Merge pull request #37 from isi-usc-edu/max-radin/overlaps

Updating LRE script to allow for overlaps to be specified per task

README.md

@@ -40,10 +40,11 @@ Contact the repository maintainers to get these.
 
 The [`scripts/compute_all_LREs_script.py`](scripts/compute_all_LREs_script.py) script can be used to generate LQREs.
 In order to run this script, first ensure that the qb-gsee-benchmark package is installed as described above.
-Then, two additional files are required:
+Then, several files are required:
 
 * A PPK file containing a key to access the SFTP server that provides access to FCIDUMP files as described above.
-* A configuration file similar to [`scripts/LRE_config.json`](scripts/LRE_config.json) which specifies algorithm parameters, solver UUID, and other information.
+* A configuration file similar to [`scripts/LRE_config.json`](scripts/LRE_config.json) or [`scripts/LRE_config.json`](scripts/LRE_config_overlaps.json) which specifies algorithm parameters, solver UUID, and other information.
+* If the `algorithm_parameters` object in the config file specifies a value for `overlap_csv`, then this CSV file must also be present. See for example [`scripts/overlaps.csv`](scripts/overlaps.csv)
 
 The paths to these files must be passed as arguments to the script, for example:
 ```bash

scripts/LRE_config.json

@@ -5,6 +5,7 @@
     "df_threshold": 1e-3,
     "max_orbitals": 32
   },
+  "algorithm_description": "Double factorized QPE resource estimates based on methodology of arXiv:2406.06335. Note that the truncation error is not included in the error bounds and that the SCF compute time is not included in the preprocessing time. Also note that unlike arXiv:2406.06335, the ground-state overlaps are not computed from DMRG but instead set to a nominal value.",
   "contact_info": [
     {
       "name": "Max Radin",

scripts/LRE_config_overlaps.json

@@ -0,0 +1,16 @@
+{
+  "solver_uuid": "f2d73e1f-3058-43c4-a634-b6c267c84ff1",
+  "algorithm_parameters": {
+    "overlap_csv": "overlaps.csv",
+    "df_threshold": 1e-3,
+    "max_orbitals": 32
+  },
+  "algorithm_description": "Double factorized QPE resource estimates based on methodology of arXiv:2406.06335. Note that the truncation error is not included in the error bounds and that the SCF compute time is not included in the preprocessing time. Ground-state overlap is taken to be that estimated for the dominant CSF as estimated by DMRG and that this DMRG runtime is not included in the classical compute costs.",
+  "contact_info": [
+    {
+      "name": "Max Radin",
+      "email": "[email protected]",
+      "institution": "L3Harris"
+    }
+  ]
+}

scripts/compute_all_LREs_script.py

@@ -22,11 +22,13 @@
 import math
 import os
 import sys
+from collections import defaultdict
 from importlib.metadata import version
 from typing import Any
 from urllib.parse import urlparse
 from uuid import uuid4
 
+import pandas as pd
 from pyLIQTR.utils.resource_analysis import estimate_resources
 
 from qb_gsee_benchmark.qre import get_df_qpe_circuit
@@ -53,10 +55,32 @@ def get_lqre(
     num_hams = len(problem_instance["tasks"])
     logging.info(f"contains {num_hams} associated Hamiltonians.")
 
+    if (
+        "overlap" in config["algorithm_parameters"]
+        and "overlap_csv" in config["algorithm_parameters"]
+    ):
+        raise ValueError("Config cannot specify both 'overlap' and 'overlap_csv'.")
+
+    if config["algorithm_parameters"].get("overlap_csv"):
+        overlap_df = pd.read_csv(config["algorithm_parameters"]["overlap_csv"])
+        overlaps = {
+            row["task_uuid"]: row["overlap"] for index, row in overlap_df.iterrows()
+        }
+    else:
+        overlaps = defaultdict(lambda: config["algorithm_parameters"]["overlap"])
+
     solution_data: list[dict[str, Any]] = []
     results: dict[str, Any] = {}
 
     for task in problem_instance["tasks"]:
+        if not overlaps.get(task["task_uuid"]):
+            logging.info(
+                f"Skipping task {task['task_uuid']} because no overlap was provided."
+            )
+            continue
+
+        logging.info(f"Analyzing task {task['task_uuid']}...")
+
         num_supporting_files = len(task["supporting_files"])
         logging.info(f"number of supporting files: {num_supporting_files}")
 
@@ -113,12 +137,12 @@ def get_lqre(
                 fci=fci,
                 error_tolerance=error_tolerance,
                 failure_tolerance=failure_tolerance,
-                square_overlap=config["algorithm_parameters"]["overlap"] ** 2,
+                square_overlap=overlaps[task["task_uuid"]] ** 2,
                 df_threshold=config["algorithm_parameters"]["df_threshold"],
             )
             circuit_generation_end_time = datetime.datetime.now()
             logging.info(
-                f"Circuit initialization time: {(circuit_generation_end_time - circuit_generation_end_time).total_seconds()} seconds."
+                f"Circuit initialization time: {(circuit_generation_end_time - circuit_generation_start_time).total_seconds()} seconds."
             )
             logging.info(f"Estimating logical resources...")
             resource_estimation_start_time = datetime.datetime.now()
@@ -170,7 +194,7 @@ def get_lqre(
         "solver_short_name": "DF QPE",
         "compute_hardware_type": "quantum_computer",
         "algorithm_details": {
-            "algorithm_description": "Double factorized QPE resource estimates based on methodology of arXiv:2406.06335. Note that the truncation error is not included in the error bounds and that the SCF compute time is not included in the preprocessing time. Also note that unlike arXiv:2406.06335, the ground-state overlaps are not computed from DMRG but instead set to a nominal value.",
+            "algorithm_description": config["algorithm_description"],
             "algorithm_parameters": config["algorithm_parameters"],
         },
         "software_details": [

scripts/overlaps.csv

@@ -0,0 +1,4 @@
+task_uuid,overlap
+be3812d3-5975-4736-8a23-96d64e50e66a,0.7085974535652702
+472bbfb1-c081-4df6-b966-32ce1194ab8c,0.9774145158534091
+a41a284c-7baa-438e-9323-532091c7a78f,0.9985372277228635

setup.cfg

@@ -30,6 +30,7 @@ install_requires =
     openfermionpyscf
     numpy
     paramiko
+    pandas
 
 
 [options.packages.find]