confidence intervals corrected

deployment/nbeats/Dockerfile

@@ -2,11 +2,11 @@ FROM python:3.8-slim-buster
 
 # Install Python dependencies.
 WORKDIR /wd
-COPY deployment/tft/requirements.txt .
+COPY deployment/nbeats/requirements.txt .
 RUN pip3 install --no-cache-dir -r requirements.txt && mkdir models
 
 # Copy the rest of the codebase into the image
-COPY deployment/tft ./
+COPY deployment/nbeats ./
 COPY morphemic-datasetmaker ./morphemic-datasetmaker
 COPY amq-message-python-library ./amq-message-python-library
 

deployment/nbeats/main.py

@@ -76,6 +76,7 @@ class Msg(object):
 
 
 def main():
+    print("STARTTTTT")
     logging.getLogger().setLevel(logging.DEBUG)
 
     start_app_conn = morphemic.Connection(
@@ -95,17 +96,18 @@ def main():
         "2", StartForecastingListener(start_conn.conn, START_TOPIC)
     )
 
-    # msg1 = Msg()
-    # msg1.body = '[{"metric": "cpu_usage", "level": 3, "publish_rate": 60000}]'
-    # msg2 = Msg()
-    # msg2.body = (
-    #     "{"
-    #     + f'"metrics": ["cpu_usage"],"timestamp": {int(time.time())}, "epoch_start": {int(time.time()) + 30}, "number_of_forward_predictions": 8,"prediction_horizon": 60'
-    #     + "}"
-    # )
-
-    # StartListener(start_conn.conn, START_APP_TOPIC).on_message(msg1)
-    # StartForecastingListener(start_conn.conn, START_APP_TOPIC).on_message(msg2)
+    msg1 = Msg()
+    msg1.body = '[{"metric": "cpu_usage", "level": 3, "publish_rate": 60000}]'
+    msg2 = Msg()
+    msg2.body = (
+        "{"
+        + f'"metrics": ["cpu_usage"],"timestamp": {int(time.time())}, "epoch_start": {int(time.time()) + 30}, "number_of_forward_predictions": 8,"prediction_horizon": 60'
+        + "}"
+    )
+
+    print(msg1)
+    StartListener(start_conn.conn, START_APP_TOPIC).on_message(msg1)
+    StartForecastingListener(start_conn.conn, START_APP_TOPIC).on_message(msg2)
 
     while True:
         pass

deployment/nbeats/model.yaml

@@ -2,8 +2,8 @@ data:
   csv_path: demo.csv
 training:
     bs: 8
-    max_epochs: 1
-    loss: quantile
+    max_epochs: 5
+    loss: rmse
 dataset:
     tv_unknown_reals: []
     known_reals: []

deployment/nbeats/predict.py

@@ -66,12 +66,12 @@ class CustomListener(stomp.ConnectionListener):
 def main():
     logging.debug(f"metrics to predict {predicted_metrics}")
     start_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     start_conn.connect()
 
     stop_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     stop_conn.connect()
 

deployment/nbeats/retrain.py

@@ -10,7 +10,7 @@ from datetime import timedelta
 from src.dataset_maker import CSVData
 
 TOPIC_NAME = "training_models"
-RETRAIN_CYCLE = 2  # minutes
+RETRAIN_CYCLE = 10  # minutes
 HOSTS = (os.environ.get("AMQ_HOSTNAME", "localhost"), os.environ.get("AMQ_PORT", 61613))
 AMQ_USER = os.environ.get("AMQ_USER", "admin")
 AMQ_PASSWORD = os.environ.get("AMQ_PASSWORD", "admin")
@@ -28,10 +28,6 @@ def main(predicted_metrics, prediction_horizon):
     dataset_preprocessor = CSVData(APP_NAME)
     dataset_preprocessor.prepare_csv()
 
-    for metric in predicted_metrics:
-        retrain_msg = train(metric, prediction_horizon)
-        start_conn.send_to_topic(TOPIC_NAME, retrain_msg)
-
     while True:
         start_time = int(time.time())
         logging.debug("TRAINING")

deployment/nbeats/src/model_predict.py

@@ -78,13 +78,13 @@ def predict(
 
     msg = {
         target_column: {
-            "metricValue": prediction[-1][-1][2].item(),
+            "metricValue": prediction[-1].item(),
             "level": 1,  # TODO
             "timestamp": int(time.time()),
             "probability": 0.95,
             "confidence_interval": [
-                prediction[-1][-1][0].item(),
-                prediction[-1][-1][-1].item(),
+                prediction[-1].item(),
+                prediction[-1].item(),
             ],  # quantiles difference
             "predictionTime": timestamp + prediction_hor * m // 1000,
             "refersTo": "TODO",
@@ -95,5 +95,5 @@ def predict(
     logging.debug(f"prediction msg: {msg}")
 
     future_df["split"] = "val"
-    future_df[target_column] = prediction.permute(0, 2, 1)[0][2]
+    future_df[target_column] = prediction[-1].item()
     return (msg, future_df)

deployment/nbeats/src/model_train.py

@@ -33,7 +33,7 @@ def train(target_column, prediction_length, yaml_file="model.yaml"):
         os.environ.get("DATA_PATH", "./"), f'{os.environ.get("APP_NAME", "demo")}.csv'
     )
 
-    dataset = pd.read_csv(data_path)
+    dataset = pd.read_csv(data_path).head(1000)
 
     ts_dataset = Dataset(dataset, target_column=target_column, **params["dataset"])
 

deployment/nbeats/src/preprocess_dataset.py

@@ -72,23 +72,14 @@ class Dataset(object):
             self.dataset[lambda x: x.split == "train"],
             time_idx="time_idx",
             target=self.target_column,
+            categorical_encoders={"series": NaNLabelEncoder().fit(self.dataset.series)},
             group_ids=["series"],
-            min_encoder_length=self.context_length,  # keep encoder length long (as it is in the validation set)
+            time_varying_unknown_reals=[self.target_column],
+            min_encoder_length=self.context_length,
             max_encoder_length=self.context_length,
-            min_prediction_length=self.prediction_length,
             max_prediction_length=self.prediction_length,
-            static_categoricals=[],
-            static_reals=self.static_reals,
-            time_varying_known_categoricals=[],
-            categorical_encoders={"series": NaNLabelEncoder().fit(self.dataset.series)},
-            variable_groups={},  # group of categorical variables can be treated as one variable
-            time_varying_known_reals=["time_idx"] + self.known_reals,
-            time_varying_unknown_categoricals=self.tv_unknown_cat,
-            time_varying_unknown_reals=[self.target_column] + self.tv_unknown_reals,
-            add_relative_time_idx=True,
-            add_target_scales=True if not self.classification else False,
-            add_encoder_length=True,
-            allow_missings=True,
+            min_prediction_length=self.prediction_length,
+            add_relative_time_idx=False,
         )
         return ts_dataset
 

deployment/tft/main.py

@@ -79,11 +79,11 @@ def main():
     logging.getLogger().setLevel(logging.DEBUG)
 
     start_app_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     start_app_conn.connect()
     start_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     start_conn.connect()
 

deployment/tft/predict.py

@@ -66,12 +66,12 @@ class CustomListener(stomp.ConnectionListener):
 def main():
     logging.debug(f"metrics to predict {predicted_metrics}")
     start_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     start_conn.connect()
 
     stop_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     stop_conn.connect()
 

deployment/tft/retrain.py

@@ -21,7 +21,7 @@ APP_NAME = os.environ.get("APP_NAME", "demo")
 
 def main(predicted_metrics, prediction_horizon):
     start_conn = morphemic.Connection(
-        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER
+        AMQ_USER, AMQ_PASSWORD, host=AMQ_HOST, port=AMQ_PORT_BROKER, debug=True
     )
     start_conn.connect()
 

deployment/tft/src/model_predict.py

@@ -83,15 +83,23 @@ def predict(
     prediction_input = prediction_input.to_dataloader(train=False)
     prediction = tft.predict(prediction_input, mode="raw")["prediction"]
 
+    predicted_values = [
+        prediction[-1][-1][0].item(),
+        prediction[-1][-1][3].item(),
+        prediction[-1][-1][-1].item(),
+    ]
+
+    predicted_values.sort()
+
     msg = {
         target_column: {
-            "metricValue": prediction[-1][-1][2].item(),
+            "metricValue": predicted_values[1],
             "level": 1,  # TODO
             "timestamp": int(time.time()),
             "probability": 0.95,
             "confidence_interval": [
-                prediction[-1][-1][0].item(),
-                prediction[-1][-1][-1].item(),
+                predicted_values[0],
+                predicted_values[-1],
             ],  # quantiles difference
             "predictionTime": timestamp,
             "refersTo": "TODO",
@@ -102,5 +110,5 @@ def predict(
     logging.debug(f"prediction msg: {msg}")
 
     future_df["split"] = "val"
-    future_df[target_column] = prediction.permute(0, 2, 1)[0][2]
+    future_df[target_column] = prediction.permute(0, 2, 1)[0][3]
     return (msg, future_df)