uuu

app.py

@@ -21,6 +21,7 @@ import gradio as gr
 import torch
 from pipeline_stable_diffusion_3_ipa import StableDiffusion3Pipeline
 import os
+import numpy as np
 import spaces
 from huggingface_hub import login
 token = os.getenv("HF_TOKEN")
@@ -62,23 +63,33 @@ pipe.init_ipadapter(
     nb_token=64,
 )
 
-
 @spaces.GPU
 def gui_generation(prompt, ref_img):
     """
     Generate images using Stable Diffusion 3.5
     """
-    image = pipe(
-        width=1024,
-        height=1024,
-        prompt=prompt,
-        negative_prompt="lowres, low quality, worst quality",
-        num_inference_steps=24,
-        guidance_scale=5.0,
-        generator=torch.Generator("cuda").manual_seed(42),
-        clip_image=ref_img,
-        ipadapter_scale=0.5,
-    ).images[0]
+    # Ensure reference image is in the correct format
+    if ref_img:
+        ref_img = ref_img.convert("RGB")
+        ref_img_tensor = torch.tensor(
+            np.array(ref_img), dtype=torch.bfloat16, device="cuda"
+        )
+    else:
+        raise ValueError("Reference image is required.")
+
+    # Ensure the pipeline runs with correct dtype and device
+    with torch.autocast("cuda", dtype=torch.bfloat16):
+        image = pipe(
+            width=1024,
+            height=1024,
+            prompt=prompt,
+            negative_prompt="lowres, low quality, worst quality",
+            num_inference_steps=24,
+            guidance_scale=5.0,
+            generator=torch.Generator("cuda").manual_seed(42),
+            clip_image=ref_img_tensor,
+            ipadapter_scale=0.5,
+        ).images[0]
 
     return image