Update app.py

app.py

@@ -87,7 +87,13 @@ pipe.vae=vaeX.to(device)
 text_encoder=CLIPTextModelWithProjection.from_pretrained("ford442/stable-diffusion-3.5-large-bf16", subfolder='text_encoder', token=True).to(device=device, dtype=torch.bfloat16)
 text_encoder_2=CLIPTextModelWithProjection.from_pretrained("ford442/stable-diffusion-3.5-large-bf16", subfolder='text_encoder_2',token=True).to(device=device, dtype=torch.bfloat16)
 text_encoder_3=T5EncoderModel.from_pretrained("ford442/stable-diffusion-3.5-large-bf16", subfolder='text_encoder_3',token=True).to(device=device, dtype=torch.bfloat16)
-    
+       
+pipe.init_ipadapter(
+    ip_adapter_path=ipadapter_path, 
+    image_encoder_path=image_encoder_path, 
+    nb_token=64, 
+)
+
 upscaler_2 = UpscaleWithModel.from_pretrained("Kim2091/ClearRealityV1").to(torch.device("cuda:0"))
 
 MAX_SEED = np.iinfo(np.int32).max
@@ -151,11 +157,6 @@ def infer(
             sd_image_e.resize((height,width), Image.LANCZOS)
         else:
             sd_image_e = None
-        pipe.init_ipadapter(
-        ip_adapter_path=ipadapter_path, 
-        image_encoder_path=image_encoder_path, 
-        nb_token=64, 
-        )
         print('-- generating image --')
         sd_image = pipe(
             width=width,
@@ -180,38 +181,19 @@ def infer(
             scale_5=latent_file_5_scale,
         ).images[0]
         timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
-        rv_path = f"sd35_{timestamp}.png"
+        rv_path = f"sd35IP_{timestamp}.png"
         sd_image.save(rv_path,optimize=False,compress_level=0)
         upload_to_ftp(rv_path)
+        upscaler_2.to(torch.device('cuda'))
+        with torch.no_grad():
+            upscale2 = upscaler_2(sd_image, tiling=True, tile_width=256, tile_height=256)
+        print('-- got upscaled image --')
+        downscale2 = upscale2.resize((upscale2.width // 4, upscale2.height // 4),Image.LANCZOS)
+        upscale_path = f"sd35l_upscale_{seed}.png"
+        downscale2.save(upscale_path,optimize=False,compress_level=0)
+        upload_to_ftp(upscale_path)
     else:
-        print('-- generating image --')
-        sd_image = pipe(
-            prompt=prompt,
-            prompt_2=enhanced_prompt_2,
-            prompt_3=enhanced_prompt,
-            negative_prompt=negative_prompt_1,
-            negative_prompt_2=negative_prompt_2,
-            negative_prompt_3=negative_prompt_3,
-            guidance_scale=guidance_scale,
-            num_inference_steps=num_inference_steps,
-            width=width,
-            height=height,
-            generator=generator,
-            max_sequence_length=512
-        ).images[0]
-        print('-- got image --')
-        timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
-        sd35_path = f"sd35l_{timestamp}.png"
-        sd_image.save(sd35_path,optimize=False,compress_level=0)
-        upload_to_ftp(sd35_path)
-    upscaler_2.to(torch.device('cuda'))
-    with torch.no_grad():
-        upscale2 = upscaler_2(sd_image, tiling=True, tile_width=256, tile_height=256)
-    print('-- got upscaled image --')
-    downscale2 = upscale2.resize((upscale2.width // 4, upscale2.height // 4),Image.LANCZOS)
-    upscale_path = f"sd35l_upscale_{seed}.png"
-    downscale2.save(upscale_path,optimize=False,compress_level=0)
-    upload_to_ftp(upscale_path)
+        print('-- at least one input image required --')
     return sd_image, enhanced_prompt
 
 examples = [