fix(app): handle dtype alignment for style/subject reference pipelines

app.py

@@ -313,6 +313,10 @@ def infer_compo(style_description, ref_style_file, caption, ref_sub_file, use_lo
         batch['style'] = ref_style_for_clip
         batch['images'] = ref_images_for_clip
 
+        batch_b = {'captions': [caption] * batch_size}
+        batch_b['style'] = ref_style_for_effnet
+        batch_b['images'] = ref_images_for_effnet
+
         progress(0.2, "Processing reference images")
         x0_forward = models_rbm.effnet(
             extras.effnet_preprocess(ref_images_for_effnet)
@@ -340,11 +344,11 @@ def infer_compo(style_description, ref_style_file, caption, ref_sub_file, use_lo
             eval_image_embeds=False, eval_subject_style=True
         )
         conditions_b = core_b.get_conditions(
-            batch, models_b, extras_b,
+            batch_b, models_b, extras_b,
             is_eval=True, is_unconditional=False
         )
         unconditions_b = core_b.get_conditions(
-            batch, models_b, extras_b,
+            batch_b, models_b, extras_b,
             is_eval=True, is_unconditional=True
         )