Adding CIS, CISD and other options for GAS-ACI calculations #355

huangm29 · 2023-10-25T21:33:02Z

Description

Adding options for initial guesses of ACI using GAS. This PR was previously #341 (comment)

User Notes

Adding CIS, CISD, and CID options for ACTIVE_REF_TYPE, which are initial guess of GAS-ACI
Adding GAS_REF_COUNT option which can be combined with GAS option of ACTIVE_REF_TYPE to specify a full GAS for initial guess of GAS-ACI
Slightly optimize the GAS-ACI code

Checklist

Re-enable GASACI-2 caused by previous bug, and also use it to test the GAS-CISD option
Added GASACI-6 test case for GAS-CIS option
Added GASACI-7 test case for GAS_REF_COUNT option
Modified GASACI-3 and GASACI-4 for GAS-CISD options
Ready to go!

fevangelista · 2023-10-27T18:17:41Z

forte/register_forte_options.py

@@ -78,9 +78,11 @@ def register_driver_options(options):
    options.add_double("MS", None, "Projection of spin onto the z axis")

    options.add_str(
-        "ACTIVE_REF_TYPE", "CAS", ["HF", "CAS", "GAS", "GAS_SINGLE", "CIS", "CID", "CISD", "DOCI"],
+        "ACTIVE_REF_TYPE", "CAS", ["HF", "CAS", "GAS", "GAS_SINGLE", "CIS", "CID", "CISD", "DOCI","GAS_CIS", "GAS_CISD", "GAS_CID"],


Why do we need "GAS_CIS", "GAS_CISD", "GAS_CID" when we already have "CIS", "CID", "CISD"? Can't the user just use these last ones and the code automatically understands if one is using GAS or not? Unless these are supposed to be both valid for GAS (e.g. "CIS" and "GAS_CIS" are both valid and give different results in a GAS computation) then this is potentially a confusing redundancy. Also, I would suggest using "GAS_SINGLE_DET" instead of "GAS_SINGLE".

fevangelista · 2023-10-27T18:19:01Z

forte/sci/aci.cc

+        options_->get_str("ACTIVE_REF_TYPE") == "GAS" or
+        options_->get_str("ACTIVE_REF_TYPE") == "GAS_CIS" or
+        options_->get_str("ACTIVE_REF_TYPE") == "GAS_CISD" or
+        options_->get_str("ACTIVE_REF_TYPE") == "GAS_CID") {
        gas_iteration_ = true;


This connects to the point above. The use of GAS should be triggered by the user passing a GAS space, not by choosing an additional keyword.

fevangelista · 2023-10-27T18:20:44Z

forte/sci/gasaci_build_F.cc

+
+                // gas_single_criterion_.first stores all allowed i,a for alpha electron transition
+                // from GAS_count_i to GAS_count_a
+                // For example, the current gas_configuration allows an alpha electron from GAS1


Suggested change

// For example, the current gas_configuration allows an alpha electron from GAS1

// For example, if the current gas_configuration allows an alpha electron from GAS1

fevangelista · 2023-10-27T18:21:30Z

forte/sci/gasaci_build_F.cc

@@ -157,32 +168,53 @@ void AdaptiveCI::get_gas_excited_determinants_sr(
            }

            // Generate aa excitations
-            for (auto& gas_count : std::get<0>(gas_double_criterion_)[gas_configuration]) {
+            for (auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :


Suggested change

for (auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :

for (const auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :

fevangelista · 2023-10-27T18:23:00Z

tests/methods/gasaci-7/input.dat

@@ -0,0 +1,46 @@
+# GAS ACI calculation for core excited state of COO


Suggested change

# GAS ACI calculation for core excited state of COO

# GAS ACI calculation for core excited state of CO

fevangelista · 2023-10-27T18:24:42Z

forte/sparse_ci/ci_reference.h

@@ -154,6 +157,9 @@ class CI_Reference {
    /// Build the complete GAS reference
    void build_gas_reference(std::vector<Determinant>& ref_space);

+    /// Build GAS-CIS/CID/CISD reference
+    void build_gas_ci_reference(std::vector<Determinant>& ref_space);
+
    /// Build single lowest energy state
    void build_gas_single(std::vector<Determinant>& ref_space);


Suggested change

void build_gas_single(std::vector<Determinant>& ref_space);

void build_gas_single_determinant(std::vector<Determinant>& ref_space);

I prefer the more verbose version so it's clear what single means. Propagate this change. I also see, the name of the option makes sense but it is not clear on its own.

fevangelista · 2023-10-27T18:26:57Z

forte/sparse_ci/ci_reference.cc

    //    for (const auto& det : ref_space) {
    //        outfile->Printf("\n    %s", str(det, nact_).c_str());
    //    }


Suggested change

// for (const auto& det : ref_space) {

// outfile->Printf("\n %s", str(det, nact_).c_str());

// }

fevangelista · 2023-10-27T18:27:48Z

forte/sci/gasaci_build_F.cc

@@ -803,24 +903,21 @@ void AdaptiveCI::get_gas_excited_determinants_core(
            }

            // Genearte bb excitations
-            for (auto& gas_count : std::get<1>(gas_double_criterion_)[gas_configuration]) {
+            for (auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :


Great use of unpacking!

Suggested change

for (auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :

for (const auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :

…dding_gas_options

huangm29 added 15 commits June 3, 2023 03:55

added gasci stuff

30ed0db

major implementations

b3a773a

revoke the changings in tdci

edc1f0b

adding a test cas with gas-cis

8487ce9

fixing a bug that generates wrong number of GAS

fb8121f

Merge remote-tracking branch 'upstream/master' into HEAD

f11c8d5

adding comments to the code

6ead8b3

fixing a bug

fdee8e4

minor corrections

08ffec9

bug fixing for gasaci-4

a5d6f19

adding a method which uses a small full gas as an initial guess

9f802c4

Merge remote-tracking branch 'origin/master' into HEAD

3781fa4

modify gasaci-2 to make sure it is no longer bugged

d742c22

modify tests.yaml to re-enable gasaci-2

105760a

modify gasaci-2 to test GAS_CISD option

d14201b

huangm29 requested review from marink2 and fevangelista October 25, 2023 22:33

fevangelista requested changes Nov 2, 2023

View reviewed changes

Merge branch 'main' of https://github.com/evangelistalab/forte into a…

462225d

…dding_gas_options

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adding CIS, CISD and other options for GAS-ACI calculations #355

Adding CIS, CISD and other options for GAS-ACI calculations #355

huangm29 commented Oct 25, 2023 •

edited

Loading

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

fevangelista Oct 27, 2023

	// For example, the current gas_configuration allows an alpha electron from GAS1
	// For example, if the current gas_configuration allows an alpha electron from GAS1

	for (auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :
	for (const auto& [gas_count_i, gas_count_j, gas_count_a, gas_count_b] :

		@@ -0,0 +1,46 @@
		# GAS ACI calculation for core excited state of COO

	void build_gas_single(std::vector<Determinant>& ref_space);
	void build_gas_single_determinant(std::vector<Determinant>& ref_space);

	// for (const auto& det : ref_space) {
	// outfile->Printf("\n %s", str(det, nact_).c_str());
	// }

Adding CIS, CISD and other options for GAS-ACI calculations #355

Are you sure you want to change the base?

Adding CIS, CISD and other options for GAS-ACI calculations #355

Conversation

huangm29 commented Oct 25, 2023 • edited Loading

Description

User Notes

Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

huangm29 commented Oct 25, 2023 •

edited

Loading