adds local entrypoint for playground, matches infra with deployed ver…

…sion, pins deps (#623)

* adds local entrypoint for playground, matches infra with deployed version

* pins dependencies

* pins python version

06_gpu_and_ml/stable_diffusion/playground.py

@@ -1,17 +1,24 @@
+# ---
+# output-directory: "/tmp/playground-2-5"
+# args: ["--prompt", "A cinematic shot of a baby raccoon wearing an intricate Italian priest robe."]
+# ---
+
 from pathlib import Path
 
 import modal
 
 stub = modal.Stub("playground-2-5")
 
+DIFFUSERS_GIT_SHA = "2e31a759b5bd8ca2b288b5c61709636a96c4bae9"
+
 image = (
-    modal.Image.debian_slim()
+    modal.Image.debian_slim(python_version="3.10")
     .apt_install("git")
     .pip_install(
-        "git+https://github.com/huggingface/diffusers.git",
-        "transformers",
-        "accelerate",
-        "safetensors",
+        f"git+https://github.com/huggingface/diffusers.git@{DIFFUSERS_GIT_SHA}",
+        "transformers~=4.38.1",
+        "accelerate==0.27.2",
+        "safetensors==0.4.2",
     )
 )
 
@@ -24,7 +31,7 @@
     from fastapi import Response
 
 
-@stub.cls(image=image, gpu="a100")
+@stub.cls(image=image, gpu="H100")
 class Model:
     @modal.build()
     @modal.enter()
@@ -39,19 +46,33 @@ def load_weights(self):
         # from diffusers import EDMDPMSolverMultistepScheduler
         # pipe.scheduler = EDMDPMSolverMultistepScheduler()
 
-    @modal.web_endpoint()
-    def inference(
-        self,
-        prompt="A cinematic shot of a baby racoon wearing an intricate italian priest robe.",
-    ):
+    def _inference(self, prompt, n_steps=24, high_noise_frac=0.8):
         image = self.pipe(
-            prompt, num_inference_steps=50, guidance_scale=3
+            prompt,
+            negative_prompt="disfigured, ugly, deformed",
+            num_inference_steps=50,
+            guidance_scale=3,
         ).images[0]
 
         buffer = io.BytesIO()
         image.save(buffer, format="JPEG")
 
-        return Response(buffer.getvalue(), media_type="image/jpeg")
+        return buffer
+
+    @modal.method()
+    def inference(self, prompt, n_steps=24, high_noise_frac=0.8):
+        return self._inference(
+            prompt, n_steps=n_steps, high_noise_frac=high_noise_frac
+        ).getvalue()
+
+    @modal.web_endpoint()
+    def web_inference(self, prompt, n_steps=24, high_noise_frac=0.8):
+        return Response(
+            content=self._inference(
+                prompt, n_steps=n_steps, high_noise_frac=high_noise_frac
+            ).getvalue(),
+            media_type="image/jpeg",
+        )
 
 
 frontend_path = Path(__file__).parent / "frontend"
@@ -79,9 +100,9 @@ def app():
 
     with open("/assets/index.html", "w") as f:
         html = template.render(
-            inference_url=Model.inference.web_url,
+            inference_url=Model.web_inference.web_url,
             model_name="Playground 2.5",
-            default_prompt="Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
+            default_prompt="Astronaut in the ocean, cold color palette, muted colors, detailed, 8k",
         )
         f.write(html)
 
@@ -90,3 +111,17 @@ def app():
     )
 
     return web_app
+
+
+@stub.local_entrypoint()
+def main(prompt: str):
+    image_bytes = Model().inference.remote(prompt)
+
+    dir = Path("/tmp/playground-2-5")
+    if not dir.exists():
+        dir.mkdir(exist_ok=True, parents=True)
+
+    output_path = dir / "output.png"
+    print(f"Saving it to {output_path}")
+    with open(output_path, "wb") as f:
+        f.write(image_bytes)