Changing experiment.algorithms to experiment.algorithm (#1014)

* changing experiment.algorithms to experiment.algorithm

* fixing building_experiment

* changing some more algorithms to algorithm

* fixing experiments variable name forgotten

* sending algorithms in benchmark creation of experiment

* changing algorithms in config files for tests

* change algorithm_name for experiment in benchmark

* exp.configuration had algorithms keys

* study test with algorithms

* add warning and make algorithms still work but deprecated

* Add backward compatibility

* disable too-many-locals for pylint

* fixing line too long

* stop sending None algorithm when there isnt one

* adding algorithms to test

* backward python_api algorithms

* Fixing upgrade for algorithms

* Adding check for setting algorithm or algorithms to None

* setting up instantiated_algorithms if both algorithms are None

* sending only algorithm to Experiment

* instanciated algorithms cannot be None by default

* fixing comment, restarting pipeline

* changing algorithms in docs

* adding check on upgrade algorithm

* verifying pre-commit

* removing algorithms that was added with rebase

* Empty commit to spawn github-actions...

* calling orion db upgrade directly instead of execute

* good import

* rebasing for tests

* fixing some rebased algorithms + rebase

Co-authored-by: Xavier Bouthillier <xavier.bouthillier@umontreal.ca>
Co-authored-by: Simnol <simon-olivier.duguay.1@ens.etsmtl.ca>

docs/src/plugins/install.rst

@@ -36,7 +36,7 @@ Next we define the file ``bayes.yaml`` as this
 
     experiment:
         name: orion-with-bayes
-        algorithms: BayesianOptimizer
+        algorithm: BayesianOptimizer
 
 Then call ``orion hunt`` with the configuration file.
 

docs/src/user/algorithms.rst

@@ -1,4 +1,4 @@
-.. _Setup Algorithms:
+.. _Setup algorithm:
 
 **********
 Algorithms
@@ -19,7 +19,7 @@ In a Oríon configuration YAML, define:
 .. code-block:: yaml
 
    experiment:
-       algorithms:
+       algorithm:
            gradient_descent:
                learning_rate: 0.1
 
@@ -51,7 +51,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             random:
                 seed: null
 
@@ -91,7 +91,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             gridsearch:
                 n_values: 100
 
@@ -140,7 +140,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             hyperband:
                 seed: null
                 repetitions: 1
@@ -196,7 +196,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             asha:
                 seed: null
                 num_rungs: null
@@ -235,7 +235,7 @@ Configuration
 .. code-block:: yaml
 
   experiment:
-    algorithms:
+    algorithm:
       bohb:
         min_points_in_model: 20
         top_n_percent: 15
@@ -285,7 +285,7 @@ Configuration
 .. code-block:: yaml
 
   experiment:
-    algorithms:
+    algorithm:
       dehb:
         seed: null
         mutation_factor: 0.5
@@ -330,7 +330,7 @@ Configuration
 
   experiment:
 
-    algorithms:
+    algorithm:
       pbt:
         population_size: 50
         generations: 10
@@ -400,7 +400,7 @@ Configuration
 
   experiment:
 
-    algorithms:
+    algorithm:
       pb2:
         population_size: 50
         generations: 10
@@ -459,7 +459,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             tpe:
                 seed: null
                 n_initial_points: 20
@@ -503,7 +503,7 @@ Configuration
 .. code-block:: yaml
 
   experiment:
-    algorithms:
+    algorithm:
       ax:
         seed: 1234
         n_initial_trials: 5,
@@ -573,7 +573,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             EvolutionES:
                 seed: null
                 repetitions: 1
@@ -626,7 +626,7 @@ Configuration
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             MOFA:
                seed: null
                index: 1
@@ -654,7 +654,7 @@ a library of algorithms for hyperparameter search.
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             nevergrad:
                 seed: null
                 budget: 1000
@@ -692,7 +692,7 @@ Configuration
 .. code-block:: yaml
 
    experiment:
-      algorithms:
+      algorithm:
          hebo:
             seed: 1234
             parameters:

docs/src/user/config.rst

@@ -93,7 +93,7 @@ Full Example of Global Configuration
         type: pickleddb
 
     experiment:
-        algorithms:
+        algorithm:
             random:
                 seed: None
         max_broken: 3
@@ -205,7 +205,7 @@ Experiment
 .. code-block:: yaml
 
     experiment:
-        algorithms:
+        algorithm:
             random:
                 seed: None
         max_broken: 3
@@ -320,7 +320,7 @@ working_dir
 
 .. _config_experiment_algorithms:
 
-algorithms
+algorithm
 ~~~~~~~~~~
 
 :Type: dict
@@ -346,7 +346,7 @@ strategy
 :Env var:
 :Description:
     (DEPRECATED) This argument will be removed in v0.4. Parallel strategies are now handled by
-    algorithms directly and should be set in algorithm configuration when they support it.
+    algorithm directly and should be set in algorithm configuration when they support it.
 
 ----
 

examples/dask_example.py

@@ -56,7 +56,7 @@ def hpo(n_workers=16):
             "tol": "loguniform(1e-4, 1e-1, precision=None)",
             "class_weight": "choices([None, 'balanced'])",
         },
-        algorithms={"random": {"seed": 1}},
+        algorithm={"random": {"seed": 1}},
     )
 
     with experiment.tmp_executor("dask", n_workers=n_workers):

examples/tutorials/code_1_python_api.py

@@ -104,7 +104,7 @@ def rosenbrock(x, noise=None):
 experiment = build_experiment(
     "tpe-rosenbrock",
     space=space,
-    algorithms={"tpe": {"n_initial_points": 5}},
+    algorithm={"tpe": {"n_initial_points": 5}},
     storage=storage,
 )
 

examples/tutorials/code_2_hyperband_checkpoint.py

@@ -304,7 +304,7 @@ def run_hpo():
             "weight_decay": "loguniform(1e-10, 1e-2)",
             "gamma": "loguniform(0.97, 1)",
         },
-        algorithms={
+        algorithm={
             "hyperband": {
                 "seed": 1,
                 "repetitions": 5,

src/orion/benchmark/__init__.py

@@ -212,9 +212,7 @@ def get_experiments(self, silent=True):
             for _, exp in study.get_experiments():
                 exp_column = dict()
                 stats = exp.stats
-                exp_column["Algorithm"] = list(exp.configuration["algorithms"].keys())[
-                    0
-                ]
+                exp_column["Algorithm"] = list(exp.configuration["algorithm"].keys())[0]
                 exp_column["Experiment Name"] = exp.name
                 exp_column["Number Trial"] = len(exp.fetch_trials())
                 exp_column["Best Evaluation"] = stats.best_evaluation
@@ -405,7 +403,7 @@ def setup_experiments(self):
                 experiment = create_experiment(
                     experiment_name,
                     space=space,
-                    algorithms=algorithm.experiment_algorithm,
+                    algorithm=algorithm.experiment_algorithm,
                     max_trials=max_trials,
                     storage=self.benchmark.storage,
                     executor=executor,
@@ -430,7 +428,7 @@ def status(self):
         for _, experiment in self.get_experiments():
             trials = experiment.fetch_trials()
 
-            algorithm_name = list(experiment.configuration["algorithms"].keys())[0]
+            algorithm_name = list(experiment.configuration["algorithm"].keys())[0]
 
             if algorithm_tasks.get(algorithm_name, None) is None:
                 task_state = {
@@ -482,7 +480,7 @@ def get_experiments(self, algorithms=None):
         for repetition_index, experiment in self.experiments_info:
             if (
                 algorithms is None
-                or list(experiment.algorithms.configuration.keys())[0] in algorithms
+                or list(experiment.algorithm.configuration.keys())[0] in algorithms
             ):
                 exps.append((repetition_index, experiment))
         return exps

src/orion/benchmark/assessment/averagerank.py

@@ -34,7 +34,7 @@ def analysis(self, task, experiments):
         algorithm_groups = defaultdict(list)
 
         for _, exp in experiments:
-            algorithm_name = list(exp.configuration["algorithms"].keys())[0]
+            algorithm_name = list(exp.configuration["algorithm"].keys())[0]
             algorithm_groups[algorithm_name].append(exp)
 
         return {rankings.__name__: rankings(algorithm_groups)}

src/orion/benchmark/assessment/averageresult.py

@@ -32,7 +32,7 @@ def analysis(self, task, experiments):
         algorithm_groups = defaultdict(list)
 
         for _, exp in experiments:
-            algorithm_name = list(exp.configuration["algorithms"].keys())[0]
+            algorithm_name = list(exp.configuration["algorithm"].keys())[0]
             algorithm_groups[algorithm_name].append(exp)
 
         return {regrets.__name__: regrets(algorithm_groups)}

src/orion/benchmark/assessment/parallelassessment.py

@@ -62,7 +62,7 @@ def analysis(self, task, experiments):
         algorithm_groups = defaultdict(list)
         algorithm_worker_groups = defaultdict(list)
         for repetition_index, exp in experiments:
-            algorithm_name = list(exp.configuration["algorithms"].keys())[0]
+            algorithm_name = list(exp.configuration["algorithm"].keys())[0]
             algorithm_groups[algorithm_name].append(exp)
 
             n_worker = self.workers[repetition_index]

src/orion/client/__init__.py

@@ -55,6 +55,7 @@ def build_experiment(
     name: str,
     version: int | None = None,
     space: dict[str, Any] | None = None,
+    algorithm: type[BaseAlgorithm] | dict | None = None,
     algorithms: type[BaseAlgorithm] | dict | None = None,
     strategy: str | dict | None = None,
     max_trials: int | None = None,
@@ -81,7 +82,7 @@ def build_experiment(
     ``name`` and ``space`` arguments are required, otherwise ``NoConfigurationError`` will be
     raised.
 
-    All other arguments (``algorithms``, ``strategy``, ``max_trials``, ``storage``, ``branching``
+    All other arguments (``algorithm``, ``strategy``, ``max_trials``, ``storage``, ``branching``
     and ``working_dir``) will be replaced by system's defaults if omitted. The system's defaults can
     also be overridden in global configuration file as described for the database in
     :ref:`Database Configuration`. We do not recommend overriding the algorithm configuration using
@@ -133,7 +134,7 @@ def build_experiment(
         or 1 for new experiment.
     space: dict, optional
         Optimization space of the algorithm. Should have the form ``dict(name='<prior>(args)')``.
-    algorithms: str or dict, optional
+    algorithm: str or dict, optional
         Algorithm used for optimization.
     strategy: str or dict, optional
         Deprecated and will be remove in v0.4. It should now be set in algorithm configuration
@@ -220,6 +221,7 @@ def build_experiment(
             name,
             version=version,
             space=space,
+            algorithm=algorithm,
             algorithms=algorithms,
             max_trials=max_trials,
             max_broken=max_broken,
@@ -235,6 +237,7 @@ def build_experiment(
                 name,
                 version=version,
                 space=space,
+                algorithm=algorithm,
                 algorithms=algorithms,
                 strategy=strategy,
                 max_trials=max_trials,
@@ -293,6 +296,7 @@ def workon(
     function: Callable,
     space: dict,
     name: str = "loop",
+    algorithm: type[BaseAlgorithm] | str | dict | None = None,
     algorithms: type[BaseAlgorithm] | str | dict | None = None,
     max_trials: int | None = None,
     max_broken: int | None = None,
@@ -302,7 +306,7 @@ def workon(
 
     This will create a new experiment with an in-memory storage and optimize the given function
     until `max_trials` is reached or the `algorithm` is done
-    (some algorithms like random search are never done).
+    (some algorithm like random search are never done).
 
     For information on how to fetch results, see
     :py:class:`orion.client.experiment.ExperimentClient`.
@@ -320,7 +324,7 @@ def workon(
         or 1 for new experiment.
     space: dict, optional
         Optimization space of the algorithm. Should have the form `dict(name='<prior>(args)')`.
-    algorithms: str or dict, optional
+    algorithm: str or dict, optional
         Algorithm used for optimization.
     max_trials: int, optional
         Maximum number or trials before the experiment is considered done.
@@ -337,6 +341,7 @@ def workon(
         name,
         version=1,
         space=space,
+        algorithm=algorithm,
         algorithms=algorithms,
         max_trials=max_trials,
         max_broken=max_broken,

src/orion/client/experiment.py

@@ -169,9 +169,9 @@ def space(self) -> Space:
         return space
 
     @property
-    def algorithms(self):
-        """Algorithms of the experiment."""
-        return self._experiment.algorithms
+    def algorithm(self):
+        """Algorithm of the experiment."""
+        return self._experiment.algorithm
 
     @property
     def refers(self):
@@ -807,8 +807,8 @@ def workon(
 
         if self.max_trials > max_trials:
             self._experiment.max_trials = max_trials
-            assert self._experiment.algorithms is not None
-            self._experiment.algorithms.max_trials = max_trials
+            assert self._experiment.algorithm is not None
+            self._experiment.algorithm.max_trials = max_trials
 
         with SetupWorkingDir(self):
             runner = Runner(