Update src/streamlit_app.py

src/streamlit_app.py

@@ -55,21 +55,18 @@ if train_btn:
             st.session_state.data = (X, y, y_pred, mse, model)
 
         else:
-            # Lower resolution grid → MUCH faster rendering
-            grid_n = int(num_points / 3)
-
-            x1 = np.linspace(0, 10, grid_n)
-            x2 = np.linspace(0, 10, grid_n)
+            x1 = np.linspace(0, 10, num_points)
+            x2 = np.linspace(0, 10, num_points)
             X1, X2 = np.meshgrid(x1, x2)
 
-            noise = np.random.randn(grid_n, grid_n) * noise_level
+            noise = np.random.randn(num_points, num_points) * noise_level
             Z = 3 * X1 + 2 * X2 + 10 + noise
 
             X_flat = np.column_stack((X1.ravel(), X2.ravel()))
             Z_flat = Z.ravel()
 
             model = LinearRegression().fit(X_flat, Z_flat)
-            Z_pred = model.predict(X_flat).reshape(grid_n, grid_n)
+            Z_pred = model.predict(X_flat).reshape(num_points, num_points)
             mse = mean_squared_error(Z_flat, Z_pred.ravel())
 
             st.session_state.data = (X1, X2, Z, Z_pred, mse, model)
@@ -84,14 +81,14 @@ if st.session_state.trained:
 
     st.success("🎉 Model trained successfully!")
 
-    # ----------------- 2D -----------------
+    # ----------------- 2D Regression -----------------
     if mode == "2D Regression":
         X, y, y_pred, mse, model = st.session_state.data
 
         col1, col2 = st.columns([2, 1])
 
         with col1:
-            fig, ax = plt.subplots(figsize=(4, 3.5))
+            fig, ax = plt.subplots(figsize=(4.5, 4))
             ax.scatter(X, y, color="orange", label="Data", s=18)
             ax.plot(X, y_pred, color="blue", linewidth=2, label="Regression Line")
             ax.set_title("2D Linear Regression")
@@ -102,7 +99,7 @@ if st.session_state.trained:
             st.metric("MSE", f"{mse:.4f}")
             st.code(f"y = {model.coef_[0]:.3f}x + {model.intercept_:.3f}")
 
-    # ----------------- 3D -----------------
+    # ----------------- 3D Regression -----------------
     else:
         X1, X2, Z, Z_pred, mse, model = st.session_state.data
 
@@ -112,36 +109,40 @@ if st.session_state.trained:
 
             # Static 3D plot
             if not rotate_3d:
-                fig = plt.figure(figsize=(4, 3.5))
+                fig = plt.figure(figsize=(4.5, 4))
                 ax = fig.add_subplot(111, projection="3d")
 
-                idx = np.random.choice(len(Z.ravel()), min(200, len(Z.ravel())), replace=False)
-                ax.scatter(X1.ravel()[idx], X2.ravel()[idx], Z.ravel()[idx],
-                           color="orange", alpha=0.22, s=6)
+                idx = np.random.choice(len(Z.ravel()), min(350, len(Z.ravel())), replace=False)
+                ax.scatter(
+                    X1.ravel()[idx], X2.ravel()[idx], Z.ravel()[idx],
+                    color="orange", alpha=0.25, s=8
+                )
 
-                ax.plot_surface(X1, X2, Z_pred, alpha=0.7, color="blue")
+                ax.plot_surface(X1, X2, Z_pred, alpha=0.75, color="blue")
                 ax.set_title("3D Linear Regression")
                 st.pyplot(fig, clear_figure=True)
 
-            # Smooth rotation animation (PC + HF optimized)
+            # Rotating 3D animation (HuggingFace-friendly)
             else:
                 placeholder = st.empty()
 
-                for angle in range(0, 360, 6):  # fewer steps = smoother on PC
-                    fig = plt.figure(figsize=(4, 3.5))
+                for angle in range(0, 360, 5):
+                    fig = plt.figure(figsize=(4.5, 4))
                     ax = fig.add_subplot(111, projection="3d")
 
-                    idx = np.random.choice(len(Z.ravel()), min(150, len(Z.ravel())), replace=False)
-                    ax.scatter(X1.ravel()[idx], X2.ravel()[idx], Z.ravel()[idx],
-                               alpha=0.2, color="orange", s=5)
-
-                    ax.plot_surface(X1, X2, Z_pred, alpha=0.70, color="blue")
+                    idx = np.random.choice(len(Z.ravel()), min(300, len(Z.ravel())), replace=False)
+                    ax.scatter(
+                        X1.ravel()[idx], X2.ravel()[idx], Z.ravel()[idx],
+                        alpha=0.2, color="orange", s=6
+                    )
 
+                    ax.plot_surface(X1, X2, Z_pred, alpha=0.75, color="blue")
                     ax.view_init(elev=25, azim=angle)
-                    ax.set_title("Rotating 3D Regression Model")
+
+                    ax.set_title("🔄 Rotating 3D Regression Model")
 
                     placeholder.pyplot(fig, clear_figure=True)
-                    time.sleep(0.05)  # PC-friendly speed
+                    time.sleep(0.07)
 
         with col2:
             st.metric("MSE", f"{mse:.4f}")