Merge zhaiwenxi/master into master

Incorporate zhaiwenxi's CodeQL-fix PR (#49). Conflict resolution:

- finetuner.py: keep our structural fix for the desc_cache <-> finetuner
  import cycle (load_or_extract / ensure_per_system_cache now live in
  finetuner.py) rather than their lazy
  `from dpa_adapt.data.desc_cache import load_or_extract`, which would
  ImportError since that symbol moved. Their swallowed-exception comments
  on the cache read/write paths are kept (auto-merged).
- test_split_cv.py: keep our bare-skip stub, which fully removes the
  unused `systems` flagged by CodeQL; their variant deleted only the
  rng/n_total lines and left `systems` assigned-but-unused.

cv.py merged cleanly: their unused-`n_splits` removal plus our redirect of
the ensure_per_system_cache import to dpa_adapt.finetuner.

Author: zirenjin

dpa_adapt/cv.py

@@ -458,8 +458,6 @@ def cross_validate(
             train_groups -= test_formulas
             if val_groups and train_groups:
                 fold_assignments.append((train_groups, val_groups))
-
-        n_splits = len(fold_assignments)
     else:
         # Deterministic GroupKFold: sort groups, split by index (no shuffle).
         # Reproducible given the same set of systems and groups.

dpa_adapt/data/smiles.py

@@ -362,9 +362,11 @@ def smiles_to_3d_coords(
         else:
             AllChem.UFFOptimizeMolecule(mol, maxIters=500)
     except Exception:
+        # MMFF optimization failed; fall back to UFF.
         try:
             AllChem.UFFOptimizeMolecule(mol, maxIters=500)
         except Exception:
+            # Even UFF failed — proceed with unoptimized conformer.
             pass
 
     conf = mol.GetConformer()

dpa_adapt/finetuner.py

@@ -1022,6 +1022,7 @@ def _extract_features_cached(self, systems: list[dpdata.System]) -> np.ndarray:
         ``self._extract_features()`` call below.
         """
         try:
+            # Lazy import to avoid circular dependency: finetuner → desc_cache → finetuner.
             from dpa_adapt.data.desc_cache import (
                 _cache_dir,
                 _cache_key,
@@ -1038,13 +1039,17 @@ def _extract_features_cached(self, systems: list[dpdata.System]) -> np.ndarray:
             if cache_path.is_file():
                 return np.load(cache_path)
         except Exception:
+            # Cache read failed (e.g. corrupted file, permissions) —
+            # fall through and recompute features from scratch.
             pass
 
         features = self._extract_features(systems)
         try:
             cache_path.parent.mkdir(parents=True, exist_ok=True)
             np.save(cache_path, features)
         except Exception:
+            # Cache write is best-effort — silently skip on permission errors
+            # or disk-full conditions; the features are already in memory.
             pass
         return features
 

dpa_adapt/mft.py

@@ -261,6 +261,9 @@ def _validate_and_resolve_type_map(
         )
 
         # Read elements from both datasets.
+        # If data cannot be loaded (e.g. glob hasn't resolved yet, or the
+        # data directory does not exist), fall back to empty lists — the
+        # type_map will still be resolved from the checkpoint below.
         try:
             train_systems = load_data(train_data)
         except Exception:

dpa_adapt/predictor.py

@@ -194,6 +194,8 @@ def fit(
             try:
                 est[-1].set_params(random_state=seed)
             except ValueError:
+                # Estimator does not support random_state (e.g. KNeighborsRegressor);
+                # training with the default clone is fine — no ensemble diversity needed.
                 pass
             est.fit(features, y_flat)
             self.estimators_.append(est)

source/tests/dpa_adapt/test_auto_convert.py

@@ -5,18 +5,16 @@
     annotations,
 )
 
+from importlib.util import (
+    find_spec,
+)
 from pathlib import (
     Path,
 )
 
 import pytest
 
-try:
-    import rdkit  # noqa: F401
-
-    _HAS_RDKIT = True
-except ImportError:
-    _HAS_RDKIT = False
+_HAS_RDKIT = find_spec("rdkit") is not None
 
 from dpa_adapt.data.convert import (
     _is_smiles_input,

source/tests/dpa_adapt/test_backend_contract.py

@@ -120,7 +120,8 @@ class _HeavyContract:
 
     def test_real_checkpoint_descriptor_shape(
         self,
-    ): ...  # placeholder for future Bohrium-only tests
+    ):  # placeholder for future Bohrium-only tests
+        pass
 
 
 class _HookOwner:

source/tests/dpa_adapt/test_finetuner_strategies.py

@@ -454,7 +454,7 @@ def test_extract_features_detaches_grad_tensors_before_numpy(monkeypatch):
     import numpy as np
     import torch
 
-    import dpa_adapt.finetuner as finetuner_mod
+    from dpa_adapt import finetuner as finetuner_mod
 
     class FakeExtractor:
         def __init__(self, model):

source/tests/dpa_adapt/test_mft_evaluate.py

@@ -404,7 +404,8 @@ def _fake_run(cmd, *args, **kwargs):
     cmd = captured["cmd"]
     f_idx = cmd.index("-f")
     datafile = cmd[f_idx + 1]
-    lines = [l for l in open(datafile).read().split("\n") if l.strip()]
+    with open(datafile) as f:
+        lines = [l for l in f.read().split("\n") if l.strip()]
     assert lines == [test_data]
     assert out["mae"] == pytest.approx(7.0e-03)
     assert out["n_systems"] == 1
@@ -444,7 +445,8 @@ def _fake_run(cmd, *args, **kwargs):
 
     cmd = captured["cmd"]
     datafile = cmd[cmd.index("-f") + 1]
-    lines = [l for l in open(datafile).read().split("\n") if l.strip()]
+    with open(datafile) as f:
+        lines = [l for l in f.read().split("\n") if l.strip()]
     assert lines == paths
     assert out["n_systems"] == 4
 

source/tests/dpa_adapt/test_trainer.py

@@ -525,7 +525,8 @@ def _capture(cmd, *args, **kwargs):
     f_idx = captured_cmd.index("-f")
     datafile = captured_cmd[f_idx + 1]
     assert os.path.isfile(datafile), f"datafile not written: {datafile}"
-    lines = [l for l in open(datafile).read().split("\n") if l.strip()]
+    with open(datafile) as f:
+        lines = [l for l in f.read().split("\n") if l.strip()]
     assert len(lines) == 5, f"Expected 5 systems in datafile, got {len(lines)}"
 
     assert out["mae"] == pytest.approx(0.01)