Option to create sample points outside gen (#1697)

Author: shuds13

libensemble/libE.py

@@ -242,9 +242,11 @@ def libE(
     ]
     exit_criteria = specs_dump(ensemble.exit_criteria, by_alias=True, exclude_none=True)
 
-    # Restore the generator object (don't use serialized version)
+    # Restore objects that don't survive serialization via model_dump
     if hasattr(ensemble.gen_specs, "generator") and ensemble.gen_specs.generator is not None:
         gen_specs["generator"] = ensemble.gen_specs.generator
+    if hasattr(ensemble.gen_specs, "vocs") and ensemble.gen_specs.vocs is not None:
+        gen_specs["vocs"] = ensemble.gen_specs.vocs
 
     # Extract platform info from settings or environment
     platform_info = get_platform(libE_specs)

libensemble/specs.py

@@ -216,6 +216,16 @@ class GenSpecs(BaseModel):
     batch sizes via ``gen_specs["user"]`` or other methods.
     """
 
+    initial_sample_method: str | None = None
+    """
+    Method for producing initial sample points before starting the generator.
+    If None (default), the generator is responsible for producing its own initial
+    sample via ``suggest()``. Set to ``"uniform"`` to have libEnsemble generate
+    uniform random samples from VOCS bounds, evaluate them, and ingest the results
+    into the generator before optimization begins. The number of sample points is
+    determined by ``initial_batch_size``.
+    """
+
     threaded: bool | None = False
     """
     Instruct Worker process to launch user function to a thread.

libensemble/tests/regression_tests/test_xopt_EI_initial_sample.py

@@ -0,0 +1,87 @@
+"""
+Tests libEnsemble with Xopt ExpectedImprovementGenerator using
+initial_sample_method="uniform" to produce initial sample points.
+
+EI requires pre-evaluated data before it can suggest points. This test
+verifies that setting initial_sample_method="uniform" in GenSpecs causes
+libEnsemble to generate uniform random samples, evaluate them through
+the sim, and ingest results into the generator before optimization begins.
+
+Execute via one of the following commands (e.g. 4 workers):
+   mpiexec -np 5 python test_xopt_EI_initial_sample.py
+   python test_xopt_EI_initial_sample.py -n 4
+
+"""
+
+# Do not change these lines - they are parsed by run-tests.sh
+# TESTSUITE_COMMS: local
+# TESTSUITE_NPROCS: 4
+# TESTSUITE_EXTRA: true
+# TESTSUITE_EXCLUDE: true
+
+import numpy as np
+from gest_api.vocs import VOCS
+from xopt.generators.bayesian.expected_improvement import ExpectedImprovementGenerator
+
+from libensemble import Ensemble
+from libensemble.alloc_funcs.start_only_persistent import only_persistent_gens as alloc_f
+from libensemble.specs import AllocSpecs, ExitCriteria, GenSpecs, LibeSpecs, SimSpecs
+
+
+def xtest_sim(H, persis_info, sim_specs, _):
+    """y1 = x2, c1 = x1"""
+    batch = len(H)
+    H_o = np.zeros(batch, dtype=sim_specs["out"])
+    for i in range(batch):
+        H_o["y1"][i] = H["x2"][i]
+        H_o["c1"][i] = H["x1"][i]
+    return H_o, persis_info
+
+
+if __name__ == "__main__":
+
+    batch_size = 4
+
+    libE_specs = LibeSpecs(gen_on_manager=True, nworkers=batch_size)
+    libE_specs.reuse_output_dir = True
+
+    vocs = VOCS(
+        variables={"x1": [0, 1.0], "x2": [0, 10.0]},
+        objectives={"y1": "MINIMIZE"},
+        constraints={"c1": ["GREATER_THAN", 0.5]},
+        constants={"constant1": 1.0},
+    )
+
+    gen = ExpectedImprovementGenerator(vocs=vocs)
+
+    # NO pre-ingested data — libEnsemble handles initial sampling.
+    gen_specs = GenSpecs(
+        generator=gen,
+        initial_batch_size=batch_size,
+        initial_sample_method="uniform",
+        batch_size=batch_size,
+        vocs=vocs,
+    )
+
+    sim_specs = SimSpecs(
+        sim_f=xtest_sim,
+        vocs=vocs,
+    )
+
+    alloc_specs = AllocSpecs(alloc_f=alloc_f)
+    exit_criteria = ExitCriteria(sim_max=20)
+
+    workflow = Ensemble(
+        libE_specs=libE_specs,
+        sim_specs=sim_specs,
+        alloc_specs=alloc_specs,
+        gen_specs=gen_specs,
+        exit_criteria=exit_criteria,
+    )
+
+    H, _, _ = workflow.run()
+
+    if workflow.is_manager:
+        print(f"Completed {len(H)} simulations")
+        assert len(H) >= 8, f"Expected at least 8 sims, got {len(H)}"
+        print("Test passed")

libensemble/utils/runners.py

@@ -158,6 +158,19 @@ def _start_generator_loop(self, tag, Work, H_in):
         self._convert_initial_ingest(H_in)
         return self._loop_over_gen(tag, Work, H_in)
 
+    def _create_initial_sample(self, sample_method, num_points):
+        """Create initial sample points using the specified sampling method."""
+        from libensemble.gen_classes.sampling import UniformSample
+
+        vocs = self.specs.get("vocs")
+        samplers = {
+            "uniform": UniformSample,
+        }
+        if sample_method not in samplers:
+            raise ValueError(f"Unknown initial_sample_method: {sample_method!r}. Supported: {list(samplers.keys())}")
+        sampler = samplers[sample_method](vocs=vocs)
+        return sampler.suggest(num_points)
+
     def _persistent_result(self, calc_in, persis_info, libE_info):
         """Setup comms with manager, setup gen, loop gen to completion, return gen's results"""
         self.ps = PersistentSupport(libE_info, EVAL_GEN_TAG)
@@ -166,14 +179,32 @@ def _persistent_result(self, calc_in, persis_info, libE_info):
         if calc_in is not None and len(calc_in) > 0:
             self._convert_initial_ingest(calc_in)
 
-        # libE gens will hit the following line, but list_dicts_to_np will passthrough if the output is a numpy array
-        H_out = list_dicts_to_np(
-            self._get_initial_suggest(libE_info),
-            dtype=self.specs.get("out"),
-            mapping=getattr(self.gen, "variables_mapping", {}),
-        )
-        tag, Work, H_in = self.ps.send_recv(H_out)  # evaluate the initial sample
-        final_H_out = self._start_generator_loop(tag, Work, H_in)
+        sample_method = self.specs.get("initial_sample_method")
+        if sample_method is not None:
+            # libEnsemble produces the initial sample, evaluates it, and
+            # ingests results into the generator before optimization begins.
+            initial_batch = self.specs.get("initial_batch_size")
+            if not initial_batch:
+                raise ValueError("initial_sample_method requires initial_batch_size to be set in GenSpecs.")
+            H_sample = list_dicts_to_np(
+                self._create_initial_sample(sample_method, initial_batch),
+                dtype=self.specs.get("out"),
+                mapping=getattr(self.gen, "variables_mapping", {}),
+            )
+            tag, Work, H_in = self.ps.send_recv(H_sample)
+            self._convert_initial_ingest(H_in)
+            # Generator now has evaluated data — enter the normal loop
+            final_H_out = self._loop_over_gen(tag, Work, H_in)
+        else:
+            # Generator handles its own initial sampling
+            H_out = list_dicts_to_np(
+                self._get_initial_suggest(libE_info),
+                dtype=self.specs.get("out"),
+                mapping=getattr(self.gen, "variables_mapping", {}),
+            )
+            tag, Work, H_in = self.ps.send_recv(H_out)  # evaluate the initial sample
+            final_H_out = self._start_generator_loop(tag, Work, H_in)
+
         self.gen.finalize()
         return final_H_out, FINISHED_PERSISTENT_GEN_TAG