Merge pull request #3549 from vladmandic/dev

refresh master

.eslintrc.json

@@ -90,6 +90,8 @@
     "getENActiveTab": "readonly",
     "quickApplyStyle": "readonly",
     "quickSaveStyle": "readonly",
+    "setupExtraNetworks": "readonly",
+    "showNetworks": "readonly",
     // from python
     "localization": "readonly",
     // progressbar.js

CHANGELOG.md

@@ -1,14 +1,10 @@
 # Change Log for SD.Next
 
-## Update for 2024-11-01
+## Update for 2024-11-02
 
-Smaller release just 3 days after the last one, but with some important fixes and improvements.  
+Smaller release just few days after the last one, but with some important fixes and improvements.  
 This release can be considered an LTS release before we kick off the next round of major updates.  
 
-- XYZ grid:
-  - optional per-image time benchmark info
-- UI:
-  - add additional [hotkeys](https://github.com/vladmandic/automatic/wiki/Hotkeys)
 - Docs:
   - add built-in [changelog](https://github.com/vladmandic/automatic/blob/master/CHANGELOG.md) search  
     since changelog is the best up-to-date source of info
@@ -21,6 +17,13 @@ This release can be considered an LTS release before we kick off the next round
 - SD3: all-in-one safetensors
   - *examples*: [large](https://civitai.com/models/882666/sd35-large-google-flan?modelVersionId=1003031), [medium](https://civitai.com/models/900327)
   - *note*: enable *bnb* on-the-fly quantization for even bigger gains  
+- UI:
+  - add additional [hotkeys](https://github.com/vladmandic/automatic/wiki/Hotkeys)
+  - add show networks on startup setting
+  - better mapping of networks previews  
+  - optimize networks display load  
+- XYZ grid:
+  - optional per-image time benchmark info
 - CLI:
   - refactor command line params  
     run `webui.sh`/`webui.bat` with `--help` to see all options  
@@ -28,6 +31,7 @@ This release can be considered an LTS release before we kick off the next round
   - Repo: move screenshots to GH pages
   - Update requirements
 - Fixes:
+  - custom watermark add alphablending  
   - detailer min/max size as fractions of image size  
   - ipadapter load on-demand  
   - ipadapter face use correct yolo model  
@@ -36,7 +40,10 @@ This release can be considered an LTS release before we kick off the next round
   - fix diffusers load from folder  
   - fix lora enum logging on windows  
   - fix xyz grid with batch count  
-  - move dowwloads of some auxillary models to hfcache instead of models folder  
+  - fix vqa models ignoring hfcache folder setting  
+  - fix network height in standard vs modern ui  
+  - fix k-diff enum on startup  
+  - move downloads of some auxillary models to hfcache instead of models folder  
 
 ## Update for 2024-10-29
 

cli/image-watermark.py

@@ -82,6 +82,7 @@ def watermark(params, file):
 
     exif = get_exif(image)
 
+    wm = None
     if params.command == 'read':
         fn = params.input
         wm = get_watermark(image, params)

html/previews.json

@@ -0,0 +1,11 @@
+{
+  "stabilityai--stable-diffusion-3-medium-diffusers": "models/Reference/stabilityai--stable-diffusion-3.jpg",
+  "stabilityai--stable-diffusion-3.5-medium": "models/Reference/stabilityai--stable-diffusion-3_5.jpg",
+  "stabilityai--stable-diffusion-3.5-large": "models/Reference/stabilityai--stable-diffusion-3_5.jpg",
+  "Disty0--FLUX.1-dev-qint8": "models/Reference/black-forest-labs--FLUX.1-dev.jpg",
+  "Disty0--FLUX.1-dev-qint4": "models/Reference/black-forest-labs--FLUX.1-dev.jpg",
+  "sayakpaul--flux.1-dev-nf4": "models/Reference/black-forest-labs--FLUX.1-dev.jpg",
+  "THUDM--CogVideoX-2b": "models/Reference/THUDM--CogView3-Plus-3B.jpg",
+  "THUDM--CogVideoX-5b": "models/Reference/THUDM--CogView3-Plus-3B.jpg",
+  "THUDM--CogVideoX-5b-I2V": "models/Reference/THUDM--CogView3-Plus-3B.jpg"
+}

javascript/extraNetworks.js

@@ -375,6 +375,7 @@ function setupExtraNetworksForTab(tabname) {
   if (!tabs) return;
 
   // buttons
+  const btnShow = gradioApp().getElementById(`${tabname}_extra_networks_btn`);
   const btnRefresh = gradioApp().getElementById(`${tabname}_extra_refresh`);
   const btnScan = gradioApp().getElementById(`${tabname}_extra_scan`);
   const btnSave = gradioApp().getElementById(`${tabname}_extra_save`);
@@ -448,50 +449,69 @@ function setupExtraNetworksForTab(tabname) {
 
   // en style
   if (!en) return;
+  let lastView;
+  let heightInitialized = false;
   const intersectionObserver = new IntersectionObserver((entries) => {
-    for (const el of Array.from(gradioApp().querySelectorAll('.extra-networks-page'))) {
-      el.style.height = `${window.opts.extra_networks_height}vh`;
-      el.parentElement.style.width = '-webkit-fill-available';
+    if (!heightInitialized) {
+      heightInitialized = true;
+      let h = 0;
+      const target = window.opts.extra_networks_card_cover === 'sidebar' ? 0 : window.opts.extra_networks_height;
+      if (window.opts.theme_type === 'Standard') h = target > 0 ? target : 55;
+      else h = target > 0 ? target : 87;
+      for (const el of Array.from(gradioApp().getElementById(`${tabname}_extra_tabs`).querySelectorAll('.extra-networks-page'))) {
+        if (h > 0) el.style.height = `${h}vh`;
+        el.parentElement.style.width = '-webkit-fill-available';
+      }
     }
-    if (entries[0].intersectionRatio > 0) {
-      refreshENpage();
-      // sortExtraNetworks('fixed');
-      if (window.opts.extra_networks_card_cover === 'cover') {
-        en.style.transition = '';
-        en.style.zIndex = 100;
-        en.style.top = '13em';
-        en.style.position = 'absolute';
-        en.style.right = 'unset';
-        en.style.width = 'unset';
-        en.style.height = 'unset';
-        gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = 'unset';
-      } else if (window.opts.extra_networks_card_cover === 'sidebar') {
-        en.style.zIndex = 100;
-        en.style.position = 'absolute';
-        en.style.right = '0';
-        en.style.top = '13em';
-        en.style.height = 'auto';
-        en.style.transition = 'width 0.3s ease';
-        en.style.width = `${window.opts.extra_networks_sidebar_width}vw`;
-        gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = `${100 - 2 - window.opts.extra_networks_sidebar_width}vw`;
+    if (lastView !== entries[0].intersectionRatio > 0) {
+      lastView = entries[0].intersectionRatio > 0;
+      if (lastView) {
+        refreshENpage();
+        // sortExtraNetworks('fixed');
+        if (window.opts.extra_networks_card_cover === 'cover') {
+          en.style.position = 'absolute';
+          en.style.height = 'unset';
+          en.style.width = 'unset';
+          en.style.right = 'unset';
+          en.style.top = '13em';
+          en.style.transition = '';
+          en.style.zIndex = 100;
+          gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = 'unset';
+        } else if (window.opts.extra_networks_card_cover === 'sidebar') {
+          en.style.position = 'absolute';
+          en.style.height = 'auto';
+          en.style.width = `${window.opts.extra_networks_sidebar_width}vw`;
+          en.style.right = '0';
+          en.style.top = '13em';
+          en.style.transition = 'width 0.3s ease';
+          en.style.zIndex = 100;
+          gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = `${100 - 2 - window.opts.extra_networks_sidebar_width}vw`;
+        } else {
+          en.style.position = 'relative';
+          en.style.height = 'unset';
+          en.style.width = 'unset';
+          en.style.right = 'unset';
+          en.style.top = 0;
+          en.style.transition = '';
+          en.style.zIndex = 0;
+          gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = 'unset';
+        }
       } else {
-        en.style.transition = '';
-        en.style.zIndex = 0;
-        en.style.top = 0;
-        en.style.position = 'relative';
-        en.style.right = 'unset';
-        en.style.width = 'unset';
-        en.style.height = 'unset';
+        if (window.opts.extra_networks_card_cover === 'sidebar') en.style.width = 0;
         gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = 'unset';
       }
-    } else {
-      if (window.opts.extra_networks_card_cover === 'sidebar') en.style.width = 0;
-      gradioApp().getElementById(`${tabname}_settings`).parentNode.style.width = 'unset';
     }
   });
   intersectionObserver.observe(en); // monitor visibility
 }
 
+async function showNetworks() {
+  for (const tabname of ['txt2img', 'img2img', 'control']) {
+    if (window.opts.extra_networks_show) gradioApp().getElementById(`${tabname}_extra_networks_btn`).click();
+  }
+  log('showNetworks');
+}
+
 async function setupExtraNetworks() {
   setupExtraNetworksForTab('txt2img');
   setupExtraNetworksForTab('img2img');

javascript/loader.js

@@ -3,7 +3,7 @@ const appStartTime = performance.now();
 async function preloadImages() {
   const dark = window.matchMedia && window.matchMedia('(prefers-color-scheme: dark)').matches;
   const imagePromises = [];
-  const num = Math.floor(10 * Math.random());
+  const num = Math.floor(9.99 * Math.random());
   const imageUrls = [
     `file=html/logo-bg-${dark ? 'dark' : 'light'}.jpg`,
     `file=html/logo-bg-${num}.jpg`,
@@ -27,7 +27,7 @@ async function preloadImages() {
 async function createSplash() {
   const dark = window.matchMedia && window.matchMedia('(prefers-color-scheme: dark)').matches;
   log('createSplash', { theme: dark ? 'dark' : 'light' });
-  const num = Math.floor(11 * Math.random());
+  const num = Math.floor(9.99 * Math.random());
   const splash = `
     <div id="splash" class="splash" style="background: ${dark ? 'black' : 'white'}">
       <div class="loading"><div class="loader"></div></div>

javascript/sdnext.css

@@ -203,7 +203,7 @@ table.settings-value-table td { padding: 0.4em; border: 1px solid #ccc; max-widt
 #extensions .date { opacity: 0.85; font-size: var(--text-sm); }
 
 /* extra networks */
-#txt2img_extra_networks, #img2img_extra_networks, #control_extra_networks { width: 0; }
+.extra_networks_root { width: 0; position: absolute; height: auto; right: 0; top: 13em; z-index: 100; } /* default is sidebar view */
 .extra-networks { background: var(--background-color); padding: var(--block-label-padding); }
 .extra-networks > div { margin: 0; border-bottom: none !important; gap: 0.3em 0; }
 .extra-networks .second-line { display: flex; width: -moz-available; width: -webkit-fill-available; gap: 0.3em; box-shadow: var(--input-shadow); }

javascript/startup.js

@@ -32,6 +32,7 @@ async function initStartup() {
   removeSplash();
 
   // post startup tasks that may take longer but are not critical
+  showNetworks();
   setHints();
   applyStyles();
   initIndexDB();

modules/images.py

@@ -374,6 +374,8 @@ def set_watermark(image, watermark):
         wm_image = None
         try:
             wm_image = Image.open(shared.opts.image_watermark_image)
+            if wm_image.mode != 'RGBA':
+                wm_image = wm_image.convert('RGBA')
         except Exception as e:
             shared.log.warning(f'Set image watermark: fn="{shared.opts.image_watermark_image}" {e}')
         if wm_image is not None:
@@ -392,8 +394,14 @@ def set_watermark(image, watermark):
             try:
                 for x in range(wm_image.width):
                     for y in range(wm_image.height):
-                        r, g, b, _a = wm_image.getpixel((x, y))
-                        if not (r == 0 and g == 0 and b == 0):
+                        rgba = wm_image.getpixel((x, y))
+                        orig = image.getpixel((x+position[0], y+position[1]))
+                        # alpha blend
+                        a = rgba[3] / 255
+                        r = int(rgba[0] * a + orig[0] * (1 - a))
+                        g = int(rgba[1] * a + orig[1] * (1 - a))
+                        b = int(rgba[2] * a + orig[2] * (1 - a))
+                        if not a == 0:
                             image.putpixel((x+position[0], y+position[1]), (r, g, b))
                 shared.log.debug(f'Set image watermark: fn="{shared.opts.image_watermark_image}" image={wm_image} position={position}')
             except Exception as e:

modules/postprocess/yolo.py

@@ -25,7 +25,6 @@ class YoloResult:
         self.box = box
         self.mask = mask
         self.item = item
-        self.size = size
         self.width = width
         self.height = height
         self.args = args
@@ -127,17 +126,17 @@ class YoloRestorer(Detailer):
                 box = box.tolist()
                 mask_image = None
                 w, h = box[2] - box[0], box[3] - box[1]
-                size = w * h / (image.width * image.height)
-                min_size = (shared.opts.detailer_min_size if shared.opts.detailer_min_size > 0 else 0) * min(w, h)
-                max_size = (shared.opts.detailer_max_size if shared.opts.detailer_max_size > 0 else 1) * max(w, h)
-                if (min(w, h) > min_size) and (max(w, h) < max_size):
+                x_size, y_size = w/image.width, h/image.height
+                min_size = shared.opts.detailer_min_size if shared.opts.detailer_min_size > 0 and shared.opts.detailer_min_size < 1 else 0
+                max_size = shared.opts.detailer_max_size if shared.opts.detailer_max_size > 0 and shared.opts.detailer_max_size < 1 else 1
+                if x_size >= min_size and y_size >=min_size and x_size <= max_size and y_size <= max_size:
                     if mask:
                         mask_image = image.copy()
                         mask_image = Image.new('L', image.size, 0)
                         draw = ImageDraw.Draw(mask_image)
                         draw.rectangle(box, fill="white", outline=None, width=0)
                         cropped = image.crop(box)
-                        result.append(YoloResult(cls=cls, label=label, score=round(score, 2), box=box, mask=mask_image, item=cropped, size=size, width=w, height=h, args=args))
+                        result.append(YoloResult(cls=cls, label=label, score=round(score, 2), box=box, mask=mask_image, item=cropped, width=w, height=h, args=args))
                 if len(result) >= shared.opts.detailer_max:
                     break
         return result

modules/shared.py

@@ -877,8 +877,9 @@ options_templates.update(options_section(('interrogate', "Interrogate"), {
 }))
 
 options_templates.update(options_section(('extra_networks', "Networks"), {
-    "extra_networks_sep1": OptionInfo("<h2>Extra networks UI</h2>", "", gr.HTML),
-    "extra_networks": OptionInfo(["All"], "Networks", gr.Dropdown, lambda: {"multiselect":True, "choices": ['All'] + [en.title for en in extra_networks]}),
+    "extra_networks_sep1": OptionInfo("<h2>Networks UI</h2>", "", gr.HTML),
+    "extra_networks_show": OptionInfo(True, "UI show on startup"),
+    "extra_networks": OptionInfo(["All"], "Available networks", gr.Dropdown, lambda: {"multiselect":True, "choices": ['All'] + [en.title for en in extra_networks]}),
     "extra_networks_sort": OptionInfo("Default", "Sort order", gr.Dropdown, {"choices": ['Default', 'Name [A-Z]', 'Name [Z-A]', 'Date [Newest]', 'Date [Oldest]', 'Size [Largest]', 'Size [Smallest]']}),
     "extra_networks_view": OptionInfo("gallery", "UI view", gr.Radio, {"choices": ["gallery", "list"]}),
     "extra_networks_card_cover": OptionInfo("sidebar", "UI position", gr.Radio, {"choices": ["cover", "inline", "sidebar"]}),
@@ -888,13 +889,18 @@ options_templates.update(options_section(('extra_networks', "Networks"), {
     "extra_networks_card_square": OptionInfo(True, "UI disable variable aspect ratio"),
     "extra_networks_fetch": OptionInfo(True, "UI fetch network info on mouse-over"),
     "extra_networks_card_fit": OptionInfo("cover", "UI image contain method", gr.Radio, {"choices": ["contain", "cover", "fill"], "visible": False}),
-    "extra_networks_sep2": OptionInfo("<h2>Extra networks general</h2>", "", gr.HTML),
-    "extra_network_reference": OptionInfo(False, "Use reference values when available", gr.Checkbox),
     "extra_network_skip_indexing": OptionInfo(False, "Build info on first access", gr.Checkbox),
-    "extra_networks_default_multiplier": OptionInfo(1.0, "Default strength", gr.Slider, {"minimum": 0.0, "maximum": 2.0, "step": 0.01}),
+
+    "extra_networks_model_sep": OptionInfo("<h2>Models</h2>", "", gr.HTML),
+    "extra_network_reference": OptionInfo(False, "Use reference values when available", gr.Checkbox),
+    "extra_networks_embed_sep": OptionInfo("<h2>Embeddings</h2>", "", gr.HTML),
     "diffusers_convert_embed": OptionInfo(False, "Auto-convert SD 1.5 embeddings to SDXL ", gr.Checkbox, {"visible": native}),
-    "extra_networks_sep3": OptionInfo("<h2>Extra networks settings</h2>", "", gr.HTML),
+    "extra_networks_styles_sep": OptionInfo("<h2>Styles</h2>", "", gr.HTML),
     "extra_networks_styles": OptionInfo(True, "Show built-in styles"),
+    "extra_networks_wildcard_sep": OptionInfo("<h2>Wildcards</h2>", "", gr.HTML),
+    "wildcards_enabled": OptionInfo(True, "Enable file wildcards support"),
+    "extra_networks_lora_sep": OptionInfo("<h2>LoRA</h2>", "", gr.HTML),
+    "extra_networks_default_multiplier": OptionInfo(1.0, "Default strength", gr.Slider, {"minimum": 0.0, "maximum": 2.0, "step": 0.01}),
     "lora_preferred_name": OptionInfo("filename", "LoRA preferred name", gr.Radio, {"choices": ["filename", "alias"]}),
     "lora_add_hashes_to_infotext": OptionInfo(False, "LoRA add hash info"),
     "lora_force_diffusers": OptionInfo(False if not cmd_opts.use_openvino else True, "LoRA force loading of all models using Diffusers"),
@@ -905,9 +911,9 @@ options_templates.update(options_section(('extra_networks', "Networks"), {
     "lora_quant": OptionInfo("NF4","LoRA precision in quantized models", gr.Radio, {"choices": ["NF4", "FP4"]}),
     "lora_functional": OptionInfo(False, "Use Kohya method for handling multiple LoRA", gr.Checkbox, { "visible": False }),
     "lora_load_gpu": OptionInfo(True if not cmd_opts.lowvram else False, "Load LoRA directly to GPU"),
-    "hypernetwork_enabled": OptionInfo(False, "Enable Hypernetwork support"),
+
+    "hypernetwork_enabled": OptionInfo(False, "Enable Hypernetwork support", gr.Checkbox, {"visible": False}),
     "sd_hypernetwork": OptionInfo("None", "Add hypernetwork to prompt", gr.Dropdown, { "choices": ["None"], "visible": False }),
-    "wildcards_enabled": OptionInfo(True, "Enable file wildcards support"),
 }))
 
 options_templates.update(options_section((None, "Hidden options"), {

modules/ui_control.py

@@ -152,7 +152,7 @@ def create_ui(_blocks: gr.Blocks=None):
             with gr.Row():
                 override_settings = ui_common.create_override_inputs('control')
 
-            with gr.Row(variant='compact', elem_id="control_extra_networks", visible=False) as extra_networks_ui:
+            with gr.Row(variant='compact', elem_id="control_extra_networks", elem_classes=["extra_networks_root"], visible=False) as extra_networks_ui:
                 from modules import timer, ui_extra_networks
                 extra_networks_ui = ui_extra_networks.create_ui(extra_networks_ui, btn_extra, 'control', skip_indexing=shared.opts.extra_network_skip_indexing)
                 timer.startup.record('ui-networks')

modules/ui_extra_networks.py

@@ -50,6 +50,7 @@ card_list = '''
         </div>
     </div>
 '''
+preview_map = None
 
 
 def init_api(app):
@@ -350,6 +351,9 @@ class ExtraNetworksPage:
         return self.link_preview(preview_file)
 
     def update_all_previews(self, items):
+        global preview_map # pylint: disable=global-statement
+        if preview_map is None:
+            preview_map = shared.readfile('html/previews.json', silent=True)
         t0 = time.time()
         reference_path = os.path.abspath(os.path.join('models', 'Reference'))
         possible_paths = list(set([os.path.dirname(item['filename']) for item in items] + [reference_path]))
@@ -378,8 +382,14 @@ class ExtraNetworksPage:
                         self.missing_thumbs.append(all_previews[file_idx])
                     item['preview'] = self.link_preview(all_previews[file_idx])
                     break
+            if item.get('preview', None) is None:
+                found = preview_map.get(base, None)
+                if found is not None:
+                    item['preview'] = self.link_preview(found)
+                    debug(f'EN mapped-preview: {item["name"]}={found}')
             if item.get('preview', None) is None:
                 item['preview'] = self.link_preview('html/card-no-preview.png')
+                debug(f'EN missing-preview: {item["name"]}')
         self.preview_time += time.time() - t0
 
 

modules/ui_img2img.py

@@ -41,7 +41,7 @@ def create_ui():
         img2img_prompt, img2img_prompt_styles, img2img_negative_prompt, img2img_submit, img2img_reprocess, img2img_paste, img2img_extra_networks_button, img2img_token_counter, img2img_token_button, img2img_negative_token_counter, img2img_negative_token_button = ui_sections.create_toprow(is_img2img=True, id_part="img2img")
         img2img_prompt_img = gr.File(label="", elem_id="img2img_prompt_image", file_count="single", type="binary", visible=False)
 
-        with gr.Row(variant='compact', elem_id="img2img_extra_networks", visible=False) as extra_networks_ui:
+        with gr.Row(variant='compact', elem_id="img2img_extra_networks", elem_classes=["extra_networks_root"], visible=False) as extra_networks_ui:
             from modules import ui_extra_networks
             extra_networks_ui_img2img = ui_extra_networks.create_ui(extra_networks_ui, img2img_extra_networks_button, 'img2img', skip_indexing=shared.opts.extra_network_skip_indexing)
             timer.startup.record('ui-networks')

modules/ui_txt2img.py

@@ -25,7 +25,7 @@ def create_ui():
         txt_prompt_img = gr.File(label="", elem_id="txt2img_prompt_image", file_count="single", type="binary", visible=False)
         txt_prompt_img.change(fn=modules.images.image_data, inputs=[txt_prompt_img], outputs=[txt2img_prompt, txt_prompt_img])
 
-        with gr.Row(variant='compact', elem_id="txt2img_extra_networks", visible=False) as extra_networks_ui:
+        with gr.Row(variant='compact', elem_id="txt2img_extra_networks", elem_classes=["extra_networks_root"], visible=False) as extra_networks_ui:
             from modules import ui_extra_networks
             extra_networks_ui = ui_extra_networks.create_ui(extra_networks_ui, txt2img_extra_networks_button, 'txt2img', skip_indexing=shared.opts.extra_network_skip_indexing)
             timer.startup.record('ui-networks')

modules/vqa.py

@@ -30,8 +30,8 @@ MODELS = {
 def git(question: str, image: Image.Image, repo: str = None):
     global processor, model, loaded # pylint: disable=global-statement
     if model is None or loaded != repo:
-        model = transformers.GitForCausalLM.from_pretrained(repo)
-        processor = transformers.GitProcessor.from_pretrained(repo)
+        model = transformers.GitForCausalLM.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
+        processor = transformers.GitProcessor.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
         loaded = repo
     model.to(devices.device, devices.dtype)
     shared.log.debug(f'VQA: class={model.__class__.__name__} processor={processor.__class__} model={repo}')
@@ -55,8 +55,8 @@ def git(question: str, image: Image.Image, repo: str = None):
 def blip(question: str, image: Image.Image, repo: str = None):
     global processor, model, loaded # pylint: disable=global-statement
     if model is None or loaded != repo:
-        model = transformers.BlipForQuestionAnswering.from_pretrained(repo)
-        processor = transformers.BlipProcessor.from_pretrained(repo)
+        model = transformers.BlipForQuestionAnswering.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
+        processor = transformers.BlipProcessor.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
         loaded = repo
     model.to(devices.device, devices.dtype)
     inputs = processor(image, question, return_tensors="pt")
@@ -73,8 +73,8 @@ def blip(question: str, image: Image.Image, repo: str = None):
 def vilt(question: str, image: Image.Image, repo: str = None):
     global processor, model, loaded # pylint: disable=global-statement
     if model is None or loaded != repo:
-        model = transformers.ViltForQuestionAnswering.from_pretrained(repo)
-        processor = transformers.ViltProcessor.from_pretrained(repo)
+        model = transformers.ViltForQuestionAnswering.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
+        processor = transformers.ViltProcessor.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
         loaded = repo
     model.to(devices.device)
     shared.log.debug(f'VQA: class={model.__class__.__name__} processor={processor.__class__} model={repo}')
@@ -94,8 +94,8 @@ def vilt(question: str, image: Image.Image, repo: str = None):
 def pix(question: str, image: Image.Image, repo: str = None):
     global processor, model, loaded # pylint: disable=global-statement
     if model is None or loaded != repo:
-        model = transformers.Pix2StructForConditionalGeneration.from_pretrained(repo)
-        processor = transformers.Pix2StructProcessor.from_pretrained(repo)
+        model = transformers.Pix2StructForConditionalGeneration.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
+        processor = transformers.Pix2StructProcessor.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
         loaded = repo
     model.to(devices.device)
     shared.log.debug(f'VQA: class={model.__class__.__name__} processor={processor.__class__} model={repo}')
@@ -115,8 +115,8 @@ def pix(question: str, image: Image.Image, repo: str = None):
 def moondream(question: str, image: Image.Image, repo: str = None):
     global processor, model, loaded # pylint: disable=global-statement
     if model is None or loaded != repo:
-        model = transformers.AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True) # revision = "2024-03-05"
-        processor = transformers.AutoTokenizer.from_pretrained(repo) # revision = "2024-03-05"
+        model = transformers.AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True, cache_dir=shared.opts.hfcache_dir) # revision = "2024-03-05"
+        processor = transformers.AutoTokenizer.from_pretrained(repo, cache_dir=shared.opts.hfcache_dir)
         loaded = repo
         model.eval()
     model.to(devices.device, devices.dtype)
@@ -142,8 +142,8 @@ def florence(question: str, image: Image.Image, repo: str = None, revision: str
         return R
     if model is None or loaded != repo:
         transformers.dynamic_module_utils.get_imports = get_imports
-        model = transformers.AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True, revision=revision)
-        processor = transformers.AutoProcessor.from_pretrained(repo, trust_remote_code=True, revision=revision)
+        model = transformers.AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True, revision=revision, cache_dir=shared.opts.hfcache_dir)
+        processor = transformers.AutoProcessor.from_pretrained(repo, trust_remote_code=True, revision=revision, cache_dir=shared.opts.hfcache_dir)
         transformers.dynamic_module_utils.get_imports = _get_imports
         loaded = repo
         model.eval()

scripts/k_diff.py

@@ -1,5 +1,4 @@
 import inspect
-import importlib
 import gradio as gr
 import diffusers
 from modules import scripts, processing, shared, sd_models
@@ -8,10 +7,7 @@ from modules import scripts, processing, shared, sd_models
 class Script(scripts.Script):
     supported_models = ['sd', 'sdxl']
     orig_pipe = None
-    try:
-        library = importlib.import_module('k_diffusion')
-    except Exception:
-        library = None
+    library = None
 
     def title(self):
         return 'K-Diffusion'
@@ -28,10 +24,8 @@ class Script(scripts.Script):
 
     def samplers(self):
         samplers = []
-        sampling = getattr(self.library, 'sampling', None)
-        if sampling is None:
-            return samplers
-        for s in dir(sampling):
+        from modules import sd_samplers_kdiffusion
+        for s in dir(sd_samplers_kdiffusion.k_sampling):
             if s.startswith('sample_'):
                 samplers.append(s.replace('sample_', ''))
         return samplers
@@ -43,8 +37,6 @@ class Script(scripts.Script):
         if shared.sd_model_type not in self.supported_models:
             shared.log.warning(f'K-Diffusion: class={shared.sd_model.__class__.__name__} model={shared.sd_model_type} required={self.supported_models}')
             return None
-        if self.library is None:
-            return
         cls = None
         if shared.sd_model_type == "sd":
             cls = diffusers.pipelines.StableDiffusionKDiffusionPipeline