Merge pull request #3682 from vladmandic/dev

refresh master

CHANGELOG.md

@@ -1,6 +1,6 @@
 # Change Log for SD.Next
 
-## Update for 2025-01-01
+## Update for 2025-01-02
 
 - [Allegro Video](https://huggingface.co/rhymes-ai/Allegro)  
   - optimizations: full offload and quantization support  
@@ -15,11 +15,17 @@
   - update installer messages  
 - **Detailer**:
   - add explicit detailer steps setting  
+- **SysInfo**:
+  - update to collected data and benchmarks  
 - **Fixes**:
   - explict clear caches on model load  
   - lock adetailer commit: `#a89c01d`  
-  - xyzgrid fix progress calculation  
+  - xyzgrid progress calculation  
+  - xyzgrid detailer
   - vae tiling use default value if not set  
+  - sd35 img2img
+  - samplers test for scale noise before using  
+  - scheduler api  
 
 ## Update for 2024-12-31
 

README.md

@@ -31,8 +31,9 @@ All individual features are not listed here, instead check [ChangeLog](CHANGELOG
 - Multiplatform!  
  ▹ **Windows | Linux | MacOS | nVidia | AMD | IntelArc/IPEX | DirectML | OpenVINO | ONNX+Olive | ZLUDA**
 - Platform specific autodetection and tuning performed on install  
-- Optimized processing with latest `torch` developments with built-in support for `torch.compile`  
-  and multiple compile backends: *Triton, StableFast, DeepCache, NNCF, OneDiff*  
+- Optimized processing with latest `torch` developments with built-in support for model compile, quantize and compress  
+  Compile backends: *Triton | StableFast | DeepCache | OneDiff*  
+  Quantization and compression methods: *BitsAndBytes | TorchAO | Optimum-Quanto | NNCF*  
 - Built-in queue management  
 - Built in installer with automatic updates and dependency management  
 - Mobile compatible  

modules/api/api.py

@@ -99,6 +99,10 @@ class Api:
         # gallery api
         gallery.register_api(app)
 
+        # compatibility api
+        self.text2imgapi = self.generate.post_text2img
+        self.img2imgapi = self.generate.post_img2img
+
     def add_api_route(self, path: str, endpoint, **kwargs):
         if (shared.cmd_opts.auth or shared.cmd_opts.auth_file) and shared.cmd_opts.api_only:
             return self.app.add_api_route(path, endpoint, dependencies=[Depends(self.auth)], **kwargs)

modules/api/models.py

@@ -1,5 +1,5 @@
 import inspect
-from typing import Any, Optional, Dict, List, Type, Callable
+from typing import Any, Optional, Dict, List, Type, Callable, Union
 from pydantic import BaseModel, Field, create_model # pylint: disable=no-name-in-module
 from inflection import underscore
 from modules.processing import StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img
@@ -194,7 +194,7 @@ ReqTxt2Img = PydanticModelGenerator(
     "StableDiffusionProcessingTxt2Img",
     StableDiffusionProcessingTxt2Img,
     [
-        {"key": "sampler_index", "type": int, "default": 0},
+        {"key": "sampler_index", "type": Union[int, str], "default": 0},
         {"key": "sampler_name", "type": str, "default": "UniPC"},
         {"key": "hr_sampler_name", "type": str, "default": "Same as primary"},
         {"key": "script_name", "type": str, "default": "none"},
@@ -218,7 +218,7 @@ ReqImg2Img = PydanticModelGenerator(
     "StableDiffusionProcessingImg2Img",
     StableDiffusionProcessingImg2Img,
     [
-        {"key": "sampler_index", "type": int, "default": 0},
+        {"key": "sampler_index", "type": Union[int, str], "default": 0},
         {"key": "sampler_name", "type": str, "default": "UniPC"},
         {"key": "hr_sampler_name", "type": str, "default": "Same as primary"},
         {"key": "script_name", "type": str, "default": "none"},
@@ -405,10 +405,6 @@ class ResNVML(BaseModel): # definition of http response
     state: str = Field(title="State")
 
 
-# compatibility items
-StableDiffusionTxt2ImgProcessingAPI = ResTxt2Img
-StableDiffusionImg2ImgProcessingAPI = ResImg2Img
-
 # helper function
 
 def create_model_from_signature(func: Callable, model_name: str, base_model: Type[BaseModel] = BaseModel, additional_fields: List = [], exclude_fields: List[str] = []):

modules/postprocess/yolo.py

@@ -1,5 +1,6 @@
 from typing import TYPE_CHECKING
 import os
+from copy import copy
 import numpy as np
 import gradio as gr
 from PIL import Image, ImageDraw
@@ -259,23 +260,23 @@ class YoloRestorer(Detailer):
 
             report = [{'label': i.label, 'score': i.score, 'size': f'{i.width}x{i.height}' } for i in items]
             shared.log.info(f'Detailer: model="{name}" items={report} args={items[0].args} denoise={p.denoising_strength} blur={p.mask_blur} width={p.width} height={p.height} padding={p.inpaint_full_res_padding}')
-            shared.log.debug(f'Detailer: prompt="{prompt}" negative="{negative}"')
+            # shared.log.debug(f'Detailer: prompt="{prompt}" negative="{negative}"')
             models_used.append(name)
 
             mask_all = []
             p.state = ''
             prev_state = shared.state.job
+            pc = copy(p)
             for item in items:
                 if item.mask is None:
                     continue
-                p.init_images = [image]
-                p.image_mask = [item.mask]
-                # mask_all.append(item.mask)
-                p.recursion = True
+                pc.init_images = [image]
+                pc.image_mask = [item.mask]
+                pc.overlay_images = []
+                pc.recursion = True
                 shared.state.job = 'Detailer'
-                pp = processing.process_images_inner(p)
-                del p.recursion
-                p.overlay_images = None # skip applying overlay twice
+                pp = processing.process_images_inner(pc)
+                del pc.recursion
                 if pp is not None and pp.images is not None and len(pp.images) > 0:
                     image = pp.images[0] # update image to be reused for next item
                     if len(pp.images) > 1:
@@ -298,13 +299,10 @@ class YoloRestorer(Detailer):
             if len(mask_all) > 0 and shared.opts.include_mask:
                 from modules.control.util import blend
                 p.image_mask = blend([np.array(m) for m in mask_all])
-                # combined = blend([np_image, p.image_mask])
-                # combined = Image.fromarray(combined)
-                # combined.save('/tmp/item.png')
                 p.image_mask = Image.fromarray(p.image_mask)
 
-        if len(models_used) > 0:
-            shared.log.debug(f'Detailer processed: models={models_used}')
+        # if len(models_used) > 0:
+        #     shared.log.debug(f'Detailer processed: models={models_used}')
         return np_image
 
     def ui(self, tab: str):

modules/processing.py

@@ -463,7 +463,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed:
         ipadapter.unapply(shared.sd_model, unload=getattr(p, 'ip_adapter_unload', False))
 
     if shared.opts.include_mask:
-        if shared.opts.mask_apply_overlay and p.overlay_images is not None and len(p.overlay_images):
+        if shared.opts.mask_apply_overlay and p.overlay_images is not None and len(p.overlay_images) > 0:
             p.image_mask = create_binary_mask(p.overlay_images[0])
             p.image_mask = ImageOps.invert(p.image_mask)
             output_images.append(p.image_mask)

modules/processing_args.py

@@ -307,6 +307,9 @@ def set_pipeline_args(p, model, prompts:list, negative_prompts:list, prompts_2:t
             if isinstance(args['image'], torch.Tensor) or isinstance(args['image'], np.ndarray):
                 args['width'] = 8 * args['image'].shape[-1]
                 args['height'] = 8 * args['image'].shape[-2]
+            elif isinstance(args['image'][0], torch.Tensor) or isinstance(args['image'][0], np.ndarray):
+                args['width'] = 8 * args['image'][0].shape[-1]
+                args['height'] = 8 * args['image'][0].shape[-2]
             else:
                 args['width'] = 8 * math.ceil(args['image'][0].width / 8)
                 args['height'] = 8 * math.ceil(args['image'][0].height / 8)

modules/processing_class.py

@@ -502,7 +502,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing):
                 image = images.resize_image(self.resize_mode, image, self.width, self.height, upscaler_name=self.resize_name, context=self.resize_context)
                 self.width = image.width
                 self.height = image.height
-            if self.image_mask is not None and shared.opts.mask_apply_overlay and not hasattr(self, 'xyz'):
+            if self.image_mask is not None and shared.opts.mask_apply_overlay:
                 image_masked = Image.new('RGBa', (image.width, image.height))
                 image_to_paste = image.convert("RGBA").convert("RGBa")
                 image_to_mask = ImageOps.invert(self.mask_for_overlay.convert('L')) if self.mask_for_overlay is not None else None

modules/sd_models.py

@@ -1288,7 +1288,7 @@ def set_diffuser_pipe(pipe, new_pipe_type):
         new_pipe.image_encoder = image_encoder
     if feature_extractor is not None:
         new_pipe.feature_extractor = feature_extractor
-    if new_pipe.__class__.__name__ == 'FluxPipeline':
+    if new_pipe.__class__.__name__ in ['FluxPipeline', 'StableDiffusion3Pipeline']:
         new_pipe.register_modules(image_encoder = image_encoder)
         new_pipe.register_modules(feature_extractor = feature_extractor)
     new_pipe.is_sdxl = getattr(pipe, 'is_sdxl', False) # a1111 compatibility item

modules/sd_samplers_diffusers.py

@@ -271,11 +271,16 @@ class DiffusionSampler:
         sampler = constructor(**self.config)
         accept_sigmas = "sigmas" in set(inspect.signature(sampler.set_timesteps).parameters.keys())
         accepts_timesteps = "timesteps" in set(inspect.signature(sampler.set_timesteps).parameters.keys())
+        accept_scale_noise = hasattr(sampler, "scale_noise")
         debug(f'Sampler: sampler="{name}" sigmas={accept_sigmas} timesteps={accepts_timesteps}')
         if ('Flux' in model.__class__.__name__) and (not accept_sigmas):
             shared.log.warning(f'Sampler: sampler="{name}" does not accept sigmas')
             self.sampler = None
             return
+        if ('StableDiffusion3' in model.__class__.__name__) and (not accept_scale_noise):
+            shared.log.warning(f'Sampler: sampler="{name}" does not implement scale noise')
+            self.sampler = None
+            return
         self.sampler = sampler
         if name == 'DC Solver':
             if not hasattr(self.sampler, 'dc_ratios'):

modules/shared.py

@@ -303,8 +303,12 @@ class OptionInfo:
                     # return False
         minimum = args.get("minimum", None)
         maximum = args.get("maximum", None)
-        if (minimum is not None and value < minimum) or (maximum is not None and value > maximum):
-            log.error(f'Setting validation: "{opt}"={value} default={self.default} minimum={minimum} maximum={maximum}')
+        try:
+            if (minimum is not None and value < minimum) or (maximum is not None and value > maximum):
+                log.error(f'Setting validation: "{opt}"={value} default={self.default} minimum={minimum} maximum={maximum}')
+                return False
+        except Exception as err:
+            log.error(f'Setting validation: "{opt}"={value} default={self.default} minimum={minimum} maximum={maximum} error={err}')
             return False
         return True
 

scripts/xyz_grid_on.py

@@ -325,12 +325,14 @@ class Script(scripts.Script):
             x_opt.apply(pc, x, xs)
             y_opt.apply(pc, y, ys)
             z_opt.apply(pc, z, zs)
+
             try:
                 processed = processing.process_images(pc)
             except Exception as e:
                 shared.log.error(f"XYZ grid: Failed to process image: {e}")
                 errors.display(e, 'XYZ grid')
                 processed = None
+
             if ix == 0 and iy == 0: # create subgrid info text
                 pc.extra_generation_params = copy(pc.extra_generation_params)
                 pc.extra_generation_params['Script'] = self.title()