Upload 2 files

app.py

@@ -44,31 +44,31 @@ class UNetNoCondWrapper(nn.Module):
     
 # device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
-device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
-
-model_id = "CarolineM5/InstructPix2Pix_WithoutPrompt_4_faces"
-vae = AutoencoderKL.from_pretrained(model_id, subfolder="vae").to(device)
-scheduler = DDPMScheduler.from_pretrained(model_id, subfolder="scheduler")
-tokenizer = CLIPTokenizer.from_pretrained(model_id, subfolder="tokenizer")
-text_encoder = CLIPTextModel.from_pretrained(model_id, subfolder="text_encoder").to(device)
-feature_extractor = CLIPImageProcessor.from_pretrained(model_id, subfolder="feature_extractor")
-
-# 2) Chargez votre UNet non‑conditionné et wrappez‑le
-base_unet = UNet2DModel.from_pretrained(model_id, subfolder="unet").to(device)
-wrapped_unet = UNetNoCondWrapper(base_unet).to(device)
-
-# 3) Construisez la pipeline manuellement
-pipe = StableDiffusionInstructPix2PixPipeline(
-    vae=vae,
-    text_encoder=text_encoder,
-    tokenizer=tokenizer,
-    unet=wrapped_unet,
-    scheduler=scheduler,
-    safety_checker=None,
-    feature_extractor=feature_extractor,
-)
-
-pipe = pipe.to(torch.float32).to(device)
+# device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+# model_id = "CarolineM5/InstructPix2Pix_WithoutPrompt_4_faces"
+# vae = AutoencoderKL.from_pretrained(model_id, subfolder="vae").to(device)
+# scheduler = DDPMScheduler.from_pretrained(model_id, subfolder="scheduler")
+# tokenizer = CLIPTokenizer.from_pretrained(model_id, subfolder="tokenizer")
+# text_encoder = CLIPTextModel.from_pretrained(model_id, subfolder="text_encoder").to(device)
+# feature_extractor = CLIPImageProcessor.from_pretrained(model_id, subfolder="feature_extractor")
+
+# # 2) Chargez votre UNet non‑conditionné et wrappez‑le
+# base_unet = UNet2DModel.from_pretrained(model_id, subfolder="unet").to(device)
+# wrapped_unet = UNetNoCondWrapper(base_unet).to(device)
+
+# # 3) Construisez la pipeline manuellement
+# pipe = StableDiffusionInstructPix2PixPipeline(
+#     vae=vae,
+#     text_encoder=text_encoder,
+#     tokenizer=tokenizer,
+#     unet=wrapped_unet,
+#     scheduler=scheduler,
+#     safety_checker=None,
+#     feature_extractor=feature_extractor,
+# )
+
+# pipe = pipe.to(torch.float32).to(device)
 # @spaces.GPU
 
 def pil_to_data_uri(img: Image.Image) -> str:
@@ -122,7 +122,8 @@ def pil_to_data_uri(img: Image.Image) -> str:
 def run(fibers: Image.Image, rings: Image.Image, num_steps: int):
     try:
         # appelle la fonction d'inference de l'utilisateur
-        outputs = inference(pipe, fibers, rings, int(num_steps))
+        # outputs = inference(pipe, fibers, rings, int(num_steps))
+        outputs = inference(fibers, rings, int(num_steps))
         if not (isinstance(outputs, (list,tuple)) and len(outputs) >= 4):
             raise ValueError("La fonction d'inference doit renvoyer une liste/tuple de 4 images.")
         # Prendre les 4 premières images

inference.py

@@ -20,8 +20,8 @@ def pil_from(x):
     if isinstance(x, str):
         return PIL.Image.open(x)
     return x
-
-def inference(pipe, fiber_imgs, ring_imgs, num_steps):
+def inference(fiber_imgs, ring_imgs, num_steps):
+# def inference(pipe, fiber_imgs, ring_imgs, num_steps):
     """
     fiber_imgs:  PIL.Image or paths
     ring_imgs:  PIL.Image or paths