Add eddiag error handling and fix AMIN error handling in VaspErrorHan…

…dler (#302)

* Added AMIN handler and unit test to VaspErrorHander logic, corrected eddrm handler to avoid custodian inducing algo_tet errors

* Added eddiag error handling and unit test

* Tuning zheev/eddiag to remove OOM issues

* pre-commit auto-fixes

* pre-commit fix

* Fix tests, ensure that custodian does not induce algo_tet errors

* simplify custodian/vasp/tests/test_validators.py

* Remove error files with too few bands.

---------

Co-authored-by: Aaron David Kaplan <esoteric@login10.chn>
Co-authored-by: Aaron Kaplan <aaronkaplan@Aarons-MBP.dhcp.lbl.gov>
Co-authored-by: Janosh Riebesell <janosh.riebesell@gmail.com>

custodian/vasp/handlers.py

@@ -99,6 +99,7 @@ class VaspErrorHandler(ErrorHandler):
         "grad_not_orth": ["EDWAV: internal error, the gradient is not orthogonal"],
         "nicht_konv": ["ERROR: SBESSELITER : nicht konvergent"],
         "zheev": ["ERROR EDDIAG: Call to routine ZHEEV failed!"],
+        "eddiag": ["ERROR in EDDIAG: call to ZHEEV/ZHEEVX/DSYEV/DSYEVX failed"],
         "elf_kpar": ["ELF: KPAR>1 not implemented"],
         "elf_ncl": ["WARNING: ELF not implemented for non collinear case"],
         "rhosyg": ["RHOSYG"],
@@ -207,6 +208,11 @@ def correct(self):
                 actions.append({"dict": "INCAR", "action": {"_set": {"ISMEAR": 0, "SIGMA": 0.05}}})
             self.error_count[err_type] += 1
 
+        # Missing AMIN error handler:
+        # previously, custodian would kill the job without letting it run if AMIN was flagged
+        if "amin" in self.errors and vi["INCAR"].get("AMIN", 0.1) > 0.01:
+            actions.append({"dict": "INCAR", "action": {"_set": {"AMIN": 0.01}}})
+
         if "inv_rot_mat" in self.errors and vi["INCAR"].get("SYMPREC", 1e-5) > 1e-8:
             actions.append({"dict": "INCAR", "action": {"_set": {"SYMPREC": 1e-8}}})
 
@@ -536,16 +542,18 @@ def correct(self):
                     UserWarning,
                 )
 
-        if "zheev" in self.errors:
+        if self.errors & {"zheev", "eddiag"}:
             # Copy CONTCAR to POSCAR if CONTCAR has already been populated.
             try:
                 is_contcar = Poscar.from_file("CONTCAR")
             except Exception:
                 is_contcar = False
             if is_contcar:
                 actions.append({"file": "CONTCAR", "action": {"_file_copy": {"dest": "POSCAR"}}})
-            if vi["INCAR"].get("ALGO", "Normal").lower() != "exact":
-                actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Exact"}}})
+            if vi["INCAR"].get("ALGO", "Normal").lower() == "fast":
+                actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Normal"}}})
+            elif vi["INCAR"].get("ALGO", "Normal").lower() == "normal":
+                actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "exact"}}})
 
         if "elf_kpar" in self.errors and vi["INCAR"].get("KPAR", 1) != 1:
             actions.append({"dict": "INCAR", "action": {"_set": {"KPAR": 1}}})
@@ -645,7 +653,7 @@ def correct(self):
                 # This time try the recovery below.
                 actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Fast"}}})
             #
-            # We will only hit the 2nd algo_teet error if the ALGO was changed back from Fast to All/Damped
+            # We will only hit the 2nd algo_tet error if the ALGO was changed back from Fast to All/Damped
             # by e.g. NonConvergingErrorHandler
             # NOTE this relies on self.errors being reset on empty set on every .check call
             if self.error_count["algo_tet"] > 0:
@@ -1052,9 +1060,8 @@ def correct(self):
 
             if (
                 v.incar.get("ISMEAR", -1) >= 0
-                or not 50 <= v.incar.get("IALGO", 38) <= 59
                 and v.incar.get("METAGGA", "--") != "--"
-                and algo != "all"
+                and (algo != "all" or (not 50 <= v.incar.get("IALGO", 38) <= 59))
             ):
                 # If meta-GGA, go straight to Algo = All only if ISMEAR is greater or equal 0.
                 # Algo = All is recommended in the VASP manual and some meta-GGAs explicitly
@@ -1083,7 +1090,9 @@ def correct(self):
                     actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Fast"}}})
                 elif algo == "fast":
                     actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Normal"}}})
-                elif algo == "normal" and (v.incar.get("ISMEAR", -1) >= 0 or not 50 <= v.incar.get("IALGO", 38) <= 59):
+                elif algo == "normal" and v.incar.get("ISMEAR", 1) >= 0:
+                    # NB: default for ISMEAR is 1. To avoid algo_tet errors, only set
+                    # ALGO = ALL if ISMEAR >= 0
                     actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "All"}}})
                 else:
                     # Try mixing as last resort
@@ -1485,9 +1494,9 @@ def correct(self):
                 actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Fast"}}})
             elif algo == "fast":
                 actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "Normal"}}})
-            elif algo == "normal":
+            elif algo == "normal" and incar.get("ISMEAR", 1) >= 0:
                 actions.append({"dict": "INCAR", "action": {"_set": {"ALGO": "All"}}})
-            elif algo == "all":
+            elif algo == "all" or (algo == "normal" and incar.get("ISMEAR", 1) < 0):
                 if amix > 0.1 and bmix > 0.01:
                     # Try linear mixing
                     actions.append(

custodian/vasp/tests/test_handlers.py

@@ -548,6 +548,38 @@ def test_read_error(self):
         assert dct["errors"] == ["read_error"]
         assert dct["actions"] is None
 
+    def test_amin(self):
+        # Cell with at least one dimension >= 50 A, but AMIN > 0.01, and calculation not yet complete
+        shutil.copy("INCAR.amin", "INCAR")
+        h = VaspErrorHandler("vasp.amin")
+        h.check()
+        d = h.correct()
+        assert d["errors"] == ["amin"]
+        assert d["actions"] == [{"action": {"_set": {"AMIN": 0.01}}, "dict": "INCAR"}]
+
+    def test_eddiag(self):
+        # subspace rotation error
+        os.remove("CONTCAR")
+        shutil.copy("INCAR.amin", "INCAR")
+        h = VaspErrorHandler("vasp.eddiag")
+        h.check()
+        d = h.correct()
+        assert d["errors"] == ["eddiag"]
+        # first check that no CONTCAR exists, only action should be updating INCAR
+        # ALGO = Fast --> ALGO = Normal
+        assert d["actions"] == [{"action": {"_set": {"ALGO": "Normal"}}, "dict": "INCAR"}]
+
+        # now copy CONTCAR and check that both CONTCAR->POSCAR
+        # and INCAR updates are included: ALGO = Normal --> ALGO = exact
+        shutil.copy("CONTCAR.eddiag", "CONTCAR")
+        h = VaspErrorHandler("vasp.eddiag")
+        h.check()
+        d = h.correct()
+        assert d["actions"] == [
+            {"file": "CONTCAR", "action": {"_file_copy": {"dest": "POSCAR"}}},
+            {"action": {"_set": {"ALGO": "exact"}}, "dict": "INCAR"},
+        ]
+
 
 class AliasingErrorHandlerTest(PymatgenTest):
     def setUp(self):

custodian/vasp/tests/test_validators.py

@@ -1,14 +1,12 @@
 import os
 import shutil
-import unittest
 
 from custodian.vasp.validators import VaspAECCARValidator, VaspFilesValidator, VaspNpTMDValidator, VasprunXMLValidator
 
 test_dir = os.path.join(os.path.dirname(__file__), "..", "..", "..", "test_files")
-cwd = os.getcwd()
 
 
-class VasprunXMLValidatorTest(unittest.TestCase):
+class TestVasprunXMLValidator:
     def test_check_and_correct(self):
         os.chdir(os.path.join(test_dir, "bad_vasprun"))
         h = VasprunXMLValidator()
@@ -26,12 +24,8 @@ def test_as_dict(self):
         h2 = VasprunXMLValidator.from_dict(d)
         assert isinstance(h2, VasprunXMLValidator)
 
-    @classmethod
-    def tearDownClass(cls):
-        os.chdir(cwd)
 
-
-class VaspFilesValidatorTest(unittest.TestCase):
+class TestVaspFilesValidator:
     def test_check_and_correct(self):
         # just an example where CONTCAR is not present
         os.chdir(os.path.join(test_dir, "positive_energy"))
@@ -47,12 +41,8 @@ def test_as_dict(self):
         h2 = VaspFilesValidator.from_dict(d)
         assert isinstance(h2, VaspFilesValidator)
 
-    @classmethod
-    def tearDownClass(cls):
-        os.chdir(cwd)
-
 
-class VaspNpTMDValidatorTest(unittest.TestCase):
+class TestVaspNpTMDValidator:
     def test_check_and_correct(self):
         # NPT-AIMD using correct VASP
         os.chdir(os.path.join(test_dir, "npt_common"))
@@ -73,18 +63,10 @@ def test_as_dict(self):
         h2 = VaspNpTMDValidator.from_dict(d)
         assert isinstance(h2, VaspNpTMDValidator)
 
-    @classmethod
-    def tearDownClass(cls):
-        os.chdir(cwd)
 
-
-class VaspAECCARValidatorTest(unittest.TestCase):
+class TestVaspAECCARValidator:
     def test_check_and_correct(self):
         # NPT-AIMD using correct VASP
         os.chdir(os.path.join(test_dir, "bad_aeccar"))
         h = VaspAECCARValidator()
         assert h.check()
-
-    @classmethod
-    def tearDownClass(cls):
-        os.chdir(cwd)

test_files/CONTCAR.eddiag

@@ -0,0 +1,87 @@
+Cs6 U6 As12 O55
+1.0
+   6.7346923911508014    0.0000000000000000    0.9285262282647135
+   0.7401923066829772    9.5890356602029918    2.1332187291702365
+   0.0000000000000000    0.0000000000000000   20.5988000000000007
+Cs U As O
+6 6 12 55
+direct
+   0.4957000000000000    0.3319500000000000    0.9441000000000001 Cs+
+   0.9702000000000000    0.6994000000000000    0.6221800000000000 Cs+
+   0.0867000000000000    0.1105000000000000    0.6551800000000000 Cs+
+   0.7371000000000000    0.0299000000000000    0.2812500000000000 Cs+
+   0.3684000000000000    0.7777800000000000    0.9883999999999999 Cs+
+   0.5954000000000000    0.4774200000000000    0.3348800000000000 Cs+
+   0.3278000000000000    0.0356500000000000    0.1334400000000000 U6+
+   0.5132600000000000    0.4086600000000000    0.1459100000000000 U6+
+   0.5191000000000000    0.7374600000000000    0.4757300000000000 U6+
+   0.3550100000000000    0.7029800000000000    0.7830900000000001 U6+
+   0.7023400000000000    0.1031800000000000    0.4972200000000000 U6+
+   0.5386900000000000    0.0726599999999999    0.8021300000000000 U6+
+   0.7045000000000000    0.4248000000000001    0.7374800000000000 As5+
+   0.8314000000000000    0.7236000000000000    0.8419200000000000 As5+
+   0.1380000000000000    0.3702000000000000    0.7862400000000000 As5+
+   0.0282000000000000    0.9725000000000000    0.8569700000000000 As5+
+   0.1601000000000000    0.6777000000000000    0.1988300000000000 As5+
+   0.2238000000000000    0.0841000000000001    0.4373500000000000 As5+
+   0.8397000000000000    0.1332000000000000    0.0763900000000000 As5+
+   0.0284000000000000    0.8342000000000001    0.4208499999999999 As5+
+   0.3544000000000000    0.3844000000000000    0.5409900000000001 As5+
+   0.9174000000000000    0.4379999999999999    0.4948500000000000 As5+
+   0.0351000000000000    0.3849000000000000    0.0865900000000000 As5+
+   0.7351000000000000    0.7370000000000000    0.1501600000000000 As5+
+   0.0140000000000000    0.7305000000000000    0.7777000000000001 O2-
+   0.6620000000000000    0.5819000000000001    0.7579000000000000 O2-
+   0.8820000000000000    0.0549999999999999    0.7982000000000000 O2-
+   0.1780000000000000    0.7551000000000000    0.4764000000000000 O2-
+   0.2537000000000000    0.9265000000000000    0.8187000000000000 O2-
+   0.5770000000000000    0.0854999999999999    0.7158000000000000 O2-
+   0.6180000000000000    0.2970999999999999    0.7952000000000000 O2-
+   0.2400000000000000    0.2114000000000000    0.7865000000000000 O2-
+   0.5500000000000000    0.4950000000000000    0.0609000000000000 O2-
+   0.5120000000000000    0.0544000000000000    0.8902000000000000 O2-
+   0.6271000000000000    0.8174000000000000    0.8173000000000000 O2-
+   0.3250000000000000    0.6163999999999999    0.8687000000000000 O2-
+   0.9150000000000000    0.8182000000000000    0.8944000000000000 O2-
+   0.2860000000000000    0.4981000000000000    0.7553000000000000 O2-
+   0.4790000000000000    0.7225000000000000    0.5642000000000000 O2-
+   0.0300000000000000    0.3731000000000000    0.8633999999999999 O2-
+   0.7730000000000000    0.3075000000000001    0.5256000000000001 O2-
+   0.0569999999999999    0.0424000000000000    0.9189000000000001 O2-
+   0.2050000000000000    0.5308000000000000    0.1727000000000000 O2-
+   0.0490000000000000    0.0733999999999999    0.5014000000000001 O2-
+   0.9920000000000000    0.7711000000000000    0.3575000000000000 O2-
+   0.3730000000000000    0.7956000000000000    0.6982999999999999 O2-
+   0.9680000000000000    0.3910000000000000    0.7289000000000000 O2-
+   0.4000000000000000    0.2285000000000000    0.5217000000000001 O2-
+   0.4820000000000000    0.3148000000000000    0.2296000000000000 O2-
+   0.6820000000000001    0.4258000000000000    0.6598999999999999 O2-
+   0.8120000000000001    0.5942000000000001    0.4958000000000000 O2-
+   0.5430000000000000    0.7559000000000000    0.3884000000000000 O2-
+   0.7920000000000000    0.5724000000000000    0.8891000000000000 O2-
+   0.7400000000000000    0.1913000000000000    0.4115000000000000 O2-
+   0.2990000000000000    0.0278000000000000    0.2205000000000000 O2-
+   0.8066000000000000    0.8840000000000000    0.4604000000000000 O2-
+   0.4430000000000000    0.5125000000000000    0.4799000000000000 O2-
+   0.0309999999999999    0.4387000000000001    0.4172000000000000 O2-
+   0.2260000000000000    0.6642000000000000    0.2741000000000000 O2-
+   0.4331000000000000    0.9927000000000000    0.4636000000000000 O2-
+   0.0960000000000001    0.4125000000000001    0.5504000000000000 O2-
+   0.8519000000000000    0.3812000000000000    0.1484000000000000 O2-
+   0.3780000000000000    0.3854000000000000    0.6205000000000001 O2-
+   0.1380000000000000    0.9863000000000000    0.3835000000000000 O2-
+   0.9855000000000000    0.0607000000000000    0.1336000000000000 O2-
+   0.6820000000000001    0.0089999999999999    0.5795000000000000 O2-
+   0.8610000000000000    0.7372000000000000    0.0755000000000000 O2-
+   0.6225000000000001    0.8922000000000000    0.1526000000000000 O2-
+   0.0734000000000000    0.5404000000000000    0.0379000000000000 O2-
+   0.8049999999999999    0.0598000000000000    0.0144000000000000 O2-
+   0.2590000000000000    0.2399000000000000    0.3884000000000000 O2-
+   0.9020000000000000    0.7111000000000000    0.2107000000000000 O2-
+   0.2452000000000000    0.2942000000000000    0.1141000000000000 O2-
+   0.6118000000000000    0.1867000000000000    0.1127000000000000 O2-
+   0.5870000000000000    0.6081000000000000    0.1783000000000000 O2-
+   0.2410000000000000    0.8121000000000000    0.1442000000000000 O2-
+   0.3547000000000000    0.0504000000000000    0.0454000000000000 O2-
+   0.9534000000000000    0.2907000000000000    0.0346000000000000 O2-
+   0.9690000000000000    0.2971000000000000    0.2809000000000000 O2-

test_files/INCAR.amin

@@ -0,0 +1,26 @@
+ADDGRID = False
+ALGO = Fast
+AMIN = 0.1
+EDIFF = 1e-06
+ENAUG = 1360
+ENCUT = 680
+ISMEAR = -5
+ISPIN = 2
+KPAR = 4
+KSPACING = 0.22
+LAECHG = False
+LASPH = False
+LCHARG = False
+LELF = False
+LMAXMIX = 6
+LORBIT = 11
+LREAL = Auto
+LVTOT = False
+LWAVE = False
+MAGMOM = 99*0.6
+NCORE = 1
+NELM = 100
+NSIM = 8
+NSW = 0
+PREC = Accurate
+SIGMA = 0.05

test_files/unconverged/vasprun.xml.electronic2

@@ -28,7 +28,7 @@
   <i name="ENCUT">    520.00000000</i>
   <v name="MAGMOM">      5.00000000      5.00000000      5.00000000      5.00000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000</v>
   <i type="string" name="LREAL"> Auto</i>
-  <i type="int" name="ISMEAR">    -5</i>
+  <i type="int" name="ISMEAR">    0</i>
   <i name="SIGMA">      0.05000000</i>
   <i type="int" name="LMAXMIX">     4</i>
   <i type="logical" name="LWAVE"> F  </i>
@@ -215,7 +215,7 @@
    <i type="int" name="NMIN">     0</i>
    <i name="EREF">      0.00000000</i>
    <separator name="electronic smearing" >
-    <i type="int" name="ISMEAR">    -5</i>
+    <i type="int" name="ISMEAR">    0</i>
     <i name="SIGMA">      0.05000000</i>
     <i name="KSPACING">      0.50000000</i>
     <i type="logical" name="KGAMMA"> T  </i>

test_files/unconverged/vasprun.xml.electronic_metagga_fast

@@ -23,7 +23,7 @@
   <i name="ENCUT">    520.00000000</i>
   <v name="MAGMOM">      5.00000000      5.00000000      5.00000000      5.00000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000</v>
   <i type="string" name="LREAL"> Auto</i>
-  <i type="int" name="ISMEAR">    -5</i>
+  <i type="int" name="ISMEAR">    0</i>
   <i name="SIGMA">      0.05000000</i>
   <i type="int" name="LMAXMIX">     4</i>
   <i type="logical" name="LWAVE"> F  </i>
@@ -210,7 +210,7 @@
    <i type="int" name="NMIN">     0</i>
    <i name="EREF">      0.00000000</i>
    <separator name="electronic smearing" >
-    <i type="int" name="ISMEAR">    -5</i>
+    <i type="int" name="ISMEAR">    0</i>
     <i name="SIGMA">      0.05000000</i>
     <i name="KSPACING">      0.50000000</i>
     <i type="logical" name="KGAMMA"> T  </i>

test_files/unconverged/vasprun.xml.electronic_normal

@@ -23,7 +23,7 @@
   <i name="ENCUT">    520.00000000</i>
   <v name="MAGMOM">      5.00000000      5.00000000      5.00000000      5.00000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000      0.60000000</v>
   <i type="string" name="LREAL"> Auto</i>
-  <i type="int" name="ISMEAR">    -5</i>
+  <i type="int" name="ISMEAR">    0</i>
   <i name="SIGMA">      0.05000000</i>
   <i type="int" name="LMAXMIX">     4</i>
   <i type="logical" name="LWAVE"> F  </i>
@@ -210,7 +210,7 @@
    <i type="int" name="NMIN">     0</i>
    <i name="EREF">      0.00000000</i>
    <separator name="electronic smearing" >
-    <i type="int" name="ISMEAR">    -5</i>
+    <i type="int" name="ISMEAR">    0</i>
     <i name="SIGMA">      0.05000000</i>
     <i name="KSPACING">      0.50000000</i>
     <i type="logical" name="KGAMMA"> T  </i>

test_files/unconverged/vasprun.xml.scan

@@ -28,7 +28,7 @@
   <i name="ENCUT">    520.00000000</i>
   <v name="MAGMOM">      0.00000000      0.00000000</v>
   <i type="string" name="LREAL"> Auto</i>
-  <i type="int" name="ISMEAR">    -5</i>
+  <i type="int" name="ISMEAR">    0</i>
   <i name="SIGMA">      0.05000000</i>
   <i type="logical" name="LWAVE"> F  </i>
   <i type="logical" name="LCHARG"> T  </i>
@@ -485,7 +485,7 @@
    <i type="int" name="NMIN">     0</i>
    <i name="EREF">      0.00000000</i>
    <separator name="electronic smearing" >
-    <i type="int" name="ISMEAR">    -5</i>
+    <i type="int" name="ISMEAR">    0</i>
     <i name="SIGMA">      0.05000000</i>
     <i name="KSPACING">      0.50000000</i>
     <i type="logical" name="KGAMMA"> T  </i>