profiling

Signed-off-by: Vladimir Mandic <mandic00@live.com>

CHANGELOG.md

@@ -124,6 +124,7 @@
 ### Fixes  
 
 - add `SD_NO_CACHE=true` env variable to disable file/folder caching  
+- add settings -> networks -> embeddings -> enable/disable
 - update `diffusers`  
 - fix README links  
 - fix sdxl controlnet single-file loader  

modules/errors.py

@@ -59,14 +59,14 @@ def exception(suppress=[]):
     console.print_exception(show_locals=False, max_frames=16, extra_lines=2, suppress=suppress, theme="ansi_dark", word_wrap=False, width=min([console.width, 200]))
 
 
-def profile(profiler, msg: str, n: int = 5):
+def profile(profiler, msg: str, n: int = 16):
     profiler.disable()
     import io
     import pstats
     stream = io.StringIO() # pylint: disable=abstract-class-instantiated
     p = pstats.Stats(profiler, stream=stream)
     p.sort_stats(pstats.SortKey.CUMULATIVE)
-    p.print_stats(100)
+    p.print_stats(200)
     # p.print_title()
     # p.print_call_heading(10, 'time')
     # p.print_callees(10)
@@ -81,6 +81,7 @@ def profile(profiler, msg: str, n: int = 5):
              and '_lsprof' not in x
              and '/profiler' not in x
              and 'rich' not in x
+             and 'profile_torch' not in x
              and x.strip() != ''
             ]
     txt = '\n'.join(lines[:min(n, len(lines))])

modules/ipadapter.py

@@ -14,6 +14,7 @@ from modules import processing, shared, devices, sd_models
 
 clip_repo = "h94/IP-Adapter"
 clip_loaded = None
+adapters_loaded = []
 ADAPTERS_NONE = {
     'None': { 'name': 'none', 'repo': 'none', 'subfolder': 'none' },
 }
@@ -129,9 +130,12 @@ def crop_images(images, crops):
 
 
 def unapply(pipe): # pylint: disable=arguments-differ
+    if len(adapters_loaded) == 0:
+        return
     try:
         if hasattr(pipe, 'set_ip_adapter_scale'):
             pipe.set_ip_adapter_scale(0)
+            pipe.unload_ip_adapter()
         if hasattr(pipe, 'unet') and hasattr(pipe.unet, 'config') and pipe.unet.config.encoder_hid_dim_type == 'ip_image_proj':
             pipe.unet.encoder_hid_proj = None
             pipe.config.encoder_hid_dim_type = None
@@ -141,7 +145,7 @@ def unapply(pipe): # pylint: disable=arguments-differ
 
 
 def apply(pipe, p: processing.StableDiffusionProcessing, adapter_names=[], adapter_scales=[1.0], adapter_crops=[False], adapter_starts=[0.0], adapter_ends=[1.0], adapter_images=[]):
-    global clip_loaded # pylint: disable=global-statement
+    global clip_loaded, adapters_loaded # pylint: disable=global-statement
     # overrides
     if hasattr(p, 'ip_adapter_names'):
         if isinstance(p.ip_adapter_names, str):
@@ -274,6 +278,7 @@ def apply(pipe, p: processing.StableDiffusionProcessing, adapter_names=[], adapt
         subfolders = [adapter['subfolder'] for adapter in adapters]
         names = [adapter['name'] for adapter in adapters]
         pipe.load_ip_adapter(repos, subfolder=subfolders, weight_name=names)
+        adapters_loaded = names
         if hasattr(p, 'ip_adapter_layers'):
             pipe.set_ip_adapter_scale(p.ip_adapter_layers)
             ip_str = ';'.join(adapter_names) + ':' + json.dumps(p.ip_adapter_layers)

modules/processing.py

@@ -179,6 +179,8 @@ def process_images(p: StableDiffusionProcessing) -> Processed:
         timer.process.record('pre')
 
         if shared.cmd_opts.profile:
+            timer.startup.profile = True
+            timer.process.profile = True
             with context_hypertile_vae(p), context_hypertile_unet(p):
                 import torch.profiler # pylint: disable=redefined-outer-name
                 activities=[torch.profiler.ProfilerActivity.CPU]
@@ -476,7 +478,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
         p.scripts.postprocess(p, processed)
     timer.process.record('post')
     if not p.disable_extra_networks:
-        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()}')
+        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()} memory={memstats.memory_stats()}')
 
     devices.torch_gc(force=True)
     return processed

modules/processing_diffusers.py

@@ -4,6 +4,7 @@ import time
 import numpy as np
 import torch
 import torchvision.transforms.functional as TF
+from PIL import Image
 from modules import shared, devices, processing, sd_models, errors, sd_hijack_hypertile, processing_vae, sd_models_compile, hidiffusion, timer, modelstats, extra_networks
 from modules.processing_helpers import resize_hires, calculate_base_steps, calculate_hires_steps, calculate_refiner_steps, save_intermediate, update_sampler, is_txt2img, is_refiner_enabled
 from modules.processing_args import set_pipeline_args
@@ -447,6 +448,9 @@ def process_diffusers(p: processing.StableDiffusionProcessing):
 
     sd_models_compile.openvino_recompile_model(p, hires=False, refiner=False) # recompile if a parameter changes
 
+    if hasattr(p, 'dummy'):
+        images = [Image.new(mode='RGB', size=(p.width, p.height))]
+        return images
     if 'base' not in p.skip:
         output = process_base(p)
     else:

modules/prompt_parser_diffusers.py

@@ -63,6 +63,8 @@ class PromptEmbedder:
         self.positive_schedule = None
         self.negative_schedule = None
         self.scheduled_prompt = False
+        if hasattr(p, 'dummy'):
+            return
         earlyout = self.checkcache(p)
         if earlyout:
             return

modules/shared.py

@@ -931,8 +931,11 @@ options_templates.update(options_section(('extra_networks', "Networks"), {
 
     "extra_networks_styles_sep": OptionInfo("<h2>Styles</h2>", "", gr.HTML),
     "extra_networks_styles": OptionInfo(True, "Show built-in styles"),
+
     "extra_networks_embed_sep": OptionInfo("<h2>Embeddings</h2>", "", gr.HTML),
-    "diffusers_convert_embed": OptionInfo(False, "Auto-convert SD15 embeddings to SDXL ", gr.Checkbox, {"visible": native}),
+    "diffusers_enable_embed": OptionInfo(True, "Enable embeddings support", gr.Checkbox, {"visible": native}),
+    "diffusers_convert_embed": OptionInfo(False, "Auto-convert SD15 embeddings to SDXL", gr.Checkbox, {"visible": native}),
+
     "extra_networks_wildcard_sep": OptionInfo("<h2>Wildcards</h2>", "", gr.HTML),
     "wildcards_enabled": OptionInfo(True, "Enable file wildcards support"),
 }))

modules/textual_inversion/textual_inversion.py

@@ -274,6 +274,8 @@ class EmbeddingDatabase:
         overwrite = bool(data)
         if not shared.sd_loaded:
             return
+        if not shared.opts.diffusers_enable_embed:
+            return
         embeddings, skipped = open_embeddings(filename) or convert_bundled(data)
         for skip in skipped:
             self.skipped_embeddings[skip.name] = skipped

modules/timer.py

@@ -7,6 +7,7 @@ class Timer:
         self.start = time.time()
         self.records = {}
         self.total = 0
+        self.profile = False
 
     def elapsed(self, reset=True):
         end = time.time()
@@ -27,11 +28,12 @@ class Timer:
             category = sys._getframe(1).f_code.co_name # pylint: disable=protected-access
         if category not in self.records:
             self.records[category] = 0
-
         self.records[category] += e + extra_time
         self.total += e + extra_time
 
     def summary(self, min_time=0.05, total=True):
+        if self.profile:
+            min_time = -1
         res = f"{self.total:.2f} " if total else ''
         additions = [x for x in self.records.items() if x[1] >= min_time]
         if not additions:
@@ -40,6 +42,8 @@ class Timer:
         return res
 
     def dct(self, min_time=0.05):
+        if self.profile:
+            return {k: round(v, 4) for k, v in self.records.items()}
         return {k: round(v, 2) for k, v in self.records.items() if v >= min_time}
 
     def reset(self):