Apply black formatting to neural-demo.py

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: mkeshavaNV

examples/neural_slang_demo/neural-demo.py

@@ -14,7 +14,12 @@
 from pathlib import Path
 
 # Create the app and load the slang module
-app = App(width=512 * 3 + 10 * 2, height=512, title="Neural Demo (FFLayer + DifferentialPtrPair)", device_type=spy.DeviceType.vulkan)
+app = App(
+    width=512 * 3 + 10 * 2,
+    height=512,
+    title="Neural Demo (FFLayer + DifferentialPtrPair)",
+    device_type=spy.DeviceType.vulkan,
+)
 module = spy.Module.load_from_file(app.device, "neural-demo.slang")
 
 # Load reference image
@@ -28,7 +33,7 @@ def create_buffer(data: np.ndarray):
         element_count=data.size,
         struct_size=4,
         usage=spy.BufferUsage.shader_resource | spy.BufferUsage.unordered_access,
-        data=data.astype("float32").flatten()
+        data=data.astype("float32").flatten(),
     )
 
 
@@ -38,7 +43,7 @@ def create_zero_buffer(size: int):
         element_count=size,
         struct_size=4,
         usage=spy.BufferUsage.shader_resource | spy.BufferUsage.unordered_access,
-        data=np.zeros(size, dtype="float32")
+        data=np.zeros(size, dtype="float32"),
     )
 
 
@@ -63,7 +68,14 @@ def __init__(self, width: int, height: int, num_latents: int):
 
     def optimize(self, learning_rate: float, iteration: int):
         # Use the actual Tensors for optimizer
-        module.optimizer_step(self._texture_tensor, self._texture_grads_tensor, self._m_texture, self._v_texture, learning_rate, iteration)
+        module.optimizer_step(
+            self._texture_tensor,
+            self._texture_grads_tensor,
+            self._m_texture,
+            self._v_texture,
+            learning_rate,
+            iteration,
+        )
 
 
 class Network(spy.InstanceList):
@@ -73,7 +85,7 @@ def __init__(self):
         super().__init__(module["Network"])
 
         # Layer sizes: 4 -> 32 -> 32 -> 3
-        layer0_params = 32 * 4 + 32   # weights + biases
+        layer0_params = 32 * 4 + 32  # weights + biases
         layer1_params = 32 * 32 + 32
         layer2_params = 3 * 32 + 3
         self.total_params = layer0_params + layer1_params + layer2_params
@@ -84,17 +96,17 @@ def __init__(self):
 
         # Layer 0: 4 -> 32
         scale = np.sqrt(6.0 / (4 + 32))
-        params_np[offset:offset + 32*4] = np.random.uniform(-scale, scale, 32*4)
-        offset += 32*4 + 32  # weights + biases
+        params_np[offset : offset + 32 * 4] = np.random.uniform(-scale, scale, 32 * 4)
+        offset += 32 * 4 + 32  # weights + biases
 
         # Layer 1: 32 -> 32
         scale = np.sqrt(6.0 / (32 + 32))
-        params_np[offset:offset + 32*32] = np.random.uniform(-scale, scale, 32*32)
-        offset += 32*32 + 32
+        params_np[offset : offset + 32 * 32] = np.random.uniform(-scale, scale, 32 * 32)
+        offset += 32 * 32 + 32
 
         # Layer 2: 32 -> 3
         scale = np.sqrt(6.0 / (32 + 3))
-        params_np[offset:offset + 3*32] = np.random.uniform(-scale, scale, 3*32)
+        params_np[offset : offset + 3 * 32] = np.random.uniform(-scale, scale, 3 * 32)
 
         # Create tensors for optimizer (underscore prefix avoids struct binding issues)
         self._params_tensor = spy.Tensor.from_numpy(app.device, params_np)
@@ -113,7 +125,14 @@ def __init__(self):
 
     def optimize(self, learning_rate: float, iteration: int):
         # Optimize MLP params using Tensors (for vectorization)
-        module.optimizer_step(self._params_tensor, self._params_grad_tensor, self._m_params, self._v_params, learning_rate, iteration)
+        module.optimizer_step(
+            self._params_tensor,
+            self._params_grad_tensor,
+            self._m_params,
+            self._v_params,
+            learning_rate,
+            iteration,
+        )
         # Optimize latent texture
         self.latent_texture.optimize(learning_rate, iteration)
 
@@ -139,12 +158,16 @@ def optimize(self, learning_rate: float, iteration: int):
     # Render using unified network
     lr_output = spy.Tensor.empty_like(image)
     module.render(pixel=spy.call_id(), resolution=res, network=network, _result=lr_output)
-    app.blit(lr_output, size=spy.int2(512), offset=spy.int2(offset, 0), tonemap=False, bilinear=True)
+    app.blit(
+        lr_output, size=spy.int2(512), offset=spy.int2(offset, 0), tonemap=False, bilinear=True
+    )
     offset += 522
 
     # Show loss
     loss_output = spy.Tensor.empty_like(image)
-    module.loss(pixel=spy.call_id(), resolution=res, network=network, reference=image, _result=loss_output)
+    module.loss(
+        pixel=spy.call_id(), resolution=res, network=network, reference=image, _result=loss_output
+    )
     app.blit(loss_output, size=spy.int2(512), offset=spy.int2(offset, 0), tonemap=False)
 
     learning_rate = 0.001