Fix test_smoke_molecule_pgd and test_standard_pgd regressions (#54)

- Restore 20-molecule SMILES lists and subsample_size=8 in
  test_molecule_metrics.py. The camera-ready commit (#48) shrunk these
  to 10 molecules with subsample_size=4, which is too small for
  StratifiedKFold(n_splits=4).

- Use np.isclose(rtol=1e-3) instead of exact equality in
  test_standard_pgd. The shared TabPFN classifier instance in
  StandardPGD causes minor float divergence (~1e-5) compared to
  fresh per-descriptor instances.

Author: Markus28

tests/test_molecule_metrics.py

@@ -20,22 +20,42 @@
     "CC(C)Cc1ccc(cc1)C(C)C(=O)O",
     "CC1(C)SC2C(NC(=O)C2=O)C1(C)C(=O)N",
     "C1C(=O)N(C2=CC=CC=C12)C3=CC=C(C=C3)C(F)(F)F",
+    "CCCCCCOc1ccc(C(=O)C=Cc2c(C=Cc3ccc(OC)cc3)cc(OC)cc2OC)cc1",
     "O=C(Nc1nc(-c2ccc(Cl)s2)cs1)c1ccncc1",
+    "COc1nc(N(C)C)ncc1-n1nc2c(c1C(C)C)C(c1ccc(C#N)c(F)c1)N(c1c[nH]c(=O)c(Cl)c1)C2=O",
     "Cc1ncc([N+](=O)[O-])n1CC(=O)Nc1ccccc1",
+    "CCOC(=O)N=C(NC(C)C)c1ccc(-c2ccc(-c3ccc(C(=NC(=O)OCC)NC(C)C)cc3)o2)cc1",
+    "COCC(=O)N1CCC(C2CC(C(F)(F)F)n3nc(C)cc3N2)CC1",
+    "Fc1ccc(C(OCCN2C3CCC2CC(Cc2ccccc2)C3)c2ccc(F)cc2)cc1",
+    "CC(C)C1CCN(C(=O)C2CCC(=O)N(C3CCCCCC3)C2)CC1",
+    "CCc1c2c(n(C)c1C)CCCC2=NOC(=O)Nc1ccc(C(C)=O)cc1",
+    "Cc1cc2c(-c3ccc(S(=O)(=O)NCCO)cc3)ccnc2[nH]1",
     "CC(C)N1CCN(C(=O)c2ccc(Oc3ccc(F)cc3)nc2)CC1",
     "O=C(Nc1ccc(Cl)c(O)c1)Nc1ccc(Cl)c(Cl)c1",
+    "O=C1NC(=NN=CC(O)C(O)C(O)C(O)CO)NC1=Cc1ccfo1",
+    "O=C(NCCO)c1c(O)c2ncc(Cc3ccc(F)cc3)cc2[nH]c1=O",
     "NC(=O)C(=O)C(Cc1ccccc1)NC(=O)C1CCN(C(=O)C=Cc2ccncc2)CC1",
 ]
 
 smiles_b = [
     "CC1=C(C=CC=C1)NC2=NC=CC(=N2)NC3=CC=CC=C3C(=O)NC4=CC=CC=N4",
     "CN1CCN(C2=CC3=C(C=C2)N=CN3C)C4=CC=CC=C14",
     "CN(C)CCCN1C2=CC=CC=C2SC3=CC=CC=C31",
+    "CC(C)C(C(=O)NCC(C)C)NC(=O)C1=CC=CC=C1C(C)C(C)NC(=O)C2=CN=CC=C2",
     "CN1C(=O)CN=C(C2=CC=CC=C12)C3=CC=CC=C3Cl",
     "O=C(c1cc(-c2ccc(Cl)cc2Cl)n[nH]1)N1CCCC1",
     "COc1cccc(OC)c1C=CC(=O)NC1CCCCC1",
     "O=C1NC(O)CCN1C1OC(CO)C(O)C1O",
+    "Cc1c2ccnc(C(=O)NCCN(C)C)c2cc2c3cc(OC(=O)CCCCC(=O)O)ccc3n(C)c12",
     "CC1(C)SC2C(NC(=O)C2=O)C1(C)C(=O)N",
+    "C1C(=O)N(C2=CC=CC=C12)C3=CC=C(C=C3)C(F)(F)F",
+    "CCCCCCOc1ccc(C(=O)C=Cc2c(C=Cc3ccc(OC)cc3)cc(OC)cc2OC)cc1",
+    "O=C(Nc1nc(-c2ccc(Cl)s2)cs1)c1ccncc1",
+    "COc1nc(N(C)C)ncc1-n1nc2c(c1C(C)C)C(c1ccc(C#N)c(F)c1)N(c1c[nH]c(=O)c(Cl)c1)C2=O",
+    "Cc1ncc([N+](=O)[O-])n1CC(=O)Nc1ccccc1",
+    "CCOC(=O)N=C(NC(C)C)c1ccc(-c2ccc(-c3ccc(C(=NC(=O)OCC)NC(C)C)cc3)o2)cc1",
+    "COCC(=O)N1CCC(C2CC(C(F)(F)F)n3nc(C)cc3N2)CC1",
+    "Fc1ccc(C(OCCN2C3CCC2CC(Cc2ccccc2)C3)c2ccc(F)cc2)cc1",
     "O=C(NCC1CCCO1)c1ccc2c(=O)n(-c3ccccc3)c(=S)[nH]c2c1",
     "CCNc1nc(C#N)nc(N2CCCCC2)n1",
 ]
@@ -103,5 +123,5 @@ def test_smoke_molecule_pgd():
     metric = MoleculePGD(mols_a)
     metric.compute(mols_b)
 
-    metric = MoleculePGDInterval(mols_a, subsample_size=4, num_samples=4)
+    metric = MoleculePGDInterval(mols_a, subsample_size=8, num_samples=4)
     metric.compute(mols_b)

tests/test_polygraphdiscrepancy.py

@@ -1,3 +1,4 @@
+import numpy as np
 import pytest
 
 from sklearn.linear_model import LogisticRegression
@@ -234,9 +235,10 @@ def test_standard_pgd(dense_graphs, sparse_graphs):
     }
 
     for name, (_, individual_result) in individual_results.items():
+        joint = result["subscores"][name]
         assert isinstance(individual_result, float)
-        assert individual_result == result["subscores"][name], (
-            f"Individual result {individual_result} for descriptor {name} does not match the overall result {result['subscores'][name]}"
+        assert np.isclose(individual_result, joint, rtol=1e-3), (
+            f"Individual result {individual_result} for descriptor {name} does not match the overall result {joint}"
         )
 
     metric = StandardPGDInterval(dense_graphs, subsample_size=10, num_samples=4)