Merge branch 'develop' into refactor/310_plus_typehints

jlnav · jlnav · commit af65de6476a0 · 2025-06-11T14:05:09.000-05:00
diff --git a/.github/workflows/basic.yml b/.github/workflows/basic.yml
@@ -129,4 +129,4 @@ jobs:
         runs-on: ubuntu-latest
         steps:
         - uses: actions/checkout@v4
-        - uses: crate-ci/typos@v1.31.2
+        - uses: crate-ci/typos@v1.33.1
diff --git a/.github/workflows/extra.yml b/.github/workflows/extra.yml
@@ -180,4 +180,4 @@ jobs:
         runs-on: ubuntu-latest
         steps:
         - uses: actions/checkout@v4
-        - uses: crate-ci/typos@v1.31.2
+        - uses: crate-ci/typos@v1.33.1
diff --git a/install/gen_deps_environment.yml b/install/gen_deps_environment.yml
@@ -13,3 +13,5 @@ dependencies:
   - DFO-LS
   - mpmath
   - ax-platform>=0.5.0
+  - petsc
+  - petsc4py
diff --git a/install/testing_requirements.txt b/install/testing_requirements.txt
@@ -1,11 +1,11 @@
 flake8==7.2.0
 coverage>=7.5
-pytest==8.3.5
+pytest==8.4.0
 pytest-cov==6.1.1
-pytest-timeout==2.3.1
+pytest-timeout==2.4.0
 mock==5.2.0
 python-dateutil==2.9.0.post0
 anyio==4.9.0
-matplotlib==3.10.1
+matplotlib==3.10.3
 mpmath==1.3.0
 rich==14.0.0
diff --git a/libensemble/gen_funcs/persistent_aposmm.py b/libensemble/gen_funcs/persistent_aposmm.py
@@ -25,7 +25,7 @@
 def cdist(XA, XB, metric="euclidean"):
     """Compute the pairwise Euclidean distances"""
 
-    # Just so dont have to change the call
+    # Just so we don't have to change the call
     if metric != "euclidean":
         raise ValueError("Only 'euclidean' metric is supported in this implementation.")
 
@@ -120,6 +120,11 @@ def aposmm(H, persis_info, gen_specs, libE_info):
         `test_persistent_aposmm_with_grad <https://github.com/Libensemble/libensemble/blob/develop/libensemble/tests/regression_tests/test_persistent_aposmm_with_grad.py>`_
         for an example where past function values are given to libEnsemble/APOSMM.
 
+    .. seealso::
+
+        `test_aposmm_starting_point_finder <https://github.com/Libensemble/libensemble/blob/develop/libensemble/tests/unit_tests/test_aposmm_starting_point_finder.py>`_
+        for an example the APOSMM r_k radius logic is adjusted to produce a certain number of localopt starting points.
+
     """
     """
     Description of intermediate variables in aposmm:
diff --git a/libensemble/tests/functionality_tests/test_uniform_sampling_cancel.py b/libensemble/tests/functionality_tests/test_uniform_sampling_cancel.py
@@ -88,38 +88,38 @@ def create_H0(persis_info, gen_specs, sim_max):
     sim_max = 500
     exit_criteria = {"sim_max": sim_max, "wallclock_max": 300}
 
-    aspec1 = {
+    a_spec_1 = {
         "alloc_f": gswf,
         "user": {
             "batch_mode": True,
             "num_active_gens": 1,
         },
     }
 
-    aspec2 = {
+    a_spec_2 = {
         "alloc_f": gswf,
         "user": {
             "batch_mode": True,
             "num_active_gens": 2,
         },
     }
 
-    aspec3 = {
+    a_spec_3 = {
         "alloc_f": fast_gswf,
         "user": {},
     }
 
-    aspec4 = {
+    a_spec_4 = {
         "alloc_f": ensure_one_active_gen,
         "user": {},
     }
 
-    aspec5 = {
+    a_spec_5 = {
         "alloc_f": give_pregenerated_sim_work,
         "user": {},
     }
 
-    allocs = {1: aspec1, 2: aspec2, 3: aspec3, 4: aspec4, 5: aspec5}
+    allocs = {1: a_spec_1, 2: a_spec_2, 3: a_spec_3, 4: a_spec_4, 5: a_spec_5}
 
     if is_manager:
         print("Testing cancellations with non-persistent gen functions")
diff --git a/libensemble/tests/regression_tests/test_persistent_aposmm_pounders.py b/libensemble/tests/regression_tests/test_persistent_aposmm_pounders.py
@@ -17,7 +17,6 @@
 # TESTSUITE_COMMS: local mpi tcp
 # TESTSUITE_NPROCS: 4
 # TESTSUITE_EXTRA: true
-# TESTSUITE_EXCLUDE: true
 
 import multiprocessing
 import sys
diff --git a/libensemble/tests/regression_tests/test_persistent_aposmm_tao_blmvm.py b/libensemble/tests/regression_tests/test_persistent_aposmm_tao_blmvm.py
@@ -15,7 +15,6 @@
 # TESTSUITE_COMMS: local mpi tcp
 # TESTSUITE_NPROCS: 4
 # TESTSUITE_EXTRA: true
-# TESTSUITE_EXCLUDE: true
 
 import multiprocessing
 import sys
diff --git a/libensemble/tests/regression_tests/test_persistent_aposmm_tao_nm.py b/libensemble/tests/regression_tests/test_persistent_aposmm_tao_nm.py
@@ -15,7 +15,6 @@
 # TESTSUITE_COMMS: local mpi tcp
 # TESTSUITE_NPROCS: 4
 # TESTSUITE_EXTRA: true
-# TESTSUITE_EXCLUDE: true
 
 import multiprocessing
 import sys
diff --git a/libensemble/tests/regression_tests/test_with_app_persistent_aposmm_tao_nm.py b/libensemble/tests/regression_tests/test_with_app_persistent_aposmm_tao_nm.py
@@ -19,7 +19,6 @@
 # TESTSUITE_COMMS: local
 # TESTSUITE_NPROCS: 4
 # TESTSUITE_EXTRA: true
-# TESTSUITE_EXCLUDE: true
 
 import multiprocessing
 import sys
diff --git a/libensemble/tests/unit_tests/test_aposmm_starting_point_finder.py b/libensemble/tests/unit_tests/test_aposmm_starting_point_finder.py
@@ -0,0 +1,120 @@
+import numpy as np
+from libensemble.gen_funcs.persistent_aposmm import (
+    initialize_APOSMM,
+    update_history_dist,
+    decide_where_to_start_localopt,
+)
+from libensemble.sim_funcs.six_hump_camel import six_hump_camel_func
+from libensemble.tests.regression_tests.support import six_hump_camel_minima as known_minima
+
+
+def setup_history_and_find_rk(n_s, num_to_start, lb, ub, f_vals, x_points):
+    """
+    Populate the history array H with n_s points and bisect over r_k to find a value
+    producing num_to_start local optimization start points using decide_where_to_start_localopt.
+
+    Parameters:
+    - n_s (int): Number of initial sample points.
+    - num_to_start (int): Desired number of starting points for local optimization.
+    - lb, ub (np.ndarray): Lower and upper bounds of the domain.
+    - f_vals (np.ndarray): Function values at each x_point.
+    - x_points (np.ndarray): n_s x d array of sample points.
+
+    Returns:
+    - H (np structured array): Updated history array.
+    - rk_final (float): Value of r_k yielding num_to_start local opt starts.
+    - inds_to_start (list): Indices in H to start local optimization.
+    """
+
+    assert x_points.shape[0] == n_s
+    assert f_vals.shape[0] == n_s
+
+    n = x_points.shape[1]
+
+    H = np.zeros(
+        n_s,
+        dtype=[
+            ("sim_id", int),
+            ("x", float, n),
+            ("x_on_cube", float, n),
+            ("f", float),
+            ("local_pt", bool),
+            ("sim_ended", bool),
+        ],
+    )
+
+    # Setup history
+    for i in range(n_s):
+        H[i]["x"] = x_points[i]
+        H[i]["sim_id"] = i
+        H[i]["x_on_cube"] = (x_points[i] - lb) / (ub - lb)
+        H[i]["f"] = f_vals[i]
+        H[i]["local_pt"] = False
+        H[i]["sim_ended"] = True  # Ensure point is considered by distance function
+
+    user_specs = {"lb": lb, "ub": ub, "initial_sample_size": n_s, "localopt_method": None}
+    local_H = initialize_APOSMM(H, user_specs, {"comm": None})[-1]
+    local_H = local_H[:n_s]  # Use only the required number of entries
+
+    update_history_dist(local_H, n)
+
+    # Search to find r_k that yields exactly num_to_start start points
+    r_low, r_high = 1e-5, 2.0  # Conservative initial bounds
+    rk_final = None
+    tol = 1e-5
+
+    while r_high - r_low > tol:
+        r_mid = (r_low + r_high) / 2.0
+        inds_to_start = decide_where_to_start_localopt(local_H, n, n_s, r_mid)
+
+        if len(inds_to_start) < num_to_start:
+            r_high = r_mid
+        elif len(inds_to_start) > num_to_start:
+            r_low = r_mid
+        else:
+            rk_final = r_mid
+            break
+
+    # Final decision (in case we didn't hit num_to_start exactly)
+    if rk_final is None:
+        rk_final = (r_low + r_high) / 2.0
+        inds_to_start = decide_where_to_start_localopt(local_H, n, n_s, rk_final)
+
+    return local_H, rk_final, inds_to_start
+
+
+if __name__ == "__main__":
+
+    # Define domain bounds
+    lb = np.array([-3.0, -2.0])
+    ub = np.array([3.0, 2.0])
+    dim = len(lb)
+
+    # Number of sample points and desired number of start points
+    num_samples = 1000
+    num_to_start = 6
+
+    # Generate random sample points uniformly in the [lb,ub] box
+    x_points = lb + (ub - lb) * np.random.uniform(size=(num_samples, dim))
+
+    # Evaluate six-hump camel function at each point
+    f_vals = np.array([six_hump_camel_func(x) for x in x_points])
+
+    # Call the history setup and bisection function
+    H, rk_final, inds_to_start = setup_history_and_find_rk(num_samples, num_to_start, lb, ub, f_vals, x_points)
+
+    assert len(inds_to_start) == num_to_start, f"Found {len(inds_to_start)} starting points instead of {num_to_start}"
+
+    starting_pts = H["x"][inds_to_start]
+    sorted_starting = starting_pts[np.lexsort(starting_pts.T[::-1])]
+    sorted_known = known_minima[np.lexsort(known_minima.T[::-1])]
+
+    print(
+        f"For this problem, we know the minima.\n"
+        f"The chosen starting points:\n{sorted_starting}\n"
+        f"should be close to the known minima:\n{sorted_known}"
+    )
+
+    # Output results
+    print(f"Chosen r_k: {rk_final:.6f}")
+    print(f"Indices to start local optimization (num_to_start={num_to_start}): {inds_to_start}")
diff --git a/libensemble/tools/alloc_support.py b/libensemble/tools/alloc_support.py
@@ -292,7 +292,7 @@ def gen_work(self, wid, H_fields, H_rows, persis_info, **libE_info):
         return work
 
     def _filter_points(self, H_in, pt_filter, low_bound):
-        """Returns H and pt_filter filted by lower bound
+        """Returns H and pt_filter filtered by lower bound
 
         :param pt_filter: (Optional) Boolean array filtering expected returned points in ``H``.
         :param low_bound: (Optional) Lower bound for testing all returned.
