Add new POTCAR validation to VASP validation plus tests (#892)

* Fix issue with using dir_name != ./ in emmet.core.tasks.TaskDoc

* Revert previous change - actual fix is using path relative to dir_name of TaskDoc.from_directory

* Add new POTCAR summary_stats check for validation, two tests for POTCAR checking/IDing

* Adding builder side of commit and test files

* Remove MP_POTCAR_ID.json.gz test file

* Fix handling of missing POTCAR lib in tests

* cleanup tests

* Hopefully fix weird behavior in failing matcalc tests

---------

Co-authored-by: esoteric-ephemera <[email protected]>

emmet-builders/emmet/builders/vasp/task_validator.py

@@ -52,7 +52,7 @@ def __init__(
                         potcar = PotcarSingle.from_symbol_and_functional(
                             symbol=potcar_symbol, functional=functional
                         )
-                        hashes[calc_type][potcar_symbol] = potcar.md5_header_hash
+                        hashes[calc_type][potcar_symbol] = potcar._summary_stats
 
                 self.potcar_hashes = potcar_hashes
         else:

emmet-core/emmet/core/tasks.py

@@ -224,7 +224,7 @@ class CustodianDoc(BaseModel):
     )
     job: Optional[Any] = Field(
         None,
-        title="Cusotodian Job Data",
+        title="Custodian Job Data",
         description="Job data logged by custodian.",
     )
 

emmet-core/emmet/core/vasp/validation.py

@@ -111,7 +111,7 @@ def from_task_doc(
                 valid_input_set = None
 
             if valid_input_set:
-                # Checking POTCAR hashes if a directory is supplied
+                # Checking POTCAR summary_stats if a directory is supplied
                 if potcar_hashes:
                     if _potcar_hash_check(task_doc, potcar_hashes):
                         if task_type in [
@@ -313,28 +313,47 @@ def _kspacing_warnings(input_set, inputs, data, warnings, kspacing_tolerance):
 
 def _potcar_hash_check(task_doc, potcar_hashes):
     """
-    Checks to make sure the POTCAR hash is equal to the correct value from the
-    pymatgen input set.
+    Checks to make sure the POTCAR summary stats is equal to the correct
+    value from the pymatgen input set.
     """
+    data_tol = 1.0e-6
 
     try:
         potcar_details = task_doc.calcs_reversed[0]["input"]["potcar_spec"]
 
-        all_match = True
+    except KeyError:
+        # Assume it is an old calculation without potcar_spec data and treat it as passing POTCAR hash check
+        return False
 
-        for entry in potcar_details:
-            symbol = entry["titel"].split(" ")[1]
-            hash = potcar_hashes[str(task_doc.calc_type)].get(symbol, None)
+    all_match = True
+    for entry in potcar_details:
+        symbol = entry["titel"].split(" ")[1]
+        ref_summ_stats = potcar_hashes[str(task_doc.calc_type)].get(symbol, None)
+        if not ref_summ_stats:
+            all_match = False
+            break
+
+        key_match = all(
+            set(ref_summ_stats["keywords"][key])
+            == set(entry["summary_stats"]["keywords"][key])
+            for key in ["header", "data"]
+        )
 
-            if not hash or hash != entry["hash"]:
-                all_match = False
-                break
+        data_match = all(
+            abs(
+                ref_summ_stats["stats"][key][stat]
+                - entry["summary_stats"]["stats"][key][stat]
+            )
+            < data_tol
+            for stat in ["MEAN", "ABSMEAN", "VAR", "MIN", "MAX"]
+            for key in ["header", "data"]
+        )
 
-        return not all_match
+        if (not key_match) or (not data_match):
+            all_match = False
+            break
 
-    except KeyError:
-        # Assume it is an old calculation without potcar_spec data and treat it as passing POTCAR hash check
-        return False
+    return not all_match
 
 
 def _magmom_check(task_doc, chemsys):

emmet-core/tests/test_calculation.py

@@ -161,3 +161,30 @@ def test_calculation(test_dir, object_name, task_name):
 
     # and decoded
     MontyDecoder().process_decoded(d)
+
+
+def test_PotcarSpec(test_dir):
+    from emmet.core.vasp.calculation import PotcarSpec
+    from pymatgen.io.vasp import PotcarSingle, Potcar
+
+    try:
+        # First test, PotcarSingle object
+        potcar = PotcarSingle.from_symbol_and_functional(symbol="Si", functional="PBE")
+        ps_spec = PotcarSpec.from_potcar_single(potcar_single=potcar)
+
+        assert ps_spec.titel == potcar.symbol
+        assert ps_spec.hash == potcar.md5_header_hash
+        assert ps_spec.summary_stats == potcar._summary_stats
+
+        # Second test, Potcar object containing mulitple PotcarSingle obejcts
+        potcars = Potcar(symbols=["Ga_d", "As"], functional="PBE")
+        ps_spec = PotcarSpec.from_potcar(potcar=potcars)
+
+        for ips, ps in enumerate(ps_spec):
+            assert ps.titel == potcars[ips].symbol
+            assert ps.hash == potcars[ips].md5_header_hash
+            assert ps.summary_stats == potcars[ips]._summary_stats
+
+    except (OSError, ValueError):
+        # missing Pymatgen POTCARs, cannot perform test
+        assert True

emmet-core/tests/test_ml.py

@@ -52,7 +52,7 @@
     ("calculator", "prop_kwargs"),
     [
         (get_universal_calculator("chgnet"), None),
-        ("m3gnet", {"ElasticityCalc": {"relax_structure": False}}),
+        ("M3GNet-MP-2021.2.8-PES", {"ElasticityCalc": {"relax_structure": False}}),
     ],
 )
 def test_ml_doc(calculator: Union[str, "Calculator"], prop_kwargs: dict) -> None:

emmet-core/tests/vasp/test_vasp.py

@@ -5,7 +5,7 @@
 
 from emmet.core.vasp.calc_types import RunType, TaskType, run_type, task_type
 from emmet.core.vasp.task_valid import TaskDocument
-from emmet.core.vasp.validation import ValidationDoc
+from emmet.core.vasp.validation import ValidationDoc, _potcar_hash_check
 
 
 def test_task_type():
@@ -85,3 +85,57 @@ def test_ldau_validation(test_dir):
     valid = ValidationDoc.from_task_doc(task)
 
     assert valid.valid
+
+
+def test_potcar_hash_check(test_dir):
+    from pymatgen.io.vasp import PotcarSingle
+
+    with zopen(test_dir / "CoF_TaskDoc.json") as f:
+        data = json.load(f)
+
+    """
+    NB: seems like TaskDoc is not fully compatible with TaskDocument
+    excluding all keys but `last_updated` ensures TaskDocument can be built
+
+    Similarly, after a TaskDoc is dumped to a file, using
+        json.dump(
+            jsanitize(
+                < Task Doc >.model_dump()
+            ),
+        < filename > )
+    I cannot rebuild the TaskDoc without excluding the `orig_inputs` key.
+    """
+    task_doc = TaskDocument(**{key: data[key] for key in data if key != "last_updated"})
+
+    # First check: generate hashes from POTCARs in TaskDoc, check should pass
+    calc_type = str(task_doc.calc_type)
+    expected_hashes = {calc_type: {}}
+    try:
+        for spec in task_doc.calcs_reversed[0]["input"]["potcar_spec"]:
+            symbol = spec["titel"].split(" ")[1]
+            expected_hashes[calc_type][
+                symbol
+            ] = PotcarSingle.from_symbol_and_functional(
+                symbol=symbol, functional="PBE"
+            )._summary_stats
+
+        assert not _potcar_hash_check(task_doc, expected_hashes)
+
+        # Second check: remove POTCAR from expected_hashes, check should fail
+
+        missing_hashes = {calc_type: {**expected_hashes[calc_type]}}
+        first_element = list(missing_hashes[calc_type])[0]
+        missing_hashes[calc_type].pop(first_element)
+        assert _potcar_hash_check(task_doc, missing_hashes)
+
+        # Third check: change data in expected hashes, check should fail
+
+        wrong_hashes = {calc_type: {**expected_hashes[calc_type]}}
+        for key in wrong_hashes[calc_type][first_element]["stats"]["data"]:
+            wrong_hashes[calc_type][first_element]["stats"]["data"][key] *= 1.1
+
+        assert _potcar_hash_check(task_doc, wrong_hashes)
+
+    except (OSError, ValueError):
+        # missing Pymatgen POTCARs, cannot perform test
+        assert True