D4CMPP2 1.0.0__tar.gz → 1.0.2__tar.gz

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/D4CMPP2.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: D4CMPP2
-Version: 1.0.0
+Version: 1.0.2
 Summary: PyTorch Geometric molecular property prediction models
 License-Expression: MIT
 Requires-Python: <3.13,>=3.10
@@ -371,7 +371,10 @@ print(result.best_params, result.best_model_path)
 `HP=None` uses the model's default optimization space. A key list uses declared
 model ranges; a dictionary supplies categorical values or numeric ranges.
 Supported strategies are `"bayesian"` and `"grid"`. Atomic JSON/CSV summaries
-and trial directories allow compatible searches to resume.
+and trial directories allow compatible searches to resume. When
+`optimization_path` is omitted, each call creates a separate timestamped
+`_Models/optimize_{network}_...` directory. To resume a search, pass its
+existing directory as `optimization_path` with `resume=True`.
 
 The legacy `grid_search()` API remains available with its historical `None`
 return value and continue-on-trial-error behavior.

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: D4CMPP2
-Version: 1.0.0
+Version: 1.0.2
 Summary: PyTorch Geometric molecular property prediction models
 License-Expression: MIT
 Requires-Python: <3.13,>=3.10
@@ -371,7 +371,10 @@ print(result.best_params, result.best_model_path)
 `HP=None` uses the model's default optimization space. A key list uses declared
 model ranges; a dictionary supplies categorical values or numeric ranges.
 Supported strategies are `"bayesian"` and `"grid"`. Atomic JSON/CSV summaries
-and trial directories allow compatible searches to resume.
+and trial directories allow compatible searches to resume. When
+`optimization_path` is omitted, each call creates a separate timestamped
+`_Models/optimize_{network}_...` directory. To resume a search, pass its
+existing directory as `optimization_path` with `resume=True`.
 
 The legacy `grid_search()` API remains available with its historical `None`
 return value and continue-on-trial-error behavior.

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/README.md

@@ -346,7 +346,10 @@ print(result.best_params, result.best_model_path)
 `HP=None` uses the model's default optimization space. A key list uses declared
 model ranges; a dictionary supplies categorical values or numeric ranges.
 Supported strategies are `"bayesian"` and `"grid"`. Atomic JSON/CSV summaries
-and trial directories allow compatible searches to resume.
+and trial directories allow compatible searches to resume. When
+`optimization_path` is omitted, each call creates a separate timestamped
+`_Models/optimize_{network}_...` directory. To resume a search, pass its
+existing directory as `optimization_path` with `resume=True`.
 
 The legacy `grid_search()` API remains available with its historical `None`
 return value and continue-on-trial-error behavior.

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "1.0.0"
+__version__ = "1.0.2"
 
 from D4CMPP2.src.api.training import train as train
 from D4CMPP2.src.api.legacy_grid_search import grid_search as grid_search
@@ -13,4 +13,3 @@ from D4CMPP2.networks import (
     MolecularNetwork as MolecularNetwork,
     register_network as register_network,
 )
-

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "D4CMPP2"
-version = "1.0.0"
+version = "1.0.2"
 description = "PyTorch Geometric molecular property prediction models"
 readme = "README.md"
 requires-python = ">=3.10,<3.13"

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/src/api/optimization.py

@@ -268,6 +268,21 @@ def _write_summary(path, summary):
             staging.unlink()
 
 
+def _create_default_optimization_root(model_dir, network):
+    """Create and return a unique directory for one optimization invocation."""
+    parent = Path(model_dir or "_Models").resolve()
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")
+    stem = f"optimize_{network}_{timestamp}"
+    for suffix in itertools.count():
+        name = stem if suffix == 0 else f"{stem}_{suffix + 1}"
+        candidate = parent / name
+        try:
+            candidate.mkdir(parents=True, exist_ok=False)
+        except FileExistsError:
+            continue
+        return candidate
+
+
 def _random_parameters(domains, rng):
     return {domain.name: domain.sample(rng) for domain in domains}
 
@@ -327,7 +342,12 @@ def optimize(
     resume=True,
     **train_kwargs,
 ):
-    """Tune one registered network and return its best completed trial."""
+    """Tune one registered network and return its best completed trial.
+
+    Calls without ``optimization_path`` create independent timestamped run
+    directories. Pass the same explicit path with ``resume=True`` to resume a
+    compatible search.
+    """
     if not isinstance(optimize_strategy, str):
         raise TypeError("optimize_strategy must be 'grid' or 'bayesian'.")
     strategy = optimize_strategy.lower()
@@ -342,12 +362,13 @@ def optimize(
     elif n_trials is not None:
         raise ValueError("n_trials is only used with optimize_strategy='bayesian'.")
 
-    root = Path(
-        optimization_path
-        or Path(train_kwargs.get("MODEL_DIR", "_Models"))
-        / f"optimize_{network}"
-    ).resolve()
-    root.mkdir(parents=True, exist_ok=True)
+    if optimization_path:
+        root = Path(optimization_path).resolve()
+        root.mkdir(parents=True, exist_ok=True)
+    else:
+        root = _create_default_optimization_root(
+            train_kwargs.get("MODEL_DIR"), network
+        )
     summary_path = root / "optimization.json"
     if resume and summary_path.exists():
         summary = json.loads(summary_path.read_text(encoding="utf-8"))

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/tests/test_optimize.py

@@ -93,6 +93,39 @@ class OptimizeTests(unittest.TestCase):
                         optimization_path=temp,
                     )
 
+    @heavy_test
+    def test_default_path_is_unique_per_optimization_invocation(self):
+        optimize_module = importlib.import_module("D4CMPP2.optimize")
+        with tempfile.TemporaryDirectory(prefix="d4cmpp2-optimize-default-") as temp:
+            with mock.patch.object(
+                optimize_module, "train", side_effect=self._fake_train
+            ) as train_mock:
+                first = optimize_module.optimize(
+                    data="fixture.csv",
+                    target=["target"],
+                    network="GCN",
+                    HP={"hidden_dim": [16]},
+                    optimize_strategy="grid",
+                    MODEL_DIR=temp,
+                )
+                second = optimize_module.optimize(
+                    data="fixture.csv",
+                    target=["target"],
+                    network="GCN",
+                    HP={"hidden_dim": [16]},
+                    optimize_strategy="grid",
+                    MODEL_DIR=temp,
+                )
+
+        first_root = Path(first.summary_path).parent
+        second_root = Path(second.summary_path).parent
+        self.assertEqual(train_mock.call_count, 2)
+        self.assertNotEqual(first_root, second_root)
+        self.assertEqual(first_root.parent, Path(temp).resolve())
+        self.assertEqual(second_root.parent, Path(temp).resolve())
+        self.assertTrue(first_root.name.startswith("optimize_GCN_"))
+        self.assertTrue(second_root.name.startswith("optimize_GCN_"))
+
     @heavy_test
     def test_bayesian_runs_requested_trials_and_is_reproducible(self):
         optimize_module = importlib.import_module("D4CMPP2.optimize")

{d4cmpp2-1.0.0 → d4cmpp2-1.0.2}/tests/test_t6_ci_contract.py

@@ -1,4 +1,5 @@
 from pathlib import Path
+import re
 import unittest
 
 
@@ -7,6 +8,19 @@ WORKFLOW = ROOT / ".github" / "workflows" / "ci.yml"
 
 
 class CIContractTests(unittest.TestCase):
+    def test_runtime_version_matches_distribution_metadata(self):
+        pyproject = (ROOT / "pyproject.toml").read_text(encoding="utf-8")
+        package_init = (ROOT / "__init__.py").read_text(encoding="utf-8")
+        metadata_version = re.search(
+            r'^version\s*=\s*"([^"]+)"', pyproject, re.MULTILINE
+        )
+        runtime_version = re.search(
+            r'^__version__\s*=\s*"([^"]+)"', package_init, re.MULTILINE
+        )
+        self.assertIsNotNone(metadata_version)
+        self.assertIsNotNone(runtime_version)
+        self.assertEqual(metadata_version.group(1), runtime_version.group(1))
+
     def test_release_workflow_uses_verified_artifact_and_job_scoped_oidc(self):
         release = (ROOT / ".github" / "workflows" / "release.yml").read_text(
             encoding="utf-8"
@@ -95,6 +109,7 @@ class CIContractTests(unittest.TestCase):
         self.assertIn("check_distribution_artifacts.py", artifact_script)
         self.assertIn("wheel_install_smoke.py", artifact_script)
         self.assertIn('"--no-deps", "--force-reinstall", str(wheel)', wheel_smoke)
+        self.assertIn("m.version('D4CMPP2')", wheel_smoke)
 
     def test_dependency_metadata_uses_approved_major_version_bounds(self):
         pyproject = (ROOT / "pyproject.toml").read_text(encoding="utf-8")