chore: import upstream snapshot with attribution

doc/source/templates/03_serving_stable_diffusion/app.py

@@ -0,0 +1,62 @@
+from io import BytesIO
+
+from ray import serve
+from fastapi import FastAPI
+from fastapi.responses import Response
+import torch
+from diffusers import EulerDiscreteScheduler, StableDiffusionPipeline
+import logging
+
+app = FastAPI()
+logger = logging.getLogger("ray.serve")
+
+
+@serve.deployment(num_replicas=1)
+@serve.ingress(app)
+class APIIngress:
+    def __init__(self, diffusion_model_handle) -> None:
+        self.handle = diffusion_model_handle
+
+    @app.get(
+        "/imagine",
+        responses={200: {"content": {"image/png": {}}}},
+        response_class=Response,
+    )
+    async def generate(self, prompt: str, img_size: int = 512):
+        assert len(prompt), "prompt parameter cannot be empty"
+
+        image = await self.handle.generate.remote(prompt, img_size=img_size)
+        file_stream = BytesIO()
+        image.save(file_stream, "PNG")
+        return Response(content=file_stream.getvalue(), media_type="image/png")
+
+
+@serve.deployment(
+    ray_actor_options={"num_gpus": 1, "num_cpus": 1},
+    max_ongoing_requests=2,
+    autoscaling_config={
+        "min_replicas": 1,
+        "max_replicas": 3,
+        "target_ongoing_requests": 1,
+    },
+)
+class StableDiffusionV2:
+    def __init__(self):
+        model_id = "stabilityai/stable-diffusion-2"
+
+        scheduler = EulerDiscreteScheduler.from_pretrained(
+            model_id, subfolder="scheduler"
+        )
+        self.pipe = StableDiffusionPipeline.from_pretrained(
+            model_id, scheduler=scheduler, revision="fp16", torch_dtype=torch.float16
+        )
+        self.pipe = self.pipe.to("cuda")
+
+    def generate(self, prompt: str, img_size: int = 512):
+        assert len(prompt), "prompt parameter cannot be empty"
+        logger.info("Prompt: [%s]", prompt)
+        image = self.pipe(prompt, height=img_size, width=img_size).images[0]
+        return image
+
+
+entrypoint = APIIngress.bind(StableDiffusionV2.bind())