docs: Update ControlNet use case docs (#4519)

* Update ControlNext use case docs

Signed-off-by: Sherlock113 <[email protected]>

* ci: auto fixes from pre-commit.ci

For more information, see https://pre-commit.ci

---------

Signed-off-by: Sherlock113 <[email protected]>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

docs/source/use-cases/diffusion-models/controlnet.rst

@@ -40,13 +40,10 @@ Create BentoML :doc:`/guides/services` in a ``service.py`` file to specify the s
 
     import typing as t
 
-    import cv2
     import numpy as np
     import PIL
     from PIL.Image import Image as PIL_Image
 
-    import torch
-    from diffusers import StableDiffusionXLControlNetPipeline, ControlNetModel, AutoencoderKL
     from pydantic import BaseModel
 
     import bentoml
@@ -55,21 +52,21 @@ Create BentoML :doc:`/guides/services` in a ``service.py`` file to specify the s
     VAE_MODEL_ID = "madebyollin/sdxl-vae-fp16-fix"
     BASE_MODEL_ID = "stabilityai/stable-diffusion-xl-base-1.0"
 
-
     @bentoml.service(
         traffic={"timeout": 600},
         workers=1,
         resources={
-            "gpu": "1",
+            "gpu": 1,
             "gpu_type": "nvidia-l4",
-            # we can also specify GPU memory requirement:
-            # "memory": "16Gi",
         }
     )
-    class SDXLControlNetService:
+    class ControlNet:
 
         def __init__(self) -> None:
 
+            import torch
+            from diffusers import StableDiffusionXLControlNetPipeline, ControlNetModel, AutoencoderKL
+
             if torch.cuda.is_available():
                 self.device = "cuda"
                 self.dtype = torch.float16
@@ -94,58 +91,34 @@ Create BentoML :doc:`/guides/services` in a ``service.py`` file to specify the s
                 torch_dtype=self.dtype
             ).to(self.device)
 
-
-        @bentoml.api
-        async def generate(
-                self,
-                prompt: str,
-                arr: np.ndarray[t.Any, np.uint8],
-                **kwargs,
-        ):
-            image = PIL.Image.fromarray(arr)
-            return self.pipe(prompt, image=image, **kwargs).to_tuple()
-
-
-    class Params(BaseModel):
-        prompt: str
-        negative_prompt: t.Optional[str]
-        controlnet_conditioning_scale: float = 0.5
-        num_inference_steps: int = 25
-
-
-    @bentoml.service(
-        name="sdxl-controlnet-service",
-        traffic={"timeout": 600},
-        workers=8,
-    resources={"cpu": "1"}
-    )
-    class ControlNet:
-        controlnet_service: SDXLControlNetService = bentoml.depends(SDXLControlNetService)
-
         @bentoml.api
         async def generate(self, image: PIL_Image, params: Params) -> PIL_Image:
+            import cv2
+
             arr = np.array(image)
             arr = cv2.Canny(arr, 100, 200)
             arr = arr[:, :, None]
             arr = np.concatenate([arr, arr, arr], axis=2)
             params_d = params.dict()
             prompt = params_d.pop("prompt")
-            res = await self.controlnet_service.generate(
+            image = PIL.Image.fromarray(arr)
+            return self.pipe(
                 prompt,
-                arr=arr,
+                image=image,
                 **params_d
-            )
-            return res[0][0]
-
-This file defines the following classes:
+            ).to_tuple()[0][0]
 
-* ``SDXLControlNetService``: A BentoML Service with custom configurations in timeout, worker count, and resources.
+    class Params(BaseModel):
+        prompt: str
+        negative_prompt: t.Optional[str]
+        controlnet_conditioning_scale: float = 0.5
+        num_inference_steps: int = 25
 
-  - It loads the three pre-trained models and configures them to use GPU if available. The main pipeline (``StableDiffusionXLControlNetPipeline``) integrates these models.
-  - It defines an API endpoint ``generate`` to process a text prompt and an image array. The processed image is converted to a tuple and returned.
+This file defines a BentoML Service ``ControlNet`` with custom :doc:`configurations </guides/configurations>` in timeout, worker count, and resources.
 
-* ``Params``: This is a ``pydantic`` model defining the structure for input parameters.
-* ``ControlNet``: A BentoML Service with custom configurations in timeout, worker count, and resources. ``ControlNet`` doesn't create images itself. Instead, it preprocesses the image and forwards it along with the text prompt to the ``SDXLControlNetService`` Service. The ``generate`` method in ``ControlNet`` then returns the final generated image.
+- It loads the three pre-trained models and configures them to use GPU if available. The main pipeline (``StableDiffusionXLControlNetPipeline``) integrates these models.
+- It defines an asynchronous API endpoint ``generate``, which takes an image and a set of parameters as input. The parameters for the generation process are extracted from a ``Params`` instance, a Pydantic model that provides automatic data validation.
+- The ``generate`` method returns the generated image by calling the pipeline with the processed image and text prompts.
 
 Run ``bentoml serve`` in your project directory to start the BentoML server.