AutoMLx Upgrade (#643)

ahosler · web-flow · commit d209cc13906f · 2024-03-18T11:51:40.000Z
diff --git a/ads/opctl/operator/lowcode/anomaly/README.md b/ads/opctl/operator/lowcode/anomaly/README.md
@@ -37,7 +37,8 @@ To run anomaly detection locally, create and activate a new conda environment (`
 ```yaml
 - datapane
 - cerberus
-- oracle-automlx==23.2.3
+- oracle-automlx==23.4.1
+- oracle-automlx[classic]==23.4.1
 - "git+https://github.com/oracle/accelerated-data-science.git@feature/anomaly#egg=oracle-ads"
 ```
 
diff --git a/ads/opctl/operator/lowcode/anomaly/environment.yaml b/ads/opctl/operator/lowcode/anomaly/environment.yaml
@@ -7,5 +7,6 @@ dependencies:
   - pip:
       - datapane
       - cerberus
-      - oracle-automlx==23.2.3
+      - oracle-automlx==23.4.1
+      - oracle-automlx[classic]==23.4.1
       - "oracle-ads[anomaly]"
diff --git a/ads/opctl/operator/lowcode/anomaly/model/automlx.py b/ads/opctl/operator/lowcode/anomaly/model/automlx.py
@@ -17,20 +17,26 @@ class AutoMLXOperatorModel(AnomalyOperatorBaseModel):
     """Class representing AutoMLX operator model."""
 
     @runtime_dependency(
-        module="automl",
+        module="automlx",
         err_msg=(
-            "Please run `pip3 install oracle-automlx==23.2.3` to "
-            "install the required dependencies for automlx."
+            "Please run `pip3 install oracle-automlx==23.4.1` and "
+            "`pip3 install oracle-automlx[classic]==23.4.1` "
+            "to install the required dependencies for automlx."
         ),
     )
     def _build_model(self) -> pd.DataFrame:
+        from automlx import init
+        try:
+            init(engine="ray", engine_opts={"ray_setup": {"_temp_dir": "/tmp/ray-temp"}})
+        except Exception as e:
+            logger.info("Ray already initialized")
         date_column = self.spec.datetime_column.name
         anomaly_output = AnomalyOutput(date_column=date_column)
+        time_budget = self.spec.model_kwargs.pop("time_budget", -1)
 
-        time_budget = self.spec.model_kwargs.pop("time_budget", None)
         # Iterate over the full_data_dict items
         for target, df in self.datasets.full_data_dict.items():
-            est = automl.Pipeline(task="anomaly_detection", **self.spec.model_kwargs)
+            est = automlx.Pipeline(task="anomaly_detection", **self.spec.model_kwargs)
             est.fit(
                 X=df,
                 X_valid=self.X_valid_dict[target]
@@ -39,10 +45,10 @@ def _build_model(self) -> pd.DataFrame:
                 y_valid=self.y_valid_dict[target]
                 if self.y_valid_dict is not None
                 else None,
-                time_budget=time_budget,
                 contamination=self.spec.contamination
                 if self.y_valid_dict is not None
                 else None,
+                time_budget=time_budget,
             )
             y_pred = est.predict(df)
             scores = est.predict_proba(df)
diff --git a/ads/opctl/operator/lowcode/forecast/README.md b/ads/opctl/operator/lowcode/forecast/README.md
@@ -38,8 +38,9 @@ To run forecasting locally, create and activate a new conda environment (`ads-fo
 - datapane
 - cerberus
 - sktime
-- optuna==2.9.0
-- oracle-automlx==23.2.3
+- optuna==3.1.0
+- oracle-automlx==23.4.1
+- oracle-automlx[forecasting]==23.4.1
 - oracle-ads>=2.9.0
 ```
 
diff --git a/ads/opctl/operator/lowcode/forecast/environment.yaml b/ads/opctl/operator/lowcode/forecast/environment.yaml
@@ -15,5 +15,6 @@ dependencies:
       - sktime
       - shap
       - autots[additional]
-      - optuna==2.9.0
-      - oracle-automlx==23.2.3
+      - optuna==3.1.0
+      - oracle-automlx==23.4.1
+      - oracle-automlx[forecasting]==23.4.1
diff --git a/ads/opctl/operator/lowcode/forecast/model/automlx.py b/ads/opctl/operator/lowcode/forecast/model/automlx.py
@@ -45,7 +45,7 @@ def set_kwargs(self):
             model_kwargs_cleaned.get("score_metric", AUTOMLX_DEFAULT_SCORE_METRIC),
         )
         model_kwargs_cleaned.pop("task", None)
-        time_budget = model_kwargs_cleaned.pop("time_budget", None)
+        time_budget = model_kwargs_cleaned.pop("time_budget", -1)
         model_kwargs_cleaned[
             "preprocessing"
         ] = self.spec.preprocessing or model_kwargs_cleaned.get("preprocessing", True)
@@ -55,9 +55,11 @@ def preprocess(self, data, series_id=None):
         return data.set_index(self.spec.datetime_column.name)
 
     @runtime_dependency(
-        module="automl",
+        module="automlx",
         err_msg=(
-            "Please run `pip3 install oracle-automlx==23.2.3` to install the required dependencies for automlx."
+            "Please run `pip3 install oracle-automlx==23.4.1` and "
+            "`pip3 install oracle-automlx[forecasting]==23.4.1` "
+            "to install the required dependencies for automlx."
         ),
     )
     @runtime_dependency(
@@ -67,15 +69,13 @@ def preprocess(self, data, series_id=None):
         ),
     )
     def _build_model(self) -> pd.DataFrame:
-        from automl import init
+        from automlx import init
         from sktime.forecasting.model_selection import temporal_train_test_split
+        try:
+            init(engine="ray", engine_opts={"ray_setup": {"_temp_dir": "/tmp/ray-temp"}})
+        except Exception as e:
+            logger.info("Ray already initialized")
 
-        init(
-            engine="local",
-            engine_opts={"n_jobs": -1, "model_n_jobs": -1},
-            check_deprecation_warnings=False,
-            logger=50,
-        )
 
         full_data_dict = self.datasets.get_data_by_series()
 
@@ -95,7 +95,7 @@ def _build_model(self) -> pd.DataFrame:
 
         for i, (s_id, df) in enumerate(full_data_dict.items()):
             try:
-                logger.debug(f"Running automl on series {s_id}")
+                logger.debug(f"Running automlx on series {s_id}")
                 model_kwargs = model_kwargs_cleaned.copy()
                 target = self.original_target_column
                 self.forecast_output.init_series_output(
@@ -110,7 +110,7 @@ def _build_model(self) -> pd.DataFrame:
                 if self.loaded_models is not None:
                     model = self.loaded_models[s_id]
                 else:
-                    model = automl.Pipeline(
+                    model = automlx.Pipeline(
                         task="forecasting",
                         **model_kwargs,
                     )
@@ -149,18 +149,7 @@ def _build_model(self) -> pd.DataFrame:
 
                 self.model_parameters[s_id] = {
                     "framework": SupportedModels.AutoMLX,
-                    "score_metric": model.score_metric,
-                    "random_state": model.random_state,
-                    "model_list": model.model_list,
-                    "n_algos_tuned": model.n_algos_tuned,
-                    "adaptive_sampling": model.adaptive_sampling,
-                    "min_features": model.min_features,
-                    "optimization": model.optimization,
-                    "preprocessing": model.preprocessing,
-                    "search_space": model.search_space,
                     "time_series_period": model.time_series_period,
-                    "min_class_instances": model.min_class_instances,
-                    "max_tuning_trials": model.max_tuning_trials,
                     "selected_model": model.selected_model_,
                     "selected_model_params": model.selected_model_params_,
                 }
diff --git a/docs/source/user_guide/operators/anomaly_detection_operator/faq.rst b/docs/source/user_guide/operators/anomaly_detection_operator/faq.rst
@@ -10,6 +10,6 @@ More details in the documentation here: https://docs.oracle.com/en-us/iaas/tools
 
 More details in the documentation here: https://winedarksea.github.io/AutoTS/build/html/source/tutorial.html
 
-**Pip Install Failing with "ERROR: No matching distribution found for oracle-automlx==23.2.3; extra == "forecast""**
+**Pip Install Failing with "ERROR: No matching distribution found for oracle-automlx==23.4.1; extra == "forecast""**
 
 Automlx only supports Python<=3.8, != 3.9, <= 3.10.7 . If you are builing in Python 3.9 or 3.10.8+, no automlx distribution will be available. It's recommended to use the conda pack available through Notebook Sessions, or to use Python 3.8
diff --git a/docs/source/user_guide/operators/forecasting_operator/faq.rst b/docs/source/user_guide/operators/forecasting_operator/faq.rst
@@ -10,6 +10,6 @@ More details in the documentation here: https://docs.oracle.com/en-us/iaas/tools
 
 More details in the documentation here: https://winedarksea.github.io/AutoTS/build/html/source/tutorial.html
 
-**Pip Install Failing with "ERROR: No matching distribution found for oracle-automlx==23.2.3; extra == "forecast""**
+**Pip Install Failing with "ERROR: No matching distribution found for oracle-automlx==23.4.1; extra == "forecast""**
 
 Automlx only supports Python<=3.8, != 3.9, <= 3.10.7 . If you are builing in Python 3.9 or 3.10.8+, no automlx distribution will be available. It's recommended to use the conda pack available through Notebook Sessions, or to use Python 3.8
diff --git a/pyproject.toml b/pyproject.toml
@@ -167,9 +167,10 @@ forecast = [
   "neuralprophet",
   "numpy",
   "oci-cli",
-  "optuna==2.9.0",
+  "optuna==3.1.0",
   "oracle-ads",
-  "oracle-automlx[forecasting]==23.2.3",
+  "oracle-automlx[forecasting]==23.4.1",
+  "oracle-automlx[classic]==23.4.1",
   "pmdarima",
   "prophet",
   "shap",
@@ -181,7 +182,8 @@ forecast = [
 anomaly  = [
   "oracle_ads[opctl]",
   "autots",
-  "oracle-automlx[forecasting]==23.2.3",
+  "oracle-automlx[forecasting]==23.4.1",
+  "oracle-automlx[classic]==23.4.1",
   "oracledb",
   "datapane",
 ]
diff --git a/tests/operators/anomaly/test_anomaly_simple.py b/tests/operators/anomaly/test_anomaly_simple.py
@@ -212,7 +212,7 @@ def test_load_datasets(model, data_dict):
 
         # run(yaml_i, backend="operator.local", debug=False)
 
-        with open(f"{tmpdirname}/anomaly.yaml", "w") as f:
+        with open(anomaly_yaml_filename, "w") as f:
             f.write(yaml.dump(yaml_i))
         sleep(0.5)
         subprocess.run(
diff --git a/tests/operators/forecast/test_datasets.py b/tests/operators/forecast/test_datasets.py
@@ -165,7 +165,7 @@ def test_load_datasets(model, dataset_name):
         if model == "autots":
             yaml_i["spec"]["model_kwargs"] = {"model_list": "superfast"}
         if model == "automlx":
-            yaml_i["spec"]["model_kwargs"] = {"time_budget": 1}
+            yaml_i["spec"]["model_kwargs"] = {"time_budget": 50}
 
         forecast_yaml_filename = f"{tmpdirname}/forecast.yaml"
         # with open(f"{tmpdirname}/forecast.yaml", "w") as f:
@@ -218,7 +218,7 @@ def run_operator(
         if model == "autots":
             yaml_i["spec"]["model_kwargs"] = {"model_list": "superfast"}
         if model == "automlx":
-            yaml_i["spec"]["model_kwargs"] = {"time_budget": 1}
+            yaml_i["spec"]["model_kwargs"] = {"time_budget": 50}
 
         forecast_yaml_filename = f"{tmpdirname}/forecast.yaml"
         with open(f"{tmpdirname}/forecast.yaml", "w") as f:
diff --git a/tests/operators/forecast/test_errors.py b/tests/operators/forecast/test_errors.py
@@ -210,7 +210,7 @@ def populate_yaml(
     if model == "autots":
         yaml_i["spec"]["model_kwargs"] = {"model_list": "superfast"}
     if model == "automlx":
-        yaml_i["spec"]["model_kwargs"] = {"time_budget": 1}
+        yaml_i["spec"]["model_kwargs"] = {"time_budget": 50}
 
     return yaml_i, output_data_path
 
diff --git a/tests/unitary/with_extras/langchain/test_serialization.py b/tests/unitary/with_extras/langchain/test_serialization.py
@@ -32,51 +32,14 @@ class ChainSerializationTest(TestCase):
     # We expect users to use the same LangChain version for serialize and de-serialize
 
     def setUp(self) -> None:
-        self.maxDiff = None
+        # self.maxDiff = None
         return super().setUp()
 
     PROMPT_TEMPLATE = "Tell me a joke about {subject}"
     COMPARTMENT_ID = "<ocid>"
     GEN_AI_KWARGS = {"service_endpoint": "https://endpoint.oraclecloud.com"}
     ENDPOINT = "https://modeldeployment.customer-oci.com/ocid/predict"
 
-    EXPECTED_LLM_CHAIN_WITH_OCI_MD = {
-        "lc": 1,
-        "type": "constructor",
-        "id": ["langchain", "chains", "llm", "LLMChain"],
-        "kwargs": {
-            "prompt": {
-                "lc": 1,
-                "type": "constructor",
-                "kwargs": {
-                    "input_variables": ["subject"],
-                    "template": "Tell me a joke about {subject}",
-                    "template_format": "f-string",
-                    "partial_variables": {},
-                },
-            },
-            "llm": {
-                "lc": 1,
-                "type": "constructor",
-                "id": ["ads", "llm", "ModelDeploymentVLLM"],
-                "kwargs": {
-                    "endpoint": "https://modeldeployment.customer-oci.com/ocid/predict",
-                    "model": "my_model",
-                },
-            },
-        },
-    }
-
-    EXPECTED_GEN_AI_LLM = {
-        "lc": 1,
-        "type": "constructor",
-        "id": ["ads", "llm", "GenerativeAI"],
-        "kwargs": {
-            "compartment_id": "<ocid>",
-            "client_kwargs": {"service_endpoint": "https://endpoint.oraclecloud.com"},
-        },
-    }
-
     EXPECTED_GEN_AI_EMBEDDINGS = {
         "lc": 1,
         "type": "constructor",
