refactor random_permutations for simplicity (#300)

Refactor random_permutations ensemble strategy to simplify code

[ committed by @ankona ]
[ reviewed by @billschereriii ]

doc/changelog.rst

@@ -37,6 +37,7 @@ A full list of changes and detailed notes can be found below:
 
 Detailed notes
 
+- Simplify code in `random_permutations` parameter generation strategy (PR300_)
 - Remove wait time associated with Experiment launch summary (PR298_)
 - Update Redis conf file to conform with Redis v7.0.5 conf file (PR293_)
 - Migrate from redis-py-cluster to redis-py for cluster status checks (PR292_)
@@ -55,6 +56,7 @@ argument name is still `interface` for backward compatibility reasons. (PR281_)
 - Typehints have been added to public APIs. A makefile target to execute static 
 analysis with mypy is available `make check-mypy`. (PR295_)
 
+.. _PR300: https://github.com/CrayLabs/SmartSim/pull/300
 .. _PR298: https://github.com/CrayLabs/SmartSim/pull/298
 .. _PR293: https://github.com/CrayLabs/SmartSim/pull/293
 .. _PR292: https://github.com/CrayLabs/SmartSim/pull/292

smartsim/entity/strategies.py

@@ -56,21 +56,10 @@ def step_values(
 def random_permutations(
     param_names: t.List[str], param_values: t.List[t.List[str]], n_models: int = 0
 ) -> t.List[t.Dict[str, str]]:
-    # first, check if we've requested more values than possible.
-    perms = list(product(*param_values))
-    if n_models >= len(perms):
-        return create_all_permutations(param_names, param_values)
-    else:
-        permutations: t.List[t.Dict[str, str]] = []
-        permutation_strings = set()
-        while len(permutations) < n_models:
-            model_dict = dict(
-                zip(
-                    param_names,
-                    map(lambda x: x[random.randint(0, len(x) - 1)], param_values),
-                )
-            )
-            if str(model_dict) not in permutation_strings:
-                permutation_strings.add(str(model_dict))
-                permutations.append(model_dict)
-        return permutations
+    permutations = create_all_permutations(param_names, param_values)
+
+    # sample from available permutations if n_models is specified
+    if n_models and n_models < len(permutations):
+        permutations = random.sample(permutations, n_models)
+
+    return permutations