bug fix

app.py

@@ -56,75 +56,7 @@ def prepare_dataset(parameters, df, rain, temperature, datepicker):
 def predict(model, dataloader): 
     return model.predict(dataloader, mode="raw", return_x=True, return_index=True)
     
-## Initiate Data
-with open('data/parameters.pkl', 'rb') as f:
-    parameters = pickle.load(f)
-model = TemporalFusionTransformer.load_from_checkpoint('model/tft_check.ckpt', map_location=torch.device('cpu'))
-
-df = pd.read_pickle('data/test_data.pkl')
-df = df.loc[(df["Branch"] == 15) & (df["Group"].isin(["6","7","4","1"]))]
-
-rain_mapping = {
-    "Yes" : 1,
-    "No" : 0
-}
-
-# Start App
-st.title("Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting")
-
-st.markdown(body = """
-    ### Abstract
-    Multi-horizon forecasting often contains a complex mix of inputs – including
-    static (i.e. time-invariant) covariates, known future inputs, and other exogenous
-    time series that are only observed in the past – without any prior information
-    on how they interact with the target. Several deep learning methods have been
-    proposed, but they are typically ‘black-box’ models which do not shed light on
-    how they use the full range of inputs present in practical scenarios. In this pa-
-    per, we introduce the Temporal Fusion Transformer (TFT) – a novel attention-
-    based architecture which combines high-performance multi-horizon forecasting
-    with interpretable insights into temporal dynamics. To learn temporal rela-
-    tionships at different scales, TFT uses recurrent layers for local processing and
-    interpretable self-attention layers for long-term dependencies. TFT utilizes spe-
-    cialized components to select relevant features and a series of gating layers to
-    suppress unnecessary components, enabling high performance in a wide range of
-    scenarios. On a variety of real-world datasets, we demonstrate significant per-
-    formance improvements over existing benchmarks, and showcase three practical
-    interpretability use cases of TFT.
-
-    ### Experiments
-""")
-
-rain = st.radio("Rain Indicator", ('Default', 'Yes', 'No'))
-
-temperature = st.slider('Change in Temperature', min_value=-10.0, max_value=10.0, value=0.0, step=0.25)
-
-datepicker = st.date_input("Start of Forecast", datetime.date(2022, 12, 24), min_value=datetime.date(2022, 6, 26) + datetime.timedelta(days = 35), max_value=datetime.date(2023, 6, 26) - datetime.timedelta(days = 30))
-
-fig, axs = plt.subplots(2, 2, figsize=(8, 6))
-
-# Plot scatter plots for each group
-axs[0, 0].scatter(df.loc[df['Group'] == '4', 'Date'], df.loc[df['Group'] == '4', 'sales'], color='red', marker='o')
-axs[0, 0].set_title('Article Group 1')
-
-axs[0, 1].scatter(df.loc[df['Group'] == '7', 'Date'], df.loc[df['Group'] == '7', 'sales'], color='blue', marker='o')
-axs[0, 1].set_title('Article Group 2')
-
-axs[1, 0].scatter(df.loc[df['Group'] == '1', 'Date'], df.loc[df['Group'] == '1', 'sales'], color='green', marker='o')
-axs[1, 0].set_title('Article Group 3')
-
-axs[1, 1].scatter(df.loc[df['Group'] == '6', 'Date'], df.loc[df['Group'] == '6', 'sales'], color='yellow', marker='o')
-axs[1, 1].set_title('Article Group 4')
-
-# Adjust spacing between subplots
-plt.tight_layout()
-
-for ax in axs.flat:
-    ax.set_xlim(df['Date'].min(), df['Date'].max())
-    ax.set_ylim(df['sales'].min(), df['sales'].max())
-
-st.pyplot(fig)
-
-st.button("Forecast Sales", type="primary") #on_click=None,
+ #on_click=None,
 
 # %%
 #preds = raw_preds_to_df(out, quantiles = None)
@@ -133,4 +65,77 @@ st.button("Forecast Sales", type="primary") #on_click=None,
 #preds.rename(columns={'time_idx_x':'time_idx'},inplace=True)
 #preds.drop(columns=['time_idx_y'],inplace=True)
 
+def main():
+    ## Initiate Data
+    with open('data/parameters.pkl', 'rb') as f:
+        parameters = pickle.load(f)
+    model = TemporalFusionTransformer.load_from_checkpoint('model/tft_check.ckpt', map_location=torch.device('cpu'))
+    
+    df = pd.read_pickle('data/test_data.pkl')
+    df = df.loc[(df["Branch"] == 15) & (df["Group"].isin(["6","7","4","1"]))]
+    
+    rain_mapping = {
+        "Yes" : 1,
+        "No" : 0
+    }
+    
+    # Start App
+    st.title("Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting")
+    
+    st.markdown(body = """
+        ### Abstract
+        Multi-horizon forecasting often contains a complex mix of inputs – including
+        static (i.e. time-invariant) covariates, known future inputs, and other exogenous
+        time series that are only observed in the past – without any prior information
+        on how they interact with the target. Several deep learning methods have been
+        proposed, but they are typically ‘black-box’ models which do not shed light on
+        how they use the full range of inputs present in practical scenarios. In this pa-
+        per, we introduce the Temporal Fusion Transformer (TFT) – a novel attention-
+        based architecture which combines high-performance multi-horizon forecasting
+        with interpretable insights into temporal dynamics. To learn temporal rela-
+        tionships at different scales, TFT uses recurrent layers for local processing and
+        interpretable self-attention layers for long-term dependencies. TFT utilizes spe-
+        cialized components to select relevant features and a series of gating layers to
+        suppress unnecessary components, enabling high performance in a wide range of
+        scenarios. On a variety of real-world datasets, we demonstrate significant per-
+        formance improvements over existing benchmarks, and showcase three practical
+        interpretability use cases of TFT.
+    
+        ### Experiments
+    """)
+    st.write(df.head(5))
+    rain = st.radio("Rain Indicator", ('Default', 'Yes', 'No'))
+    
+    temperature = st.slider('Change in Temperature', min_value=-10.0, max_value=10.0, value=0.0, step=0.25)
+    
+    datepicker = st.date_input("Start of Forecast", datetime.date(2022, 12, 24), min_value=datetime.date(2022, 6, 26) + datetime.timedelta(days = 35), max_value=datetime.date(2023, 6, 26) - datetime.timedelta(days = 30))
+    
+    fig, axs = plt.subplots(2, 2, figsize=(8, 6))
+    
+    # Plot scatter plots for each group
+    axs[0, 0].scatter(df.loc[df['Group'] == '4', 'Date'], df.loc[df['Group'] == '4', 'sales'], color='red', marker='o')
+    axs[0, 0].set_title('Article Group 1')
+    
+    axs[0, 1].scatter(df.loc[df['Group'] == '7', 'Date'], df.loc[df['Group'] == '7', 'sales'], color='blue', marker='o')
+    axs[0, 1].set_title('Article Group 2')
+    
+    axs[1, 0].scatter(df.loc[df['Group'] == '1', 'Date'], df.loc[df['Group'] == '1', 'sales'], color='green', marker='o')
+    axs[1, 0].set_title('Article Group 3')
+    
+    axs[1, 1].scatter(df.loc[df['Group'] == '6', 'Date'], df.loc[df['Group'] == '6', 'sales'], color='yellow', marker='o')
+    axs[1, 1].set_title('Article Group 4')
+    
+    # Adjust spacing between subplots
+    plt.tight_layout()
+    
+    for ax in axs.flat:
+        ax.set_xlim(df['Date'].min(), df['Date'].max())
+        ax.set_ylim(df['sales'].min(), df['sales'].max())
+    
+    st.pyplot(fig)
+    
+    st.button("Forecast Sales", type="primary")
+
+if __name__ == '__main__':
+    main()