diff --git a/modules/model_quant.py b/modules/model_quant.py
index 7e219cd8d..a71e7ee0a 100644
--- a/modules/model_quant.py
+++ b/modules/model_quant.py
@@ -40,6 +40,16 @@ def get_quant(name):
     return 'none'
 
 
+def dont_quant():
+    from modules import shared
+    models_list = re.split(r'[ ,]+', shared.opts.models_not_to_quant)
+    models_list = [m.lower().strip() for m in models_list]
+    if shared.sd_model_type.lower() in models_list:
+        shared.log.debug(f'Quantization: model={shared.sd_model_type} skip')
+        return True
+    return False
+
+
 def create_bnb_config(kwargs = None, allow: bool = True, module: str = 'Model', modules_to_not_convert: list = None):
     from modules import shared, devices
     if allow and (module == 'any' or module in shared.opts.bnb_quantization):
@@ -237,6 +247,8 @@ def check_nunchaku(module: str = ''):
 
 
 def create_config(kwargs = None, allow: bool = True, module: str = 'Model', modules_to_not_convert: list = None, modules_dtype_dict: dict = None):
+    if dont_quant():
+        return kwargs
     if kwargs is None:
         kwargs = {}
     kwargs = create_sdnq_config(kwargs, allow=allow, module=module, modules_to_not_convert=modules_to_not_convert, modules_dtype_dict=modules_dtype_dict)
@@ -757,16 +769,6 @@ def get_dit_args(load_config:dict=None, module:str=None, device_map:bool=False,
     return config, quant_args
 
 
-def dont_quant():
-    from modules import shared
-    models_list = re.split(r'[ ,]+', shared.opts.models_not_to_quant)
-    models_list = [m.lower().strip() for m in models_list]
-    if shared.sd_model_type.lower() in models_list:
-        shared.log.debug(f'Quantization: model={shared.sd_model_type} skip')
-        return True
-    return False
-
-
 def do_post_load_quant(sd_model, allow=True):
     from modules import shared
     if dont_quant():
diff --git a/modules/processing_diffusers.py b/modules/processing_diffusers.py
index c0acfb0d7..6bd9041dd 100644
--- a/modules/processing_diffusers.py
+++ b/modules/processing_diffusers.py
@@ -144,8 +144,7 @@ def process_base(p: processing.StableDiffusionProcessing):
         denoising_start=0 if use_refiner_start else p.refiner_start if use_denoise_start else None,
         denoising_end=p.refiner_start if use_refiner_start else 1 if use_denoise_start else None,
         num_frames=getattr(p, 'frames', 1),
-        # output_type='latent',
-        output_type='np',
+        output_type='latent',
         clip_skip=p.clip_skip,
         desc=desc,
     )
diff --git a/modules/shared.py b/modules/shared.py
index 01a4d59a6..a969c81b0 100644
--- a/modules/shared.py
+++ b/modules/shared.py
@@ -176,7 +176,7 @@ options_templates.update(options_section(('offload', "Model Offloading"), {
     "diffusers_offload_always": OptionInfo(startup_offload_always, "Modules to always offload"),
     "diffusers_offload_never": OptionInfo(startup_offload_never, "Modules to never offload"),
     "offload_group_sep": OptionInfo("<h2>Group Offload</h2>", "", gr.HTML),
-    "group_offload_type": OptionInfo("block_level", "Group offload type", gr.Radio, {"choices": ['block_level', 'leaf_level']}),
+    "group_offload_type": OptionInfo("leaf_level", "Group offload type", gr.Radio, {"choices": ['leaf_level', 'block_level']}),
     "group_offload_stream": OptionInfo(False, "Use torch streams", gr.Checkbox),
     'group_offload_record': OptionInfo(False, "Record torch streams", gr.Checkbox),
     'group_offload_blocks': OptionInfo(1, "Offload blocks", gr.Number),