diff --git a/api/onnx_web/chain/base.py b/api/onnx_web/chain/base.py
index 944e6116..fadce75d 100644
--- a/api/onnx_web/chain/base.py
+++ b/api/onnx_web/chain/base.py
@@ -250,12 +250,12 @@ class ChainPipeline:
                         stage_sources = stage_outputs
                         break
                     except Exception:
+                        worker.retries = worker.retries - 1
                         logger.exception(
                             "error while running stage pipeline, %s retries left", worker.retries
                         )
                         server.cache.clear()
                         run_gc([worker.get_device()])
-                        worker.retries = worker.retries - 1
 
                 if worker.retries <= 0:
                     raise RetryException("exhausted retries on stage")
diff --git a/api/onnx_web/diffusers/load.py b/api/onnx_web/diffusers/load.py
index 642ccf77..622ae341 100644
--- a/api/onnx_web/diffusers/load.py
+++ b/api/onnx_web/diffusers/load.py
@@ -262,7 +262,11 @@ def load_pipeline(
 
     for vae in VAE_COMPONENTS:
         if hasattr(pipe, vae):
-            getattr(pipe, vae).set_tiled(tiled=params.tiled_vae)
+            vae_model = getattr(pipe, vae)
+            vae_model.set_tiled(tiled=params.tiled_vae)
+            vae_model.set_window_size(
+                params.vae_tile // 8, params.vae_overlap
+            )
 
     # update panorama params
     if params.is_panorama():
@@ -276,12 +280,6 @@ def load_pipeline(
         )
         pipe.set_window_size(params.unet_tile // 8, unet_stride)
 
-        for vae in VAE_COMPONENTS:
-            if hasattr(pipe, vae):
-                getattr(pipe, vae).set_window_size(
-                    params.vae_tile // 8, params.vae_overlap
-                )
-
     run_gc([device])
 
     return pipe
diff --git a/api/onnx_web/diffusers/pipelines/panorama_xl.py b/api/onnx_web/diffusers/pipelines/panorama_xl.py
index e55a5870..7c87ed1a 100644
--- a/api/onnx_web/diffusers/pipelines/panorama_xl.py
+++ b/api/onnx_web/diffusers/pipelines/panorama_xl.py
@@ -445,8 +445,8 @@ class StableDiffusionXLPanoramaPipelineMixin(StableDiffusionXLImg2ImgPipelineMix
                 value[:, :, h_start:h_end, w_start:w_end] += latents_view_denoised
                 count[:, :, h_start:h_end, w_start:w_end] += 1
 
-            for i in range(len(regions)):
-                top, left, bottom, right, mult, prompt = regions[i]
+            for r in range(len(regions)):
+                top, left, bottom, right, mult, prompt = regions[r]
                 logger.debug("running region prompt: %s, %s, %s, %s, %s, %s", top, left, bottom, right, mult, prompt)
 
                 # convert coordinates to latent space
@@ -475,8 +475,8 @@ class StableDiffusionXLPanoramaPipelineMixin(StableDiffusionXLImg2ImgPipelineMix
                 noise_pred = self.unet(
                     sample=latent_region_input,
                     timestep=timestep,
-                    encoder_hidden_states=region_embeds[i],
-                    text_embeds=add_region_embeds[i],
+                    encoder_hidden_states=region_embeds[r],
+                    text_embeds=add_region_embeds[r],
                     time_ids=add_time_ids,
                 )
                 noise_pred = noise_pred[0]
@@ -504,8 +504,12 @@ class StableDiffusionXLPanoramaPipelineMixin(StableDiffusionXLImg2ImgPipelineMix
                 )
                 latents_region_denoised = scheduler_output.prev_sample.numpy()
 
-                value[:, :, h_start:h_end, w_start:w_end] += latents_region_denoised * mult
-                count[:, :, h_start:h_end, w_start:w_end] += mult
+                if mult > 1000.0:
+                    value[:, :, h_start:h_end, w_start:w_end] = latents_region_denoised * mult
+                    count[:, :, h_start:h_end, w_start:w_end] = mult
+                else:
+                    value[:, :, h_start:h_end, w_start:w_end] += latents_region_denoised * mult
+                    count[:, :, h_start:h_end, w_start:w_end] += mult
 
             # take the MultiDiffusion step. Eq. 5 in MultiDiffusion paper: https://arxiv.org/abs/2302.08113
             latents = np.where(count > 0, value / count, value)