UI for whitebox server using streamlit (#131)

Author: stavrostheocharis

.streamlit/config.toml

@@ -0,0 +1,6 @@
+[theme]
+base="dark"
+primaryColor="#21babe"
+backgroundColor="#1e2025"
+secondaryBackgroundColor="#252a33"
+

docs/mkdocs/docs/sdk-docs.md

@@ -4,15 +4,15 @@ This is the documentation for Whitebox's SDK. For an interactive experience, you
 
 ## Models
 
-**_create_model_**_(name, type, prediction, labels=None, description="")_
+**_create_model_**_(name, type, target_column, labels=None, description="")_
 
 Creates a model in the database. This model works as placeholder for all the actual model's metadata.
 
 | Parameter       | Type             | Description                                                               |
 | --------------- | ---------------- | ------------------------------------------------------------------------- |
 | **name**        | `str`            | The name of the model.                                                    |
 | **type**        | `str`            | The model's type. Possible values: `binary`, `multi_class`, `regression`. |
-| **prediction**  | `str`            | The prediction of the model.                                              |
+| **target_column**  | `str`            | The name of the target column (y).                             |
 | **labels**      | `Dict[str, int]` | The model's labels. Defaults to `None`.                                   |
 | **description** | `str`            | The model's description. Defaults to an empty string `""`.                |
 

docs/mkdocs/docs/tutorial/sdk.md

@@ -73,7 +73,7 @@ wb.create_model(
         'additionalProp1': 0,
         'additionalProp2': 1
     },
-    prediction="target"
+    target_column="target"
 )
 ```
 

examples/notebooks/sdk-example.ipynb

@@ -74,7 +74,7 @@
     }
    ],
    "source": [
-    "wb.create_model(name=\"Model 1\", type=\"binary\", labels={'additionalProp1': 0, 'additionalProp2': 1}, prediction=\"y_prediction_multi\")"
+    "wb.create_model(name=\"Model 1\", type=\"binary\", labels={'additionalProp1': 0, 'additionalProp2': 1}, target_column=\"y_prediction_multi\")"
    ]
   },
   {
@@ -248,7 +248,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3",
+   "display_name": ".venv",
    "language": "python",
    "name": "python3"
   },
@@ -262,12 +262,12 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.8"
+   "version": "3.8.6"
   },
   "orig_nbformat": 4,
   "vscode": {
    "interpreter": {
-    "hash": "32a5a47fe20cdfbd609287887f2e78a4d5f2f7afeda3da775d5794970f9a3f8e"
+    "hash": "6b8a8ae524dcca06b04542d2d49be160be0f11dd19d43d7b2673d555344c6092"
    }
   }
  },

requirements.txt

@@ -88,16 +88,29 @@ setuptools==67.1.0
 six==1.16.0
 sniffio==1.3.0
 SQLAlchemy==1.4.46
+stack-data==0.5.1
 starlette==0.23.1
 statsmodels==0.13.5
+streamlit==1.18.1
 tenacity==8.2.1
 threadpoolctl==3.1.0
 tifffile==2023.2.3
 tomli==2.0.1
 tqdm==4.64.1
+traitlets==5.4.0
+typed-ast==1.5.4
+typer==0.6.1
+types-python-dateutil==2.8.19
 typing_extensions==4.4.0
+unicodedata2==14.0.0
 urllib3==1.26.14
 uvicorn==0.20.0
 virtualenv==20.19.0
 watchdog==2.2.1
+wcwidth==0.2.5
 wheel==0.38.4
+wrapt==1.14.1
+xgboost==1.6.2
+yarl==1.8.1
+zipp==3.10.0
+

whitebox/.streamlit/config.toml

@@ -0,0 +1,6 @@
+[theme]
+base="dark"
+primaryColor="#21babe"
+backgroundColor="#1e2025"
+secondaryBackgroundColor="#252a33"
+

whitebox/api/v1/dataset_rows.py

@@ -45,15 +45,15 @@ async def create_dataset_rows(
     model = crud.models.get(db=db, _id=dict(body[0])["model_id"])
     if model:
         for row in body:
-            if not model.prediction in row.processed:
+            if not model.target_column in row.processed:
                 return errors.bad_request(
-                    f'Column "{model.prediction}" was not found in some or any of the rows in provided training dataset. Please try again!'
+                    f'Column "{model.target_column}" was not found in some or any of the rows in provided training dataset. Please try again!'
                 )
 
-        predictions = list(set(vars(x)["processed"][model.prediction] for x in body))
+        predictions = list(set(vars(x)["processed"][model.target_column] for x in body))
         if len(predictions) <= 1:
             return errors.bad_request(
-                f'Training dataset\'s "{model.prediction}" columns must have at least 2 different values!'
+                f'Training dataset\'s "{model.target_column}" columns must have at least 2 different values!'
             )
 
         new_dataset_rows = crud.dataset_rows.create_many(db=db, obj_list=body)
@@ -66,21 +66,21 @@ async def create_dataset_rows(
             background_tasks.add_task(
                 create_binary_classification_training_model_pipeline,
                 processed_dataset_rows_pd,
-                model.prediction,
+                model.target_column,
                 model.id,
             )
         elif model.type == ModelType.multi_class:
             background_tasks.add_task(
                 create_multiclass_classification_training_model_pipeline,
                 processed_dataset_rows_pd,
-                model.prediction,
+                model.target_column,
                 model.id,
             )
         elif model.type == ModelType.regression:
             background_tasks.add_task(
                 create_regression_training_model_pipeline,
                 processed_dataset_rows_pd,
-                model.prediction,
+                model.target_column,
                 model.id,
             )
         return new_dataset_rows

whitebox/api/v1/inference_rows.py

@@ -50,8 +50,18 @@ async def create_many_inference_rows(
 ) -> List[InferenceRow]:
     """Inserts a set of inference rows into the database."""
 
-    new_inference_rows = crud.inference_rows.create_many(db=db, obj_list=body)
-    return new_inference_rows
+    model = crud.models.get(db=db, _id=dict(body[0])["model_id"])
+    if model:
+        for row in body:
+            if not model.target_column in row.processed:
+                return errors.bad_request(
+                    f'Column "{model.target_column}" was not found in some or any of the rows in provided inference dataset. Please try again!'
+                )
+
+        new_inference_rows = crud.inference_rows.create_many(db=db, obj_list=body)
+        return new_inference_rows
+    else:
+        return errors.not_found(f"Model with id: {dict(body[0])['model_id']} not found")
 
 
 @inference_rows_router.get(
@@ -138,7 +148,7 @@ async def create_inference_row_xai_report(
 
     xai_report = create_xai_pipeline_per_inference_row(
         training_set=pd.DataFrame(dataset_rows_processed),
-        target=model.prediction,
+        target=model.target_column,
         inference_row=inference_row_series,
         type_of_task=model.type,
         model_id=model.id,

whitebox/cron_tasks/monitoring_metrics.py

@@ -51,10 +51,10 @@ async def run_calculate_drifting_metrics_pipeline(
 
     # We need to drop the target column from the data to calculate drifting metrics
     processed_inference_dropped_target_df = inference_processed_df.drop(
-        [model.prediction], axis=1
+        [model.target_column], axis=1
     )
     processed_training_dropped_target_df = training_processed_df.drop(
-        [model.prediction], axis=1
+        [model.target_column], axis=1
     )
 
     data_drift_report = run_data_drift_pipeline(
@@ -63,7 +63,7 @@ async def run_calculate_drifting_metrics_pipeline(
     concept_drift_report = run_concept_drift_pipeline(
         training_processed_df,
         inference_processed_df,
-        model.prediction,
+        model.target_column,
     )
 
     new_drifting_metric = entities.DriftingMetric(
@@ -115,7 +115,7 @@ async def run_calculate_performance_metrics_pipeline(
     if model.type == ModelType.binary:
         binary_classification_metrics_report = (
             create_binary_classification_evaluation_metrics_pipeline(
-                cleaned_actuals_df, inference_processed_df[model.prediction], labels
+                cleaned_actuals_df, inference_processed_df[model.target_column], labels
             )
         )
 
@@ -130,7 +130,7 @@ async def run_calculate_performance_metrics_pipeline(
     elif model.type == ModelType.multi_class:
         multiclass_classification_metrics_report = (
             create_multiple_classification_evaluation_metrics_pipeline(
-                cleaned_actuals_df, inference_processed_df[model.prediction], labels
+                cleaned_actuals_df, inference_processed_df[model.target_column], labels
             )
         )
 
@@ -144,7 +144,7 @@ async def run_calculate_performance_metrics_pipeline(
 
     elif model.type == ModelType.regression:
         regression_metrics_report = create_regression_evaluation_metrics_pipeline(
-            cleaned_actuals_df, inference_processed_df[model.prediction]
+            cleaned_actuals_df, inference_processed_df[model.target_column]
         )
 
         new_performance_metric = entities.RegressionMetrics(

whitebox/entities/Model.py

@@ -13,7 +13,7 @@ class Model(Base):
     description = Column(String)
     type = Column("type", Enum(ModelType))
     labels = Column(JSON, nullable=True)
-    prediction = Column(String)
+    target_column = Column(String)
     created_at = Column(DateTime)
     updated_at = Column(DateTime)