diff --git a/launch.py b/launch.py
index e00da58c7..5c8a6051a 100755
--- a/launch.py
+++ b/launch.py
@@ -192,6 +192,9 @@ def main():
     global args # pylint: disable=global-statement
     installer.ensure_base_requirements()
     init_args() # setup argparser and default folders
+    if args.malloc:
+        import tracemalloc
+        tracemalloc.start()
     installer.args = args
     installer.setup_logging()
     installer.log.info('Starting SD.Next')
diff --git a/modules/cmd_args.py b/modules/cmd_args.py
index 752ad02c0..cb4e5fc16 100644
--- a/modules/cmd_args.py
+++ b/modules/cmd_args.py
@@ -26,6 +26,7 @@ def main_args():
     group_diag.add_argument("--no-hashing", default=os.environ.get("SD_NOHASHING", False), action='store_true', help="Disable hashing of checkpoints, default: %(default)s")
     group_diag.add_argument("--no-metadata", default=os.environ.get("SD_NOMETADATA", False), action='store_true', help="Disable reading of metadata from models, default: %(default)s")
     group_diag.add_argument("--profile", default=os.environ.get("SD_PROFILE", False), action='store_true', help="Run profiler, default: %(default)s")
+    group_diag.add_argument("--malloc", default=os.environ.get("SD_PROFILE", False), action='store_true', help="Trace memory ops, default: %(default)s")
     group_diag.add_argument("--disable-queue", default=os.environ.get("SD_DISABLEQUEUE", False), action='store_true', help="Disable queues, default: %(default)s")
     group_diag.add_argument('--debug', default=os.environ.get("SD_DEBUG", False), action='store_true', help = "Run installer with debug logging, default: %(default)s")
 
diff --git a/modules/lora/networks.py b/modules/lora/networks.py
index f211149bd..23c45ff2a 100644
--- a/modules/lora/networks.py
+++ b/modules/lora/networks.py
@@ -22,7 +22,7 @@ from modules import shared, devices, sd_models, sd_models_compile, errors, files
 
 
 debug = os.environ.get('SD_LORA_DEBUG', None) is not None
-pbar = p.Progress(p.TextColumn('[cyan]LoRA apply'), p.BarColumn(), p.TaskProgressColumn(), p.TimeRemainingColumn(), p.TimeElapsedColumn(), p.TextColumn('[cyan]{task.description}'), console=shared.console)
+pbar = p.Progress(p.TextColumn('[cyan]{task.description}'), p.BarColumn(), p.TaskProgressColumn(), p.TimeRemainingColumn(), p.TimeElapsedColumn(), console=shared.console)
 extra_network_lora = None
 available_networks = {}
 available_network_aliases = {}
@@ -50,6 +50,13 @@ def total_time():
     return sum(timer.values())
 
 
+def get_timers():
+    t = { 'total': round(sum(timer.values()), 2) }
+    for k, v in timer.items():
+        t[k] = round(v, 2)
+    return t
+
+
 def assign_network_names_to_compvis_modules(sd_model):
     if sd_model is None:
         return
@@ -362,7 +369,8 @@ def network_apply_weights(self: Union[torch.nn.Conv2d, torch.nn.Linear, torch.nn
     network_layer_name = getattr(self, 'network_layer_name', None)
     current_names = getattr(self, "network_current_names", ())
     wanted_names = tuple((x.name, x.te_multiplier, x.unet_multiplier, x.dyn_dim) for x in loaded_networks)
-    maybe_backup_weights(self, wanted_names)
+    if network_layer_name is not None and any([net.modules.get(network_layer_name, None) for net in loaded_networks]): # noqa: C419
+        maybe_backup_weights(self, wanted_names)
     if current_names != wanted_names:
         batch_updown = None
         batch_ex_bias = None
@@ -414,30 +422,23 @@ def network_load(): # called from processing
     if shared.opts.diffusers_offload_mode != "none":
         sd_models.disable_offload(sd_model)
         sd_models.move_model(sd_model, device=devices.cpu)
+    modules = []
+    for component_name in ['text_encoder','text_encoder_2', 'unet', 'transformer']:
+        component = getattr(sd_model, component_name, None)
+        if component is not None and hasattr(component, 'named_modules'):
+            modules += list(component.named_modules())
     with pbar:
-        for component_name in ['text_encoder','text_encoder_2', 'unet', 'transformer']:
-            component = getattr(sd_model, component_name, None)
-            if component is not None:
-                applied = 0
-                modules = list(component.named_modules())
-                task_start = time.time()
-                task = pbar.add_task(description=component_name , total=len(modules), visible=False)
-                for _, module in modules:
-                    layer_name = getattr(module, 'network_layer_name', None)
-                    if layer_name is None:
-                        continue
-                    present = any([net.modules.get(layer_name, None) for net in loaded_networks]) # noqa: C419
-                    if present:
-                        network_apply_weights(module)
-                        applied += 1
-                    pbar.update(task, advance=1, visible=(time.time() - task_start) > 1) # progress bar becomes visible if operation takes more than 1sec
-                pbar.remove_task(task)
-                if debug:
-                    shared.log.debug(f'Load network: type=LoRA component={component_name} modules={len(modules)} applied={applied}')
+        task = pbar.add_task(description='Apply network: type=LoRA' , total=len(modules), visible=len(loaded_networks) > 0)
+        for _, module in modules:
+            network_apply_weights(module)
+            # pbar.update(task, advance=1) # progress bar becomes visible if operation takes more than 1sec
+        pbar.remove_task(task)
+    if debug:
+        shared.log.debug(f'Load network: type=LoRA modules={len(modules)}')
     if shared.opts.diffusers_offload_mode != "none":
         sd_models.set_diffuser_offload(sd_model, op="model")
     if debug:
-        shared.log.debug(f'Load network: type=LoRA total={total_time():.2f} timers={timer}')
+        shared.log.debug(f'Load network: type=LoRA timers{get_timers()}')
 
 
 def list_available_networks():
diff --git a/modules/processing.py b/modules/processing.py
index 16e7a9213..92faaee8d 100644
--- a/modules/processing.py
+++ b/modules/processing.py
@@ -473,4 +473,13 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
         p.scripts.postprocess(p, processed)
     timer.process.record('post')
     shared.log.info(f'Processed: images={len(output_images)} its={(p.steps * len(output_images)) / (t1 - t0):.2f} time={t1-t0:.2f} timers={timer.process.dct(min_time=0.02)} memory={memstats.memory_stats()}')
+
+    if shared.cmd_opts.malloc:
+        import tracemalloc
+        snapshot = tracemalloc.take_snapshot()
+        stats = snapshot.statistics('lineno')
+        shared.log.debug('Profile malloc:')
+        for stat in stats[:20]:
+            frame = stat.traceback[0]
+            shared.log.debug(f'  file="{frame.filename}":{frame.lineno} size={stat.size}')
     return processed
diff --git a/modules/processing_helpers.py b/modules/processing_helpers.py
index 22acf296c..ab08d4cc8 100644
--- a/modules/processing_helpers.py
+++ b/modules/processing_helpers.py
@@ -1,4 +1,5 @@
 import os
+import time
 import math
 import random
 import warnings
@@ -9,7 +10,7 @@ import cv2
 from PIL import Image
 from skimage import exposure
 from blendmodes.blend import blendLayers, BlendType
-from modules import shared, devices, images, sd_models, sd_samplers, sd_hijack_hypertile, processing_vae
+from modules import shared, devices, images, sd_models, sd_samplers, sd_hijack_hypertile, processing_vae, timer
 
 
 debug = shared.log.trace if os.environ.get('SD_PROCESS_DEBUG', None) is not None else lambda *args, **kwargs: None
@@ -352,6 +353,7 @@ def img2img_image_conditioning(p, source_image, latent_image, image_mask=None):
 
 
 def validate_sample(tensor):
+    t0 = time.time()
     if not isinstance(tensor, np.ndarray) and not isinstance(tensor, torch.Tensor):
         return tensor
     dtype = tensor.dtype
@@ -377,6 +379,8 @@ def validate_sample(tensor):
         if upcast is not None and not upcast:
             setattr(shared.sd_model.vae.config, 'force_upcast', True) # noqa: B010
             shared.log.warning('Decode: upcast=True set, retry operation')
+    t1 = time.time()
+    timer.process.add('validate', t1 - t0)
     return cast