diff --git a/installer.py b/installer.py
index 369afcf47..4a5195a39 100644
--- a/installer.py
+++ b/installer.py
@@ -1427,7 +1427,7 @@ def run_deferred_tasks():
     log.debug('Starting deferred tasks')
     time.sleep(1.0) # wait for server to start
     try:
-        from modules.sd_models import write_metadata
+        from modules.sd_checkpoint import write_metadata
         write_metadata()
     except Exception as e:
         log.error(f'Deferred task error: write_metadata {e}')
@@ -1446,7 +1446,6 @@ def run_deferred_tasks():
     log.debug(f'Deferred tasks complete: time={round(time.time() - t_start, 2)}')
 
 
-
 @lru_cache
 def get_state():
     state = {
diff --git a/launch.py b/launch.py
index 10c5a1136..53335d3d4 100755
--- a/launch.py
+++ b/launch.py
@@ -212,6 +212,7 @@ def start_server(immediate=True, server=None):
     log.debug(f'Starting module: {server}')
     module_spec.loader.exec_module(server)
     threading.Thread(target=installer.run_deferred_tasks, daemon=True).start()
+
     uvicorn = None
     if args.test:
         log.info("Test only")
diff --git a/modules/control/run.py b/modules/control/run.py
index 0a78f24b7..21a2f9039 100644
--- a/modules/control/run.py
+++ b/modules/control/run.py
@@ -2,6 +2,8 @@ import os
 import sys
 import cv2
 from PIL import Image
+from modules.logger import log
+from modules import devices, shared, errors, processing, sd_models, sd_vae, scripts_manager, masking
 from modules.control import util # helper functions
 from modules.control import unit # control units
 from modules.control import processors # image preprocessors
@@ -12,8 +14,6 @@ from modules.control.units import lite # Kohya ControlLLLite
 from modules.control.units import t2iadapter # TencentARC T2I-Adapter
 from modules.control.units import reference # ControlNet-Reference
 from modules.control.processor import preprocess_image
-from modules import devices, shared, errors, processing, sd_models, sd_vae, scripts_manager, masking
-from modules.logger import log
 from modules.processing_class import StableDiffusionProcessingControl
 from modules.ui_common import infotext_to_html
 from modules.api import script
@@ -449,8 +449,8 @@ def control_run(state: str = '', # pylint: disable=keyword-arg-before-vararg
         setattr(p, k, v)
     p_extra_args = {}
 
-    if shared.sd_model is None:
-        log.warning('Aborted: op=control model not loaded')
+    if shared.sd_model is None: # triggers load if not loaded
+        log.warning('Aborted: op=generate model not loaded')
         return [], '', '', 'Error: model not loaded'
 
     unit_type = unit_type.strip().lower() if unit_type is not None else ''
diff --git a/modules/face/faceid.py b/modules/face/faceid.py
index 21c8cec95..f3cd6f783 100644
--- a/modules/face/faceid.py
+++ b/modules/face/faceid.py
@@ -235,7 +235,7 @@ def face_id(
                 faceid_model_name = None
             devices.torch_gc()
 
-        ipadapter.unapply(p.sd_model)
+        ipadapter.unapply(shared.sd_model)
         extra_networks.deactivate(p, p.network_data)
 
         p.extra_generation_params["IP Adapter"] = f"{basename}:{scale}"
@@ -243,7 +243,7 @@ def face_id(
         if faceid_model is not None and original_load_ip_adapter is not None:
             faceid_model.__class__.load_ip_adapter = original_load_ip_adapter
         if shared.opts.cuda_compile_backend == 'none':
-            token_merge.remove_token_merging(p.sd_model)
+            token_merge.remove_token_merging(shared.sd_model)
         script_callbacks.after_process_callback(p)
 
     return processed_images
diff --git a/modules/lora/extra_networks_lora.py b/modules/lora/extra_networks_lora.py
index 05d0dac86..9ca2b3b9f 100644
--- a/modules/lora/extra_networks_lora.py
+++ b/modules/lora/extra_networks_lora.py
@@ -178,6 +178,8 @@ class ExtraNetworkLora(extra_networks.ExtraNetwork):
             debug_log(f'Network check: type=LoRA requested={requested} status=forced')
             return True
         sd_model = shared.sd_model.pipe if hasattr(shared.sd_model, 'pipe') else shared.sd_model
+        if sd_model is None:
+            return False
         if not hasattr(sd_model, 'loaded_loras'):
             sd_model.loaded_loras = {}
         if include is None or len(include) == 0:
diff --git a/modules/modeldata.py b/modules/modeldata.py
index a19b172a0..3c22c32b5 100644
--- a/modules/modeldata.py
+++ b/modules/modeldata.py
@@ -167,47 +167,46 @@ class ModelData:
             self.sd_refiner = v
 
 
+model_data = ModelData()
+
+
 # provides shared.sd_model field as a property
 class Shared(sys.modules[__name__].__class__):
     @property
     def sd_loaded(self):
-        import modules.sd_models # pylint: disable=W0621
-        return modules.sd_models.model_data.sd_model is not None
+        return model_data.sd_model is not None
 
     @property
     def sd_model(self):
-        import modules.sd_models # pylint: disable=W0621
-        if modules.sd_models.model_data.sd_model is None:
+        if model_data.sd_model is None:
             fn = f'{os.path.basename(sys._getframe(2).f_code.co_filename)}:{sys._getframe(2).f_code.co_name}:{sys._getframe(1).f_code.co_name}' # pylint: disable=protected-access
-            log.debug(f'Model requested: fn={fn}') # pylint: disable=protected-access
-        return modules.sd_models.model_data.get_sd_model()
+            log.debug(f'Model requested: fn={fn}')
+        model = model_data.get_sd_model()
+        return model
 
     @sd_model.setter
     def sd_model(self, value):
-        import modules.sd_models # pylint: disable=W0621
         if value is None:
             fn = f'{os.path.basename(sys._getframe(2).f_code.co_filename)}:{sys._getframe(2).f_code.co_name}:{sys._getframe(1).f_code.co_name}' # pylint: disable=protected-access
-            log.debug(f'Model unloaded: fn={fn}') # pylint: disable=protected-access
-        modules.sd_models.model_data.set_sd_model(value)
+            if model_data.sd_model is not None:
+                log.debug(f'Model unloaded: fn={fn}')
+        model_data.set_sd_model(value)
 
     @property
     def sd_refiner(self):
-        import modules.sd_models # pylint: disable=W0621
-        return modules.sd_models.model_data.get_sd_refiner()
+        return model_data.get_sd_refiner()
 
     @sd_refiner.setter
     def sd_refiner(self, value):
-        import modules.sd_models # pylint: disable=W0621
-        modules.sd_models.model_data.set_sd_refiner(value)
+        model_data.set_sd_refiner(value)
 
     @property
     def sd_model_type(self):
         try:
-            import modules.sd_models # pylint: disable=W0621
-            if modules.sd_models.model_data.sd_model is None:
+            if model_data.sd_model is None:
                 model_type = 'none'
                 return model_type
-            model_type = get_model_type(modules.sd_models.model_data.sd_model)
+            model_type = get_model_type(model_data.sd_model)
         except Exception:
             model_type = 'unknown'
         return model_type
@@ -215,11 +214,10 @@ class Shared(sys.modules[__name__].__class__):
     @property
     def sd_refiner_type(self):
         try:
-            import modules.sd_models # pylint: disable=W0621
-            if modules.sd_models.model_data.sd_refiner is None:
+            if model_data.sd_refiner is None:
                 model_type = 'none'
                 return model_type
-            model_type = get_model_type(modules.sd_models.model_data.sd_refiner)
+            model_type = get_model_type(model_data.sd_refiner)
         except Exception:
             model_type = 'unknown'
         return model_type
@@ -231,6 +229,3 @@ class Shared(sys.modules[__name__].__class__):
             return get_console()
         except ImportError:
             return None
-
-
-model_data = ModelData()
diff --git a/modules/postprocess/yolo.py b/modules/postprocess/yolo.py
index 3bda389b6..37ff1fbbb 100644
--- a/modules/postprocess/yolo.py
+++ b/modules/postprocess/yolo.py
@@ -76,8 +76,10 @@ class YoloRestorer(Detailer):
         return list(self.list)
 
     def dependencies(self):
-        import installer
-        installer.install('ultralytics==8.3.40', ignore=True, quiet=True)
+        from installer import install
+        install('ultralytics==8.3.40', ignore=True, quiet=True)
+        install('omegaconf')
+        install('antlr4-python3-runtime')
 
     def predict(
             self,
diff --git a/modules/processing.py b/modules/processing.py
index 9b2b2b6e6..a6fa07ec9 100644
--- a/modules/processing.py
+++ b/modules/processing.py
@@ -14,7 +14,6 @@ from modules.processing_class import ( # pylint: disable=unused-import
     StableDiffusionProcessingControl,
 )
 from modules.processing_info import create_infotext
-from modules.modeldata import model_data
 
 
 opt_C = 4
@@ -39,7 +38,7 @@ processed = None # last known processed results
 
 class Processed:
     def __init__(self, p: StableDiffusionProcessing, images_list, seed=-1, info=None, subseed=None, all_prompts=None, all_negative_prompts=None, all_seeds=None, all_subseeds=None, index_of_first_image=0, infotexts=None, comments="", binary=None, audio=None):
-        self.sd_model_hash = getattr(shared.sd_model, 'sd_model_hash', '') if model_data.sd_model is not None else ''
+        self.sd_model_hash = getattr(shared.sd_model, 'sd_model_hash', '') if shared.sd_loaded is not None else ''
 
         self.prompt = p.prompt or ''
         self.negative_prompt = p.negative_prompt or ''
@@ -139,8 +138,11 @@ def get_processed(*args, **kwargs):
 def process_images(p: StableDiffusionProcessing) -> Processed:
     timer.process.reset()
     debug(f'Process images: class={p.__class__.__name__} {vars(p)}')
-    if not hasattr(p.sd_model, 'sd_checkpoint_info'):
-        log.error('Processing: incomplete model')
+    if shared.sd_model is None:
+        log.warning('Aborted: op=process model not loaded')
+        return None
+    if not hasattr(shared.sd_model, 'sd_checkpoint_info'):
+        log.error('Aborted: op=process incomplete model')
         return None
     if p.abort:
         log.debug('Processing: aborted')
diff --git a/modules/processing_diffusers.py b/modules/processing_diffusers.py
index 72dd85c72..4ad11c1d9 100644
--- a/modules/processing_diffusers.py
+++ b/modules/processing_diffusers.py
@@ -70,8 +70,10 @@ def restore_state(p: processing.StableDiffusionProcessing):
 
 def process_pre(p: processing.StableDiffusionProcessing):
     from modules import ipadapter, sd_hijack_freeu, para_attention, teacache, hidiffusion, ras, pag, cfgzero, transformer_cache, token_merge, linfusion, cachedit
+    if shared.sd_model is None:
+        log.warning('Processing modifiers: model not loaded')
+        return
     log.info('Processing modifiers: apply')
-
     try:
         # apply-with-unapply
         sd_models_compile.check_deepcache(enable=True)
@@ -505,19 +507,25 @@ def process_decode(p: processing.StableDiffusionProcessing, output):
 
 
 def update_pipeline(sd_model, p: processing.StableDiffusionProcessing):
+    if sd_model is None:
+        sd_model = shared.sd_model
+    if sd_model is None:
+        shared.log.warning('Processing: op=update model not loaded')
+        return None
+    updated_model = sd_model
     if sd_models.get_diffusers_task(sd_model) == sd_models.DiffusersTaskType.INPAINTING and getattr(p, 'image_mask', None) is None and p.task_args.get('image_mask', None) is None and getattr(p, 'mask', None) is None:
         log.warning('Processing: mode=inpaint mask=None')
-        sd_model = sd_models.set_diffuser_pipe(sd_model, sd_models.DiffusersTaskType.IMAGE_2_IMAGE)
+        updated_model = sd_models.set_diffuser_pipe(sd_model, sd_models.DiffusersTaskType.IMAGE_2_IMAGE)
     if shared.opts.cuda_compile_backend == "olive-ai":
-        sd_model = olive_check_parameters_changed(p, is_refiner_enabled(p))
+        updated_model = olive_check_parameters_changed(p, is_refiner_enabled(p))
     if sd_model.__class__.__name__ == "OnnxRawPipeline":
-        sd_model = preprocess_onnx_pipeline(p)
+        updated_model = preprocess_onnx_pipeline(p)
         global orig_pipeline # pylint: disable=global-statement
-        orig_pipeline = sd_model # processed ONNX pipeline should not be replaced with original pipeline.
-    if getattr(sd_model, "current_attn_name", None) != shared.opts.cross_attention_optimization:
+        orig_pipeline = updated_model # processed ONNX pipeline should not be replaced with original pipeline.
+    if getattr(updated_model, "current_attn_name", None) != shared.opts.cross_attention_optimization:
         log.info(f"Setting attention optimization: {shared.opts.cross_attention_optimization}")
-        attention.set_diffusers_attention(sd_model)
-    return sd_model
+        attention.set_diffusers_attention(updated_model)
+    return updated_model
 
 
 def validate_pipeline(p: processing.StableDiffusionProcessing):
diff --git a/modules/prompt_parser_diffusers.py b/modules/prompt_parser_diffusers.py
index 46b195baa..553999d76 100644
--- a/modules/prompt_parser_diffusers.py
+++ b/modules/prompt_parser_diffusers.py
@@ -79,7 +79,7 @@ class PromptEmbedder:
         earlyout = self.checkcache(p)
         if earlyout:
             return
-        self.pipe = prepare_model(p.sd_model)
+        self.pipe = prepare_model(shared.sd_model)
         if self.pipe is None:
             log.error("Prompt encode: cannot find text encoder in model")
             return
diff --git a/modules/sd_hijack_freeu.py b/modules/sd_hijack_freeu.py
index 2ffa14c95..bde6242b0 100644
--- a/modules/sd_hijack_freeu.py
+++ b/modules/sd_hijack_freeu.py
@@ -146,15 +146,15 @@ def ratio_to_region(width: float, offset: float, n: int):
 
 def apply_freeu(p):
     global state_enabled # pylint: disable=global-statement
-    if hasattr(p.sd_model, 'enable_freeu'):
+    if hasattr(shared.sd_model, 'enable_freeu'):
         if shared.opts.freeu_enabled:
             freeu_device = get_fft_device()
             if freeu_device != devices.cpu:
                 p.extra_generation_params['FreeU'] = f'b1={shared.opts.freeu_b1} b2={shared.opts.freeu_b2} s1={shared.opts.freeu_s1} s2={shared.opts.freeu_s2}'
-                p.sd_model.enable_freeu(s1=shared.opts.freeu_s1, s2=shared.opts.freeu_s2, b1=shared.opts.freeu_b1, b2=shared.opts.freeu_b2)
+                shared.sd_model.enable_freeu(s1=shared.opts.freeu_s1, s2=shared.opts.freeu_s2, b1=shared.opts.freeu_b1, b2=shared.opts.freeu_b2)
                 state_enabled = True
         elif state_enabled:
-            p.sd_model.disable_freeu()
+            shared.sd_model.disable_freeu()
             state_enabled = False
     if shared.opts.freeu_enabled and state_enabled:
         log.info(f'Applying Free-U: b1={shared.opts.freeu_b1} b2={shared.opts.freeu_b2} s1={shared.opts.freeu_s1} s2={shared.opts.freeu_s2}')
diff --git a/modules/sd_hijack_hypertile.py b/modules/sd_hijack_hypertile.py
index 9ea96a193..6bf01d5ac 100644
--- a/modules/sd_hijack_hypertile.py
+++ b/modules/sd_hijack_hypertile.py
@@ -178,7 +178,7 @@ def split_attention(layer: nn.Module, tile_size: int=256, min_tile_size: int=128
 
 def context_hypertile_vae(p):
     from modules import shared
-    if p.sd_model is None or not shared.opts.hypertile_vae_enabled:
+    if shared.sd_model is None or not shared.opts.hypertile_vae_enabled:
         return nullcontext()
     if shared.opts.cross_attention_optimization == 'Sub-quadratic':
         log.warning('Hypertile UNet is not compatible with Sub-quadratic cross-attention optimization')
@@ -188,7 +188,7 @@ def context_hypertile_vae(p):
     error_reported = False
     set_resolution(p)
     max_h, max_w = 0, 0
-    vae = getattr(p.sd_model, "vae", None)
+    vae = getattr(shared.sd_model, "vae", None)
     if height == 0 or width == 0:
         log.warning('Hypertile VAE disabled: resolution unknown')
         return nullcontext()
@@ -207,7 +207,7 @@ def context_hypertile_vae(p):
 
 def context_hypertile_unet(p):
     from modules import shared
-    if p.sd_model is None or not shared.opts.hypertile_unet_enabled:
+    if shared.sd_model is None or not shared.opts.hypertile_unet_enabled:
         return nullcontext()
     if shared.opts.cross_attention_optimization == 'Sub-quadratic' and not shared.cmd_opts.experimental:
         log.warning('Hypertile UNet is not compatible with Sub-quadratic cross-attention optimization')
@@ -216,7 +216,7 @@ def context_hypertile_unet(p):
     error_reported = False
     set_resolution(p)
     max_h, max_w = 0, 0
-    unet = getattr(p.sd_model, "unet", None)
+    unet = getattr(shared.sd_model, "unet", None)
     if height == 0 or width == 0:
         log.warning('Hypertile VAE disabled: resolution unknown')
         return nullcontext()
diff --git a/modules/sd_offload.py b/modules/sd_offload.py
index 40c933f6c..70d45a115 100644
--- a/modules/sd_offload.py
+++ b/modules/sd_offload.py
@@ -478,9 +478,10 @@ def apply_balanced_offload(sd_model=None, exclude:list[str]=None, force:bool=Fal
         return sd_model
     if exclude is None:
         exclude = []
-    t0 = time.time()
     if sd_model.__class__.__name__ in balanced_offload_exclude:
         return sd_model
+
+    t0 = time.time()
     cached = True
     checkpoint_name = sd_model.sd_checkpoint_info.name if getattr(sd_model, "sd_checkpoint_info", None) is not None else sd_model.__class__.__name__
     if force or (offload_hook_instance is None) or (offload_hook_instance.min_watermark != shared.opts.diffusers_offload_min_gpu_memory) or (offload_hook_instance.max_watermark != shared.opts.diffusers_offload_max_gpu_memory) or (checkpoint_name != offload_hook_instance.checkpoint_name):
diff --git a/modules/ui_extra_networks_textual_inversion.py b/modules/ui_extra_networks_textual_inversion.py
index 91e719828..5a24e0baa 100644
--- a/modules/ui_extra_networks_textual_inversion.py
+++ b/modules/ui_extra_networks_textual_inversion.py
@@ -1,6 +1,6 @@
 import json
 import os
-from modules import shared, sd_models, ui_extra_networks, files_cache, modelstats
+from modules import shared, ui_extra_networks, files_cache, modelstats
 from modules.logger import log
 from modules.textual_inversion import Embedding
 
@@ -12,10 +12,10 @@ class ExtraNetworksPageTextualInversion(ui_extra_networks.ExtraNetworksPage):
         self.embeddings = []
 
     def refresh(self):
-        if sd_models.model_data.sd_model is None:
+        if not shared.sd_loaded:
             return
-        if hasattr(sd_models.model_data.sd_model, 'embedding_db'):
-            sd_models.model_data.sd_model.embedding_db.load_textual_inversion_embeddings(force_reload=True)
+        if hasattr(shared.sd_model, 'embedding_db'):
+            shared.sd_model.embedding_db.load_textual_inversion_embeddings(force_reload=True)
 
     def create_item(self, embedding: Embedding):
         record = None
@@ -43,15 +43,15 @@ class ExtraNetworksPageTextualInversion(ui_extra_networks.ExtraNetworksPage):
         return record
 
     def list_items(self):
-        if sd_models.model_data.sd_model is None:
+        if not shared.sd_loaded:
             candidates = list(files_cache.list_files(shared.opts.embeddings_dir, ext_filter=['.pt', '.safetensors'], recursive=files_cache.not_hidden))
             self.embeddings = [
                 Embedding(vec=0, name=os.path.basename(embedding_path), filename=embedding_path)
                 for embedding_path
                 in candidates
             ]
-        elif hasattr(sd_models.model_data.sd_model, 'embedding_db'):
-            self.embeddings = list(sd_models.model_data.sd_model.embedding_db.word_embeddings.values())
+        elif hasattr(shared.sd_model, 'embedding_db'):
+            self.embeddings = list(shared.sd_model.embedding_db.word_embeddings.values())
         else:
             self.embeddings = []
         self.embeddings = sorted(self.embeddings, key=lambda emb: emb.filename)