: include 3.10 workflow

* ci: disable clone protection

* ci: mitigate mypy failure

* chore: start fixing mypy errors

* fix: satisfy linter by using optional types

* chore: blackening

* chore: ignore gflownet foreign code with mypy

* fix: linter mypy

* fix: apply black

* fix: linter

* chore: pinning scipy

* chore: upper bound pandas for python 3.10

* fix: compatibility with python<3.10

* chore: update readme with correct version reqs

---------

Co-authored-by: Filip Skogh <[email protected]>

.github/workflows/pypi.yaml

@@ -16,6 +16,7 @@ jobs:
       max-parallel: 3
       matrix:
         python-version:
+          - 3.10
           - 3.8
           - 3.7
         os:

.github/workflows/tests.yaml

@@ -19,6 +19,8 @@ jobs:
     defaults:
       run:
         shell: bash -l {0} # for conda command
+    env:
+      GIT_CLONE_PROTECTION_ACTIVE: false
     steps:
       - uses: actions/checkout@v2
       - uses: conda-incubator/setup-miniconda@v2
@@ -48,7 +50,7 @@ jobs:
       - name: Check mypy
         run: |
           conda activate gt4sd
-          python -m mypy src/gt4sd
+          python -m mypy src/gt4sd --show-traceback
       - name: Run pytests
         run: |
           conda activate gt4sd

README.md

@@ -27,8 +27,8 @@ Almost all pretrained models are also available via `gradio`-powered [web apps](
 
 Currently `gt4sd` relies on:
 
-- python>=3.7,<3.9
-- pip>=19.1,<20.3
+- python>=3.7,<=3.10
+- pip==24.0
 
 We are actively working on relaxing these, so stay tuned or help us with this by [contributing](./CONTRIBUTING.md) to the project.
 

conda_cpu_linux.yml

@@ -5,8 +5,9 @@ channels:
   - https://conda.anaconda.org/pyg
   - https://conda.anaconda.org/conda-forge
 dependencies:
-  - python>=3.7,<3.9
-  - pip>=19.1,<20.3
+  - setuptools==69.5.1
+  - python>=3.7,<3.11
+  - pip=24.0
   - pytorch>=1.0,<=1.12.1
   - cpuonly
   - pytorch-scatter<=2.0.9=*cu102*

conda_cpu_mac.yml

@@ -5,8 +5,9 @@ channels:
   - https://conda.anaconda.org/pyg
   - https://conda.anaconda.org/conda-forge
 dependencies:
-  - python>=3.7,<3.9
-  - pip>=19.1,<20.3
+  - setuptools==69.5.1
+  - python>=3.7,<3.11
+  - pip=24.0
   - pytorch>=1.0,<=1.12.1
   - cpuonly
   - pytorch-scatter<=2.0.9

conda_gpu.yml

@@ -5,8 +5,9 @@ channels:
   - https://conda.anaconda.org/pyg
   - https://conda.anaconda.org/conda-forge
 dependencies:
-  - python>=3.7,<3.9
-  - pip>=19.1,<20.3
+  - setuptools==69.5.1
+  - python>=3.7,<3.11
+  - pip=24.0
   - pytorch>=1.0,<=1.12.1=*cu*
   - pytorch-scatter<=2.0.9=*cu102*
   - torchvision<=0.13.1=*cu*

dev_requirements.txt

@@ -6,11 +6,11 @@ flask==1.1.2
 flask_login==0.5.0
 # isort==5.7.0
 licenseheaders==0.8.8
-mypy==0.950
-myst-parser==0.13.3
-pytest==6.1.1
+mypy>=1.0.0
+myst-parser==1.0.0
+pytest==6.2.5
 pytest-cov==2.10.1
-sphinx==3.4.3
+sphinx>=5
 sphinx-autodoc-typehints==1.11.1
 jinja2<3.1.0
 sphinx_rtd_theme==0.5.1

requirements.txt

@@ -1,4 +1,5 @@
 # pypi requirements
+setuptools==69.5.1
 accelerate>=0.12,<0.20.0
 datasets>=1.11.0
 diffusers<=0.6.0
@@ -11,12 +12,13 @@ gt4sd-molformer>=0.1.1
 gt4sd-trainer-hf-pl>=0.0.2
 keras>=2.3.1,<2.11.0
 keybert>=0.7.0
-markdown-it-py<3.0.0,>=2.2.0
+markdown-it-py
 minio==7.0.1
 modlamp>=4.0.0
 molgx>=0.22.0a1
 nglview>=3.0.3
 numpy>=1.16.5,<1.24.0
+pandas<=2.0.3
 protobuf<3.20
 pyarrow>=8.0.0
 pydantic>=1.7.3,<2.0.0
@@ -25,12 +27,13 @@ PyTDC==0.3.7
 pytorch_lightning<=1.7.7
 pyyaml>=5.4.1
 rdkit>=2022.3.5
+rdkit-stubs>=0.7
 regex>=2.5.91
 reinvent-chemistry==0.0.38
 sacremoses>=0.0.41
 scikit-learn>=1.0.0,<1.3.0
 scikit-optimize>=0.8.1
-scipy>=1.0.0
+scipy>=1.0.0,<=1.11.0
 sentencepiece>=0.1.95
 sentence_transformers>1.0,<=2.2.2
 sympy>=1.10.1

setup.cfg

@@ -98,13 +98,20 @@ ignore = E203, E501, W503
 [mypy]
 check_untyped_defs = True
 plugins = pydantic.mypy
+exclude = ^src/gt4sd/frameworks/gflownet/envs
+
+[mypy-gt4sd.frameworks.gflownet.envs.*]
+follow_imports = skip
 
 [mypy-pytest.*]
 ignore_missing_imports = True
 
 [mypy-rdkit.*]
 ignore_missing_imports = True
 
+[mypy-rdkit-stubs.*]
+ignore_errors = True
+
 [mypy-setuptools.*]
 ignore_missing_imports = True
 
@@ -141,7 +148,6 @@ ignore_missing_imports = True
 [mypy-transformers.*]
 ignore_missing_imports = True
 
-# to avoid mypy from crashing (https://github.com/python/mypy/issues/11045)
 [mypy-transformers.trainer]
 check_untyped_defs = False
 

src/gt4sd/algorithms/conditional_generation/regression_transformer/implementation.py

@@ -1013,7 +1013,7 @@ def validate_output(self, sequences: List[Any]) -> Tuple[List[Any], List[int]]:
             if isinstance(self.tokenizer.text_tokenizer, PolymerGraphTokenizer):
                 # Copolymer models require specific validation
                 return validate_molecules(
-                    pattern_list=list(zip(*sequences))[0],
+                    pattern_list=list(zip(*sequences))[0],  # type: ignore
                     input_type=MoleculeFormat.copolymer,
                 )
 
@@ -1086,9 +1086,11 @@ def filter_substructures(
                     f"{keep} is not a valid SMILES/SELFIES. Instead substructure filtering "
                     f"based on sequence alone can be done and is set to: {self.text_filtering}"
                 )
-            if keep not in self.substructures_to_keep and not Chem.MolFromSmiles(
+            if keep not in self.substructures_to_keep and not Chem.MolFromSmiles(  # type: ignore
                 self.target
-            ).HasSubstructMatch(subs_mol):
+            ).HasSubstructMatch(
+                subs_mol
+            ):
                 logger.info(
                     f"{keep} could not be identified in SMILES/SELFIES on text level AND no "
                     "substructure match occurred, hence it will be ignored"
@@ -1120,7 +1122,7 @@ def filter_substructures(
                         sane = False
                         break
                 else:
-                    if not mol.HasSubstructMatch(subs_mol):
+                    if not mol.HasSubstructMatch(subs_mol):  # type: ignore
                         # Desired substructure not found
                         sane = False
                         break

src/gt4sd/algorithms/conditional_generation/regression_transformer/utils.py

@@ -65,20 +65,20 @@ def filter_stubbed(
         Tuple of tuples of length 2 with filtered, generated molecule and its properties.
     """
 
-    seed = Chem.MolFromSmiles(target)
+    seed = Chem.MolFromSmiles(target)  # type: ignore
 
-    seed_atoms = len(list(seed.GetAtoms()))
-    seed_bonds = seed.GetNumBonds()
+    seed_atoms = len(list(seed.GetAtoms()))  # type: ignore
+    seed_bonds = seed.GetNumBonds()  # type: ignore
 
     smis: List[str] = []
     props: List[str] = []
     for smi, prop in property_sequences:
         if smi == "":
             continue
         try:
-            mol = Chem.MolFromSmiles(smi)
-            num_atoms = len(list(mol.GetAtoms()))
-            num_bonds = mol.GetNumBonds()
+            mol = Chem.MolFromSmiles(smi)  # type: ignore
+            num_atoms = len(list(mol.GetAtoms()))  # type: ignore
+            num_bonds = mol.GetNumBonds()  # type: ignore
 
             if num_atoms > (threshold * seed_atoms) and num_bonds > (
                 threshold * seed_bonds

src/gt4sd/algorithms/controlled_sampling/advanced_manufacturing/implementation/core.py

@@ -154,7 +154,7 @@ def __init__(
         targets: Dict[str, float],
         property_predictors: Dict[str, PropertyPredictor],
         representations: RepresentationsDict,
-        representation_order: List[str] = None,
+        representation_order: Optional[List[str]] = None,
         scalers: Optional[Dict[str, Scaler]] = None,
         weights: Optional[Dict[str, float]] = None,
         custom_score_function: Optional[

src/gt4sd/algorithms/core.py

@@ -25,7 +25,7 @@
 
 from __future__ import annotations
 
-import collections
+from collections.abc import Hashable
 import logging
 import os
 import shutil
@@ -233,7 +233,7 @@ def sample(self, number_of_items: int = 100) -> Iterator[S]:
                 try:
                     valid_item = self.configuration.validate_item(item)
                     # check if sample is hashable
-                    if not isinstance(item, collections.Hashable):
+                    if not isinstance(item, Hashable):
                         yield valid_item
                         item_set.add(str(index))
                     else:
@@ -623,9 +623,11 @@ def save_version_from_training_pipeline_arguments(
                 target_version,
             )
             filepaths_mapping = {
-                filename: source_filepath
-                if os.path.exists(source_filepath)
-                else os.path.join(source_missing_path, filename)
+                filename: (
+                    source_filepath
+                    if os.path.exists(source_filepath)
+                    else os.path.join(source_missing_path, filename)
+                )
                 for filename, source_filepath in filepaths_mapping.items()
             }
             logger.info(f"Saving artifacts into {target_path}...")
@@ -713,9 +715,11 @@ def upload_version_from_training_pipeline_arguments(
 
             # mapping between filenames and paths for a version.
             filepaths_mapping = {
-                filename: source_filepath
-                if os.path.exists(source_filepath)
-                else os.path.join(source_missing_path, filename)
+                filename: (
+                    source_filepath
+                    if os.path.exists(source_filepath)
+                    else os.path.join(source_missing_path, filename)
+                )
                 for filename, source_filepath in filepaths_mapping.items()
             }
 

src/gt4sd/algorithms/generation/diffusion/geodiff/core.py

@@ -253,8 +253,8 @@ def visualize_2d_input(self, data: Data) -> None:
         molSize = (450, 300)
         drawer = MD2.MolDraw2DSVG(molSize[0], molSize[1])
         drawer.DrawMolecule(mc)
-        drawer.FinishDrawing()
-        svg = drawer.GetDrawingText()
+        drawer.FinishDrawing()  # type: ignore
+        svg = drawer.GetDrawingText()  # type: ignore
         display(SVG(svg.replace("svg:", "")))
 
     def visualize_3d(self, mols_gen: List[Chem.Mol]) -> None:

src/gt4sd/algorithms/generation/diffusion/geodiff/model/core.py

@@ -21,7 +21,7 @@
 # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 # SOFTWARE.
 #
-from typing import Any, Tuple, Union
+from typing import Any, Tuple, Union, Optional
 
 import torch
 from diffusers.configuration_utils import ConfigMixin
@@ -127,13 +127,13 @@ def _forward(
         bond_index: torch.Tensor,
         bond_type: torch.Tensor,
         batch: torch.Tensor,
-        edge_index: torch.Tensor = None,
-        edge_type: torch.Tensor = None,
-        edge_length: int = None,
+        edge_index: Optional[torch.Tensor] = None,
+        edge_type: Optional[torch.Tensor] = None,
+        edge_length: Optional[int] = None,
         return_edges: bool = False,
         extend_order: bool = True,
         extend_radius: bool = True,
-        is_sidechain: bool = None,
+        is_sidechain: Optional[bool] = None,
     ) -> Tuple[Any, ...]:
         """Forward pass for edges features.
 

src/gt4sd/algorithms/generation/diffusion/implementation.py

@@ -63,7 +63,7 @@ def set_seed(seed: int = 42) -> None:
     """
     np.random.seed(seed)
     torch.manual_seed(seed)
-    if torch.cuda.is_available:
+    if torch.cuda.is_available():
         torch.cuda.manual_seed_all(seed)  # type:ignore
 
 

src/gt4sd/algorithms/generation/hugging_face/implementation.py

@@ -79,7 +79,7 @@ def set_seed(seed: int = 42) -> None:
     """
     np.random.seed(seed)
     torch.manual_seed(seed)
-    if torch.cuda.is_available:
+    if torch.cuda.is_available():
         torch.cuda.manual_seed_all(seed)  # type:ignore
 
 

src/gt4sd/domains/materials/__init__.py

@@ -71,7 +71,7 @@ def validate_smiles(
         for index, molecule in enumerate(molecules)
         if molecule is not None and molecule != ""
     ]
-    return molecules, valid_ids
+    return molecules, valid_ids  # type: ignore
 
 
 def validate_selfies(

src/gt4sd/frameworks/cgcnn/data.py

@@ -31,7 +31,7 @@
 import logging
 import os
 import random
-from typing import Any, Callable, List, Tuple, Union
+from typing import Any, Callable, List, Tuple, Union, Optional
 
 import numpy as np
 import torch
@@ -49,7 +49,7 @@ def get_train_val_test_loader(
     dataset: torch.utils.data.Dataset,
     collate_fn: Callable[[List[Any]], Any] = default_collate,
     batch_size: int = 64,
-    train_ratio: float = None,
+    train_ratio: Optional[float] = None,
     val_ratio: float = 0.1,
     test_ratio: float = 0.1,
     return_test: bool = False,
@@ -212,7 +212,9 @@ class GaussianDistance:
     Unit: angstrom
     """
 
-    def __init__(self, dmin: float, dmax: float, step: float, var: float = None):
+    def __init__(
+        self, dmin: float, dmax: float, step: float, var: Optional[float] = None
+    ):
         """
         Args:
             dmin: float
@@ -333,7 +335,7 @@ def __init__(
         dmin: int = 0,
         step: float = 0.2,
         random_seed: int = 123,
-        atom_initialization: AtomCustomJSONInitializer = None,
+        atom_initialization: Optional[AtomCustomJSONInitializer] = None,
     ):
         """
         Args: