Update app.py

app.py

@@ -51,6 +51,7 @@ def load_history(history_path):
 def smooth_data(data, window_size):
     return np.convolve(data, np.ones(window_size)/window_size, mode='valid')
 
+# Streamlit app
 # Streamlit app
 st.markdown('<h1 class="big-font">TuNNe</h1>', unsafe_allow_html=True)
 st.markdown('<h2 class="center-text">Tuning a Neural Network</h2>', unsafe_allow_html=True)
@@ -82,65 +83,80 @@ learning_rates = sorted(set(lr for lr, _, _ in hyperparameters))
 
 # Select slider for learning rate
 st.markdown('<p class="slider-label">Learning Rate</p>', unsafe_allow_html=True)
-selected_lr = st.select_slider("LR", options=learning_rates)
+if len(learning_rates) > 1:
+    selected_lr = st.select_slider("Learning Rate", options=learning_rates, label_visibility="collapsed")
+else:
+    selected_lr = learning_rates[0] if learning_rates else None
+    st.write(f"Only one learning rate available: {selected_lr}")
 
 # Filter batch sizes based on selected learning rate
 filtered_bs = sorted(set(bs for lr, bs, _ in hyperparameters if lr == selected_lr))
 st.markdown('<p class="slider-label">Batch Size</p>', unsafe_allow_html=True)
-selected_bs = st.select_slider("BS", options=filtered_bs)
+if len(filtered_bs) > 1:
+    selected_bs = st.select_slider("Batch Size", options=filtered_bs, label_visibility="collapsed")
+else:
+    selected_bs = filtered_bs[0] if filtered_bs else None
+    st.write(f"Only one batch size available: {selected_bs}")
 
 # Filter epochs based on selected learning rate and batch size
 filtered_epochs = sorted(set(epochs for lr, bs, epochs in hyperparameters if lr == selected_lr and bs == selected_bs))
 st.markdown('<p class="slider-label">Epochs</p>', unsafe_allow_html=True)
-selected_epochs = st.select_slider("Epochs", options=filtered_epochs)
+if len(filtered_epochs) > 1:
+    selected_epochs = st.select_slider("Epochs", options=filtered_epochs, label_visibility="collapsed")
+else:
+    selected_epochs = filtered_epochs[0] if filtered_epochs else None
+    st.write(f"Only one epoch option available: {selected_epochs}")
 
 # Options for grid and smoothing
 enable_grid = st.checkbox("Enable Grid Lines")
-if selected_epochs > 20:
+if selected_epochs and selected_epochs > 20:
     smoothing_window = st.slider("Smoothing Window (every 4 epochs)", min_value=1, max_value=5, step=1, value=1)
 
 # Find the corresponding history file
-history_filename = f"mnist_model_lr{selected_lr}_bs{selected_bs}_epochs{selected_epochs}.json"
-history_path = os.path.join(model_dir, history_filename)
-
-if os.path.exists(history_path):
-    history = load_history(history_path)
-
-    # Plot training & validation accuracy values
-    fig, ax = plt.subplots()
-    accuracy = history['accuracy']
-    val_accuracy = history['val_accuracy']
-    if selected_epochs > 20 and smoothing_window > 1:
-        accuracy = smooth_data(accuracy, smoothing_window * 4)
-        val_accuracy = smooth_data(val_accuracy, smoothing_window * 4)
-    sns.lineplot(x=range(len(accuracy)), y=accuracy, ax=ax, label='Train Accuracy')
-    sns.lineplot(x=range(len(val_accuracy)), y=val_accuracy, ax=ax, label='Validation Accuracy')
-    ax.set_title('Model Accuracy', fontsize=15)
-    ax.set_ylabel('Accuracy', fontsize=12)
-    ax.set_xlabel('Epoch', fontsize=12)
-    ax.legend(loc='upper left', fontsize=10)
-    if enable_grid:
-        ax.grid(True)
-    st.pyplot(fig)
-
-    # Plot training & validation loss values
-    fig, ax = plt.subplots()
-    loss = history['loss']
-    val_loss = history['val_loss']
-    if selected_epochs > 20 and smoothing_window > 1:
-        loss = smooth_data(loss, smoothing_window * 4)
-        val_loss = smooth_data(val_loss, smoothing_window * 4)
-    sns.lineplot(x=range(len(loss)), y=loss, ax=ax, label='Train Loss')
-    sns.lineplot(x=range(len(val_loss)), y=val_loss, ax=ax, label='Validation Loss')
-    ax.set_title('Model Loss', fontsize=15)
-    ax.set_ylabel('Loss', fontsize=12)
-    ax.set_xlabel('Epoch', fontsize=12)
-    ax.legend(loc='upper left', fontsize=10)
-    if enable_grid:
-        ax.grid(True)
-    st.pyplot(fig)
+if selected_lr is not None and selected_bs is not None and selected_epochs is not None:
+    history_filename = f"mnist_model_lr{selected_lr}_bs{selected_bs}_epochs{selected_epochs}.json"
+    history_path = os.path.join(model_dir, history_filename)
+
+    if os.path.exists(history_path):
+        history = load_history(history_path)
+
+        # Plot training & validation accuracy values
+        fig, ax = plt.subplots()
+        accuracy = history['accuracy']
+        val_accuracy = history['val_accuracy']
+        if selected_epochs > 20 and 'smoothing_window' in locals() and smoothing_window > 1:
+            accuracy = smooth_data(accuracy, smoothing_window * 4)
+            val_accuracy = smooth_data(val_accuracy, smoothing_window * 4)
+        sns.lineplot(x=range(len(accuracy)), y=accuracy, ax=ax, label='Train Accuracy')
+        sns.lineplot(x=range(len(val_accuracy)), y=val_accuracy, ax=ax, label='Validation Accuracy')
+        ax.set_title('Model Accuracy', fontsize=15)
+        ax.set_ylabel('Accuracy', fontsize=12)
+        ax.set_xlabel('Epoch', fontsize=12)
+        ax.legend(loc='upper left', fontsize=10)
+        if enable_grid:
+            ax.grid(True)
+        st.pyplot(fig)
+
+        # Plot training & validation loss values
+        fig, ax = plt.subplots()
+        loss = history['loss']
+        val_loss = history['val_loss']
+        if selected_epochs > 20 and 'smoothing_window' in locals() and smoothing_window > 1:
+            loss = smooth_data(loss, smoothing_window * 4)
+            val_loss = smooth_data(val_loss, smoothing_window * 4)
+        sns.lineplot(x=range(len(loss)), y=loss, ax=ax, label='Train Loss')
+        sns.lineplot(x=range(len(val_loss)), y=val_loss, ax=ax, label='Validation Loss')
+        ax.set_title('Model Loss', fontsize=15)
+        ax.set_ylabel('Loss', fontsize=12)
+        ax.set_xlabel('Epoch', fontsize=12)
+        ax.legend(loc='upper left', fontsize=10)
+        if enable_grid:
+            ax.grid(True)
+        st.pyplot(fig)
+    else:
+        st.error(f"History file not found: {history_path}")
 else:
-    st.error(f"History file not found: {history_path}")
+    st.error("Unable to load model due to missing hyperparameters")
 
 
 # Final message