From 43bdaa2f0eda79c685792b06a2bd84c65806a48f Mon Sep 17 00:00:00 2001 From: missionfloyd Date: Thu, 25 May 2023 18:49:28 -0600 Subject: Make ctrl+left/right optional --- modules/shared.py | 1 + 1 file changed, 1 insertion(+) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index b3508883..280c4451 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -405,6 +405,7 @@ options_templates.update(options_section(('ui', "User interface"), { "keyedit_precision_attention": OptionInfo(0.1, "Ctrl+up/down precision when editing (attention:1.1)", gr.Slider, {"minimum": 0.01, "maximum": 0.2, "step": 0.001}), "keyedit_precision_extra": OptionInfo(0.05, "Ctrl+up/down precision when editing ", gr.Slider, {"minimum": 0.01, "maximum": 0.2, "step": 0.001}), "keyedit_delimiters": OptionInfo(".,\\/!?%^*;:{}=`~()", "Ctrl+up/down word delimiters"), + "keyedit_move": OptionInfo(False, "Ctrl+left/right moves prompt elements"), "quicksettings_list": OptionInfo(["sd_model_checkpoint"], "Quicksettings list", ui_components.DropdownMulti, lambda: {"choices": list(opts.data_labels.keys())}), "hidden_tabs": OptionInfo([], "Hidden UI tabs (requires restart)", ui_components.DropdownMulti, lambda: {"choices": [x for x in tab_names]}), "ui_reorder": OptionInfo(", ".join(ui_reorder_categories), "txt2img/img2img UI item order"), -- cgit v1.2.3 From f603275d84301b5ee952683e951dd1aad72ba615 Mon Sep 17 00:00:00 2001 From: Jared Deckard Date: Thu, 15 Jun 2023 10:55:53 -0500 Subject: Add an opt-in infotext user name setting --- modules/processing.py | 2 +- modules/shared.py | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'modules/shared.py') diff --git a/modules/processing.py b/modules/processing.py index 3e8d153e..a0cc8db2 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -580,7 +580,7 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments=None, iter "NGMS": None if p.s_min_uncond == 0 else p.s_min_uncond, **p.extra_generation_params, "Version": program_version() if opts.add_version_to_infotext else None, - "User": p.user, + "User": p.user if opts.add_user_name_to_info else None, } generation_params_text = ", ".join([k if k == v else f'{k}: {generation_parameters_copypaste.quote(v)}' for k, v in generation_params.items() if v is not None]) diff --git a/modules/shared.py b/modules/shared.py index 271a062d..4c639a21 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -496,6 +496,7 @@ options_templates.update(options_section(('ui', "User interface"), { options_templates.update(options_section(('infotext', "Infotext"), { "add_model_hash_to_info": OptionInfo(True, "Add model hash to generation information"), "add_model_name_to_info": OptionInfo(True, "Add model name to generation information"), + "add_user_name_to_info": OptionInfo(False, "Add user name to generation information when authenticated"), "add_version_to_infotext": OptionInfo(True, "Add program version to generation information"), "disable_weights_auto_swap": OptionInfo(True, "When reading generation parameters from text into UI (from PNG info or pasted text), do not change the selected model/checkpoint."), })) -- cgit v1.2.3 From 41363e0d27bbaa0c84eebe3c7c8451075390ec4e Mon Sep 17 00:00:00 2001 From: dhwz Date: Fri, 16 Jun 2023 18:10:15 +0200 Subject: fix very slow loading speed of .safetensors files --- modules/sd_models.py | 7 +++++-- modules/shared.py | 1 + 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/sd_models.py b/modules/sd_models.py index 918f6fd6..d9ac675b 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -247,8 +247,11 @@ def read_metadata_from_safetensors(filename): def read_state_dict(checkpoint_file, print_global_state=False, map_location=None): _, extension = os.path.splitext(checkpoint_file) if extension.lower() == ".safetensors": - device = map_location or shared.weight_load_location or devices.get_optimal_device_name() - pl_sd = safetensors.torch.load_file(checkpoint_file, device=device) + if not shared.opts.disable_mmap_load_safetensors: + device = map_location or shared.weight_load_location or devices.get_optimal_device_name() + pl_sd = safetensors.torch.load_file(checkpoint_file, device=device) + else: + pl_sd = safetensors.torch.load(open(checkpoint_file, 'rb').read()) else: pl_sd = torch.load(checkpoint_file, map_location=map_location or shared.weight_load_location) diff --git a/modules/shared.py b/modules/shared.py index 91c31d55..6b0ccac1 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -376,6 +376,7 @@ options_templates.update(options_section(('system', "System"), { "multiple_tqdm": OptionInfo(True, "Add a second progress bar to the console that shows progress for an entire job."), "print_hypernet_extra": OptionInfo(False, "Print extra hypernetwork information to console."), "list_hidden_files": OptionInfo(True, "Load models/files in hidden directories").info("directory is hidden if its name starts with \".\""), + "disable_mmap_load_safetensors": OptionInfo(False, "Disable memmapping for loading .safetensors files (fixes very slow loading speed in some cases)."), })) options_templates.update(options_section(('training', "Training"), { -- cgit v1.2.3 From 24129368f1b732be25ef486edb2cf5a6ace66737 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 27 Jun 2023 09:19:04 +0300 Subject: send tensors to the correct device when loading from safetensors file with memmap disabled for #11260 --- modules/sd_models.py | 4 +++- modules/shared.py | 2 +- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/sd_models.py b/modules/sd_models.py index 0391398a..f65f4e36 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -246,11 +246,13 @@ def read_metadata_from_safetensors(filename): def read_state_dict(checkpoint_file, print_global_state=False, map_location=None): _, extension = os.path.splitext(checkpoint_file) if extension.lower() == ".safetensors": + device = map_location or shared.weight_load_location or devices.get_optimal_device_name() + if not shared.opts.disable_mmap_load_safetensors: - device = map_location or shared.weight_load_location or devices.get_optimal_device_name() pl_sd = safetensors.torch.load_file(checkpoint_file, device=device) else: pl_sd = safetensors.torch.load(open(checkpoint_file, 'rb').read()) + pl_sd = {k: v.to(device) for k, v in pl_sd.items()} else: pl_sd = torch.load(checkpoint_file, map_location=map_location or shared.weight_load_location) diff --git a/modules/shared.py b/modules/shared.py index 4743a428..203ee1b9 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -376,7 +376,7 @@ options_templates.update(options_section(('system', "System"), { "multiple_tqdm": OptionInfo(True, "Add a second progress bar to the console that shows progress for an entire job."), "print_hypernet_extra": OptionInfo(False, "Print extra hypernetwork information to console."), "list_hidden_files": OptionInfo(True, "Load models/files in hidden directories").info("directory is hidden if its name starts with \".\""), - "disable_mmap_load_safetensors": OptionInfo(False, "Disable memmapping for loading .safetensors files (fixes very slow loading speed in some cases)."), + "disable_mmap_load_safetensors": OptionInfo(False, "Disable memmapping for loading .safetensors files.").info("fixes very slow loading speed in some cases"), })) options_templates.update(options_section(('training', "Training"), { -- cgit v1.2.3 From da14f6a6632e67cacaeaac7441344f0848f66114 Mon Sep 17 00:00:00 2001 From: NoCrypt <57245077+NoCrypt@users.noreply.github.com> Date: Wed, 28 Jun 2023 10:16:44 +0700 Subject: Add options to change colors in grid --- modules/images.py | 13 +++++-------- modules/shared.py | 5 ++++- 2 files changed, 9 insertions(+), 9 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/images.py b/modules/images.py index 1906e2ab..320008be 100644 --- a/modules/images.py +++ b/modules/images.py @@ -10,7 +10,7 @@ import re import numpy as np import piexif import piexif.helper -from PIL import Image, ImageFont, ImageDraw, PngImagePlugin +from PIL import Image, ImageFont, ImageDraw, ImageColor, PngImagePlugin import string import json import hashlib @@ -156,10 +156,10 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): while drawing.multiline_textsize(line.text, font=fnt)[0] > line.allowed_width and fontsize > 0: fontsize -= 1 fnt = get_font(fontsize) - drawing.multiline_text((draw_x, draw_y + line.size[1] / 2), line.text, font=fnt, fill=color_active if line.is_active else color_inactive, anchor="mm", align="center") + drawing.multiline_text((draw_x, draw_y + line.size[1] / 2), line.text, font=fnt, fill=ImageColor.getcolor(opts.grid_text_color_active, 'RGB') if line.is_active else ImageColor.getcolor(opts.grid_text_color_inactive, 'RGB'), anchor="mm", align="center") if not line.is_active: - drawing.line((draw_x - line.size[0] // 2, draw_y + line.size[1] // 2, draw_x + line.size[0] // 2, draw_y + line.size[1] // 2), fill=color_inactive, width=4) + drawing.line((draw_x - line.size[0] // 2, draw_y + line.size[1] // 2, draw_x + line.size[0] // 2, draw_y + line.size[1] // 2), fill=ImageColor.getcolor(opts.grid_text_color_inactive, 'RGB'), width=4) draw_y += line.size[1] + line_spacing @@ -168,9 +168,6 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): fnt = get_font(fontsize) - color_active = (0, 0, 0) - color_inactive = (153, 153, 153) - pad_left = 0 if sum([sum([len(line.text) for line in lines]) for lines in ver_texts]) == 0 else width * 3 // 4 cols = im.width // width @@ -179,7 +176,7 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): assert cols == len(hor_texts), f'bad number of horizontal texts: {len(hor_texts)}; must be {cols}' assert rows == len(ver_texts), f'bad number of vertical texts: {len(ver_texts)}; must be {rows}' - calc_img = Image.new("RGB", (1, 1), "white") + calc_img = Image.new("RGB", (1, 1), ImageColor.getcolor(opts.grid_background, 'RGB')) calc_d = ImageDraw.Draw(calc_img) for texts, allowed_width in zip(hor_texts + ver_texts, [width] * len(hor_texts) + [pad_left] * len(ver_texts)): @@ -200,7 +197,7 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): pad_top = 0 if sum(hor_text_heights) == 0 else max(hor_text_heights) + line_spacing * 2 - result = Image.new("RGB", (im.width + pad_left + margin * (cols-1), im.height + pad_top + margin * (rows-1)), "white") + result = Image.new("RGB", (im.width + pad_left + margin * (cols-1), im.height + pad_top + margin * (rows-1)), ImageColor.getcolor(opts.grid_background, 'RGB')) for row in range(rows): for col in range(cols): diff --git a/modules/shared.py b/modules/shared.py index 203ee1b9..4a83cca4 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -413,6 +413,10 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "CLIP_stop_at_last_layers": OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}).link("wiki", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip").info("ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer"), "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"), "randn_source": OptionInfo("GPU", "Random number generator source.", gr.Radio, {"choices": ["GPU", "CPU"]}).info("changes seeds drastically; use CPU to produce the same picture across different videocard vendors"), + "font": OptionInfo("", "Font for image grids that have text"), + "grid_text_color_active": OptionInfo("#000000", "Text color for image grids", ui_components.FormColorPicker, {}), + "grid_text_color_inactive": OptionInfo("#999999", "Inactive text color for image grids", ui_components.FormColorPicker, {}), + "grid_background": OptionInfo("#ffffff", "Background color for image grids", ui_components.FormColorPicker, {}), })) options_templates.update(options_section(('optimizations', "Optimizations"), { @@ -471,7 +475,6 @@ options_templates.update(options_section(('ui', "User interface"), { "do_not_show_images": OptionInfo(False, "Do not show any images in results for web"), "send_seed": OptionInfo(True, "Send seed when sending prompt or image to other interface"), "send_size": OptionInfo(True, "Send size when sending prompt or image to another interface"), - "font": OptionInfo("", "Font for image grids that have text"), "js_modal_lightbox": OptionInfo(True, "Enable full page image viewer"), "js_modal_lightbox_initially_zoomed": OptionInfo(True, "Show images zoomed in by default in full page image viewer"), "js_modal_lightbox_gamepad": OptionInfo(False, "Navigate image viewer with gamepad"), -- cgit v1.2.3 From 45ab7475d61fe42b70c37541974c03736cf73189 Mon Sep 17 00:00:00 2001 From: NoCrypt <57245077+NoCrypt@users.noreply.github.com> Date: Wed, 28 Jun 2023 17:55:58 +0700 Subject: Revision --- modules/images.py | 13 +++++++++---- modules/shared.py | 6 +++--- 2 files changed, 12 insertions(+), 7 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/images.py b/modules/images.py index 320008be..913c3c2f 100644 --- a/modules/images.py +++ b/modules/images.py @@ -139,6 +139,11 @@ class GridAnnotation: def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): + + color_active = ImageColor.getcolor(opts.grid_text_active_color, 'RGB') + color_inactive = ImageColor.getcolor(opts.grid_text_inactive_color, 'RGB') + color_background = ImageColor.getcolor(opts.grid_background_color, 'RGB') + def wrap(drawing, text, font, line_length): lines = [''] for word in text.split(): @@ -156,10 +161,10 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): while drawing.multiline_textsize(line.text, font=fnt)[0] > line.allowed_width and fontsize > 0: fontsize -= 1 fnt = get_font(fontsize) - drawing.multiline_text((draw_x, draw_y + line.size[1] / 2), line.text, font=fnt, fill=ImageColor.getcolor(opts.grid_text_color_active, 'RGB') if line.is_active else ImageColor.getcolor(opts.grid_text_color_inactive, 'RGB'), anchor="mm", align="center") + drawing.multiline_text((draw_x, draw_y + line.size[1] / 2), line.text, font=fnt, fill=color_active if line.is_active else color_inactive, anchor="mm", align="center") if not line.is_active: - drawing.line((draw_x - line.size[0] // 2, draw_y + line.size[1] // 2, draw_x + line.size[0] // 2, draw_y + line.size[1] // 2), fill=ImageColor.getcolor(opts.grid_text_color_inactive, 'RGB'), width=4) + drawing.line((draw_x - line.size[0] // 2, draw_y + line.size[1] // 2, draw_x + line.size[0] // 2, draw_y + line.size[1] // 2), fill=color_inactive, width=4) draw_y += line.size[1] + line_spacing @@ -176,7 +181,7 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): assert cols == len(hor_texts), f'bad number of horizontal texts: {len(hor_texts)}; must be {cols}' assert rows == len(ver_texts), f'bad number of vertical texts: {len(ver_texts)}; must be {rows}' - calc_img = Image.new("RGB", (1, 1), ImageColor.getcolor(opts.grid_background, 'RGB')) + calc_img = Image.new("RGB", (1, 1), color_background) calc_d = ImageDraw.Draw(calc_img) for texts, allowed_width in zip(hor_texts + ver_texts, [width] * len(hor_texts) + [pad_left] * len(ver_texts)): @@ -197,7 +202,7 @@ def draw_grid_annotations(im, width, height, hor_texts, ver_texts, margin=0): pad_top = 0 if sum(hor_text_heights) == 0 else max(hor_text_heights) + line_spacing * 2 - result = Image.new("RGB", (im.width + pad_left + margin * (cols-1), im.height + pad_top + margin * (rows-1)), ImageColor.getcolor(opts.grid_background, 'RGB')) + result = Image.new("RGB", (im.width + pad_left + margin * (cols-1), im.height + pad_top + margin * (rows-1)), color_background) for row in range(rows): for col in range(cols): diff --git a/modules/shared.py b/modules/shared.py index 4a83cca4..22e6bd0b 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -414,9 +414,9 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"), "randn_source": OptionInfo("GPU", "Random number generator source.", gr.Radio, {"choices": ["GPU", "CPU"]}).info("changes seeds drastically; use CPU to produce the same picture across different videocard vendors"), "font": OptionInfo("", "Font for image grids that have text"), - "grid_text_color_active": OptionInfo("#000000", "Text color for image grids", ui_components.FormColorPicker, {}), - "grid_text_color_inactive": OptionInfo("#999999", "Inactive text color for image grids", ui_components.FormColorPicker, {}), - "grid_background": OptionInfo("#ffffff", "Background color for image grids", ui_components.FormColorPicker, {}), + "grid_text_active_color": OptionInfo("#000000", "Text color for image grids", ui_components.FormColorPicker, {}), + "grid_text_inactive_color": OptionInfo("#999999", "Inactive text color for image grids", ui_components.FormColorPicker, {}), + "grid_background_color": OptionInfo("#ffffff", "Background color for image grids", ui_components.FormColorPicker, {}), })) options_templates.update(options_section(('optimizations', "Optimizations"), { -- cgit v1.2.3 From f74fb5049506b85a98b02b1c2fd7361e9f751980 Mon Sep 17 00:00:00 2001 From: NoCrypt <57245077+NoCrypt@users.noreply.github.com> Date: Wed, 28 Jun 2023 20:24:57 +0700 Subject: Move change colors options to Saving images/grids --- modules/shared.py | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index 22e6bd0b..76d8e221 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -311,6 +311,10 @@ options_templates.update(options_section(('saving-images', "Saving images/grids" "grid_prevent_empty_spots": OptionInfo(False, "Prevent empty spots in grid (when set to autodetect)"), "grid_zip_filename_pattern": OptionInfo("", "Archive filename pattern", component_args=hide_dirs).link("wiki", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Custom-Images-Filename-Name-and-Subdirectory"), "n_rows": OptionInfo(-1, "Grid row count; use -1 for autodetect and 0 for it to be same as batch size", gr.Slider, {"minimum": -1, "maximum": 16, "step": 1}), + "font": OptionInfo("", "Font for image grids that have text"), + "grid_text_active_color": OptionInfo("#000000", "Text color for image grids", ui_components.FormColorPicker, {}), + "grid_text_inactive_color": OptionInfo("#999999", "Inactive text color for image grids", ui_components.FormColorPicker, {}), + "grid_background_color": OptionInfo("#ffffff", "Background color for image grids", ui_components.FormColorPicker, {}), "enable_pnginfo": OptionInfo(True, "Save text information about generation parameters as chunks to png files"), "save_txt": OptionInfo(False, "Create a text file next to every image with generation parameters."), @@ -413,10 +417,6 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "CLIP_stop_at_last_layers": OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}).link("wiki", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip").info("ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer"), "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"), "randn_source": OptionInfo("GPU", "Random number generator source.", gr.Radio, {"choices": ["GPU", "CPU"]}).info("changes seeds drastically; use CPU to produce the same picture across different videocard vendors"), - "font": OptionInfo("", "Font for image grids that have text"), - "grid_text_active_color": OptionInfo("#000000", "Text color for image grids", ui_components.FormColorPicker, {}), - "grid_text_inactive_color": OptionInfo("#999999", "Inactive text color for image grids", ui_components.FormColorPicker, {}), - "grid_background_color": OptionInfo("#ffffff", "Background color for image grids", ui_components.FormColorPicker, {}), })) options_templates.update(options_section(('optimizations', "Optimizations"), { -- cgit v1.2.3 From f44feb6a10aacc6a5ff4c9275fba2546b2858935 Mon Sep 17 00:00:00 2001 From: Aarni Koskela Date: Fri, 30 Jun 2023 13:11:31 +0300 Subject: Add job argument to State.begin() --- modules/api/api.py | 14 +++++++------- modules/call_queue.py | 2 +- modules/extras.py | 3 +-- modules/interrogate.py | 3 +-- modules/postprocessing.py | 3 +-- modules/shared.py | 4 ++-- 6 files changed, 13 insertions(+), 16 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/api/api.py b/modules/api/api.py index 279c384a..3ea099ad 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -327,7 +327,7 @@ class Api: p.outpath_grids = opts.outdir_txt2img_grids p.outpath_samples = opts.outdir_txt2img_samples - shared.state.begin() + shared.state.begin(job="scripts_txt2img") if selectable_scripts is not None: p.script_args = script_args processed = scripts.scripts_txt2img.run(p, *p.script_args) # Need to pass args as list here @@ -384,7 +384,7 @@ class Api: p.outpath_grids = opts.outdir_img2img_grids p.outpath_samples = opts.outdir_img2img_samples - shared.state.begin() + shared.state.begin(job="scripts_img2img") if selectable_scripts is not None: p.script_args = script_args processed = scripts.scripts_img2img.run(p, *p.script_args) # Need to pass args as list here @@ -599,7 +599,7 @@ class Api: def create_embedding(self, args: dict): try: - shared.state.begin() + shared.state.begin(job="create_embedding") filename = create_embedding(**args) # create empty embedding sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings() # reload embeddings so new one can be immediately used shared.state.end() @@ -610,7 +610,7 @@ class Api: def create_hypernetwork(self, args: dict): try: - shared.state.begin() + shared.state.begin(job="create_hypernetwork") filename = create_hypernetwork(**args) # create empty embedding shared.state.end() return models.CreateResponse(info=f"create hypernetwork filename: {filename}") @@ -620,7 +620,7 @@ class Api: def preprocess(self, args: dict): try: - shared.state.begin() + shared.state.begin(job="preprocess") preprocess(**args) # quick operation unless blip/booru interrogation is enabled shared.state.end() return models.PreprocessResponse(info = 'preprocess complete') @@ -636,7 +636,7 @@ class Api: def train_embedding(self, args: dict): try: - shared.state.begin() + shared.state.begin(job="train_embedding") apply_optimizations = shared.opts.training_xattention_optimizations error = None filename = '' @@ -657,7 +657,7 @@ class Api: def train_hypernetwork(self, args: dict): try: - shared.state.begin() + shared.state.begin(job="train_hypernetwork") shared.loaded_hypernetworks = [] apply_optimizations = shared.opts.training_xattention_optimizations error = None diff --git a/modules/call_queue.py b/modules/call_queue.py index 69bf63d2..3b94f8a4 100644 --- a/modules/call_queue.py +++ b/modules/call_queue.py @@ -30,7 +30,7 @@ def wrap_gradio_gpu_call(func, extra_outputs=None): id_task = None with queue_lock: - shared.state.begin() + shared.state.begin(job=id_task) progress.start_task(id_task) try: diff --git a/modules/extras.py b/modules/extras.py index 830b53aa..e9c0263e 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -73,8 +73,7 @@ def to_half(tensor, enable): def run_modelmerger(id_task, primary_model_name, secondary_model_name, tertiary_model_name, interp_method, multiplier, save_as_half, custom_name, checkpoint_format, config_source, bake_in_vae, discard_weights, save_metadata): - shared.state.begin() - shared.state.job = 'model-merge' + shared.state.begin(job="model-merge") def fail(message): shared.state.textinfo = message diff --git a/modules/interrogate.py b/modules/interrogate.py index 9b2c5b60..a3ae1dd5 100644 --- a/modules/interrogate.py +++ b/modules/interrogate.py @@ -184,8 +184,7 @@ class InterrogateModels: def interrogate(self, pil_image): res = "" - shared.state.begin() - shared.state.job = 'interrogate' + shared.state.begin(job="interrogate") try: if shared.cmd_opts.lowvram or shared.cmd_opts.medvram: lowvram.send_everything_to_cpu() diff --git a/modules/postprocessing.py b/modules/postprocessing.py index 736315e2..544b2f72 100644 --- a/modules/postprocessing.py +++ b/modules/postprocessing.py @@ -9,8 +9,7 @@ from modules.shared import opts def run_postprocessing(extras_mode, image, image_folder, input_dir, output_dir, show_extras_results, *args, save_output: bool = True): devices.torch_gc() - shared.state.begin() - shared.state.job = 'extras' + shared.state.begin(job="extras") image_data = [] image_names = [] diff --git a/modules/shared.py b/modules/shared.py index 203ee1b9..7df2879c 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -173,7 +173,7 @@ class State: return obj - def begin(self): + def begin(self, job: str = "(unknown)"): self.sampling_step = 0 self.job_count = -1 self.processing_has_refined_job_count = False @@ -187,7 +187,7 @@ class State: self.interrupted = False self.textinfo = None self.time_start = time.time() - + self.job = job devices.torch_gc() def end(self): -- cgit v1.2.3 From 522a8b9f629940a205812b5b023f25c051f3c8d8 Mon Sep 17 00:00:00 2001 From: Aarni Koskela Date: Fri, 30 Jun 2023 13:24:17 +0300 Subject: Add a status logger in modules.shared --- modules/shared.py | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index 7df2879c..9ab9d98b 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -4,6 +4,7 @@ import os import sys import threading import time +import logging import gradio as gr import torch @@ -18,6 +19,8 @@ from modules.paths_internal import models_path, script_path, data_path, sd_confi from ldm.models.diffusion.ddpm import LatentDiffusion from typing import Optional +log = logging.getLogger(__name__) + demo = None parser = cmd_args.parser @@ -144,12 +147,15 @@ class State: def request_restart(self) -> None: self.interrupt() self.server_command = "restart" + log.info("Received restart request") def skip(self): self.skipped = True + log.info("Received skip request") def interrupt(self): self.interrupted = True + log.info("Received interrupt request") def nextjob(self): if opts.live_previews_enable and opts.show_progress_every_n_steps == -1: @@ -189,8 +195,11 @@ class State: self.time_start = time.time() self.job = job devices.torch_gc() + log.info("Starting job %s", job) def end(self): + duration = time.time() - self.time_start + log.info("Ending job %s (%.2f seconds)", self.job, duration) self.job = "" self.job_count = 0 -- cgit v1.2.3 From 7a6abc59ea1ecd8bb311de1719b018fb5960cd80 Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Sat, 8 Jul 2023 16:15:28 +0300 Subject: for #10650: change key to alt+arrows, enable by default --- javascript/edit-order.js | 6 +++++- modules/shared.py | 2 +- 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'modules/shared.py') diff --git a/javascript/edit-order.js b/javascript/edit-order.js index 50f7fe37..ad983d33 100644 --- a/javascript/edit-order.js +++ b/javascript/edit-order.js @@ -1,8 +1,12 @@ +/* alt+left/right moves text in prompt */ + function keyupEditOrder(event) { if (!opts.keyedit_move) return; + let target = event.originalTarget || event.composedPath()[0]; if (!target.matches("*:is([id*='_toprow'] [id*='_prompt'], .prompt) textarea")) return; - if (!event.metaKey && !event.ctrlKey) return; + if (!event.altKey) return; + event.preventDefault() let isLeft = event.key == "ArrowLeft"; let isRight = event.key == "ArrowRight"; diff --git a/modules/shared.py b/modules/shared.py index b9c53875..b29c3307 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -494,7 +494,7 @@ options_templates.update(options_section(('ui', "User interface"), { "keyedit_precision_attention": OptionInfo(0.1, "Ctrl+up/down precision when editing (attention:1.1)", gr.Slider, {"minimum": 0.01, "maximum": 0.2, "step": 0.001}), "keyedit_precision_extra": OptionInfo(0.05, "Ctrl+up/down precision when editing ", gr.Slider, {"minimum": 0.01, "maximum": 0.2, "step": 0.001}), "keyedit_delimiters": OptionInfo(".,\\/!?%^*;:{}=`~()", "Ctrl+up/down word delimiters"), - "keyedit_move": OptionInfo(False, "Ctrl+left/right moves prompt elements"), + "keyedit_move": OptionInfo(True, "Alt+left/right moves prompt elements"), "quicksettings_list": OptionInfo(["sd_model_checkpoint"], "Quicksettings list", ui_components.DropdownMulti, lambda: {"choices": list(opts.data_labels.keys())}).js("info", "settingsHintsShowQuicksettings").info("setting entries that appear at the top of page rather than in settings tab").needs_restart(), "ui_tab_order": OptionInfo([], "UI tab order", ui_components.DropdownMulti, lambda: {"choices": list(tab_names)}).needs_restart(), "hidden_tabs": OptionInfo([], "Hidden UI tabs", ui_components.DropdownMulti, lambda: {"choices": list(tab_names)}).needs_restart(), -- cgit v1.2.3 From d7d6e8cfc8b85a99a48f82975ee213d487783c28 Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Sat, 8 Jul 2023 16:45:59 +0300 Subject: use natural sort for shared.walk_files and shared.listfiles, as well as for dirs in extra networks --- extensions-builtin/Lora/lora.py | 2 +- modules/shared.py | 14 +++++++++++--- modules/ui_extra_networks.py | 4 ++-- 3 files changed, 14 insertions(+), 6 deletions(-) (limited to 'modules/shared.py') diff --git a/extensions-builtin/Lora/lora.py b/extensions-builtin/Lora/lora.py index 34ff57dd..cd46e6c7 100644 --- a/extensions-builtin/Lora/lora.py +++ b/extensions-builtin/Lora/lora.py @@ -443,7 +443,7 @@ def list_available_loras(): os.makedirs(shared.cmd_opts.lora_dir, exist_ok=True) candidates = list(shared.walk_files(shared.cmd_opts.lora_dir, allowed_extensions=[".pt", ".ckpt", ".safetensors"])) - for filename in sorted(candidates, key=str.lower): + for filename in candidates: if os.path.isdir(filename): continue diff --git a/modules/shared.py b/modules/shared.py index b29c3307..48478a68 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -1,6 +1,7 @@ import datetime import json import os +import re import sys import threading import time @@ -832,8 +833,12 @@ mem_mon = modules.memmon.MemUsageMonitor("MemMon", device, opts) mem_mon.start() +def natural_sort_key(s, regex=re.compile('([0-9]+)')): + return [int(text) if text.isdigit() else text.lower() for text in regex.split(s)] + + def listfiles(dirname): - filenames = [os.path.join(dirname, x) for x in sorted(os.listdir(dirname), key=str.lower) if not x.startswith(".")] + filenames = [os.path.join(dirname, x) for x in sorted(os.listdir(dirname), key=natural_sort_key) if not x.startswith(".")] return [file for file in filenames if os.path.isfile(file)] @@ -858,8 +863,11 @@ def walk_files(path, allowed_extensions=None): if allowed_extensions is not None: allowed_extensions = set(allowed_extensions) - for root, _, files in os.walk(path, followlinks=True): - for filename in files: + items = list(os.walk(path, followlinks=True)) + items = sorted(items, key=lambda x: natural_sort_key(x[0])) + + for root, _, files in items: + for filename in sorted(files, key=natural_sort_key): if allowed_extensions is not None: _, ext = os.path.splitext(filename) if ext not in allowed_extensions: diff --git a/modules/ui_extra_networks.py b/modules/ui_extra_networks.py index 1efd00b0..693cafb6 100644 --- a/modules/ui_extra_networks.py +++ b/modules/ui_extra_networks.py @@ -90,8 +90,8 @@ class ExtraNetworksPage: subdirs = {} for parentdir in [os.path.abspath(x) for x in self.allowed_directories_for_previews()]: - for root, dirs, _ in os.walk(parentdir, followlinks=True): - for dirname in dirs: + for root, dirs, _ in sorted(os.walk(parentdir, followlinks=True), key=lambda x: shared.natural_sort_key(x[0])): + for dirname in sorted(dirs, key=shared.natural_sort_key): x = os.path.join(root, dirname) if not os.path.isdir(x): -- cgit v1.2.3 From 4981c7d3704e50dd93fe1b68d299239a4ded1ec2 Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Sat, 8 Jul 2023 17:52:03 +0300 Subject: move github proxy to settings, System page. --- modules/shared.py | 1 + modules/ui_extensions.py | 33 ++++++++++++++------------------- 2 files changed, 15 insertions(+), 19 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index 48478a68..b7518de6 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -391,6 +391,7 @@ options_templates.update(options_section(('system', "System"), { "print_hypernet_extra": OptionInfo(False, "Print extra hypernetwork information to console."), "list_hidden_files": OptionInfo(True, "Load models/files in hidden directories").info("directory is hidden if its name starts with \".\""), "disable_mmap_load_safetensors": OptionInfo(False, "Disable memmapping for loading .safetensors files.").info("fixes very slow loading speed in some cases"), + "github_proxy": OptionInfo("None", "Github proxy", ui_components.DropdownEditable, lambda: {"choices": ["None", "ghproxy.com", "hub.yzuu.cf", "hub.njuu.cf", "hub.nuaa.cf"]}).info("for custom inputs will just replace github.com with the input"), })) options_templates.update(options_section(('training', "Training"), { diff --git a/modules/ui_extensions.py b/modules/ui_extensions.py index ac523bcf..a208012d 100644 --- a/modules/ui_extensions.py +++ b/modules/ui_extensions.py @@ -325,7 +325,18 @@ def normalize_git_url(url): return url -def install_extension_from_url(dirname, proxy, url, branch_name=None): +def github_proxy(url): + proxy = shared.opts.github_proxy + + if proxy == 'None': + return url + if proxy == 'ghproxy.com': + return "https://ghproxy.com/" + url + + return url.replace('github.com', proxy) + + +def install_extension_from_url(dirname, url, branch_name=None): check_access() if isinstance(dirname, str): @@ -335,18 +346,7 @@ def install_extension_from_url(dirname, proxy, url, branch_name=None): assert url, 'No URL specified' - proxy_list = { - "none": "", - "ghproxy": "https://ghproxy.com/", - "yzuu": "hub.yzuu.cf", - "njuu": "hub.njuu.cf", - "nuaa": "hub.nuaa.cf", - } - - if proxy in ['yzuu', 'njuu', 'nuaa']: - url = url.replace('github.com', proxy_list[proxy]) - elif proxy == 'ghproxy': - url = proxy_list[proxy] + url + url = github_proxy(url) if dirname is None or dirname == "": *parts, last_part = url.split('/') @@ -628,11 +628,6 @@ def create_ui(): ) with gr.TabItem("Install from URL", id="install_from_url"): - install_proxy = gr.Radio( - label="Install Proxy", choices=["none", "ghproxy", "nuaa", "yzuu", "njuu"], value="none", - info="If you can't access github.com, you can use a proxy to install extensions from github.com" - ) - install_url = gr.Text(label="URL for extension's git repository") install_branch = gr.Text(label="Specific branch name", placeholder="Leave empty for default main branch") install_dirname = gr.Text(label="Local directory name", placeholder="Leave empty for auto") @@ -641,7 +636,7 @@ def create_ui(): install_button.click( fn=modules.ui.wrap_gradio_call(lambda *args: [gr.update(), *install_extension_from_url(*args)], extra_outputs=[gr.update(), gr.update()]), - inputs=[install_dirname, install_proxy, install_url, install_branch], + inputs=[install_dirname, install_url, install_branch], outputs=[install_url, extensions_table, install_result], ) -- cgit v1.2.3 From 75f56406cede0095eb0e5dcc4e0d5759063e89dc Mon Sep 17 00:00:00 2001 From: wfjsw Date: Sun, 9 Jul 2023 22:40:23 +0800 Subject: Revert Pull Request #11244 Revert "Add github mirror for the download extension" This reverts commit 9ec2ba2d28bb0d8f01e19e2919b7bf2e3e864773. Revert "Update code style" This reverts commit de022c4c80240a430a8099fb27a41aa505bf5b2f. Revert "Update call method" This reverts commit e9bd18c57bd83363d38c7409263fe87f3ed3a7f0. Revert "move github proxy to settings, System page." This reverts commit 4981c7d3704e50dd93fe1b68d299239a4ded1ec2. --- modules/shared.py | 1 - modules/ui_extensions.py | 17 ++--------------- 2 files changed, 2 insertions(+), 16 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index b7518de6..48478a68 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -391,7 +391,6 @@ options_templates.update(options_section(('system', "System"), { "print_hypernet_extra": OptionInfo(False, "Print extra hypernetwork information to console."), "list_hidden_files": OptionInfo(True, "Load models/files in hidden directories").info("directory is hidden if its name starts with \".\""), "disable_mmap_load_safetensors": OptionInfo(False, "Disable memmapping for loading .safetensors files.").info("fixes very slow loading speed in some cases"), - "github_proxy": OptionInfo("None", "Github proxy", ui_components.DropdownEditable, lambda: {"choices": ["None", "ghproxy.com", "hub.yzuu.cf", "hub.njuu.cf", "hub.nuaa.cf"]}).info("for custom inputs will just replace github.com with the input"), })) options_templates.update(options_section(('training', "Training"), { diff --git a/modules/ui_extensions.py b/modules/ui_extensions.py index a208012d..dff522ef 100644 --- a/modules/ui_extensions.py +++ b/modules/ui_extensions.py @@ -325,17 +325,6 @@ def normalize_git_url(url): return url -def github_proxy(url): - proxy = shared.opts.github_proxy - - if proxy == 'None': - return url - if proxy == 'ghproxy.com': - return "https://ghproxy.com/" + url - - return url.replace('github.com', proxy) - - def install_extension_from_url(dirname, url, branch_name=None): check_access() @@ -346,8 +335,6 @@ def install_extension_from_url(dirname, url, branch_name=None): assert url, 'No URL specified' - url = github_proxy(url) - if dirname is None or dirname == "": *parts, last_part = url.split('/') last_part = normalize_git_url(last_part) @@ -367,12 +354,12 @@ def install_extension_from_url(dirname, url, branch_name=None): shutil.rmtree(tmpdir, True) if not branch_name: # if no branch is specified, use the default branch - with git.Repo.clone_from(url, tmpdir, filter=['blob:none'], verbose=False) as repo: + with git.Repo.clone_from(url, tmpdir, filter=['blob:none']) as repo: repo.remote().fetch() for submodule in repo.submodules: submodule.update() else: - with git.Repo.clone_from(url, tmpdir, filter=['blob:none'], branch=branch_name, verbose=False) as repo: + with git.Repo.clone_from(url, tmpdir, filter=['blob:none'], branch=branch_name) as repo: repo.remote().fetch() for submodule in repo.submodules: submodule.update() -- cgit v1.2.3 From da464a3fb39ecc6ea7b22fe87271194480d8501c Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Wed, 12 Jul 2023 23:52:43 +0300 Subject: SDXL support --- modules/launch_utils.py | 17 +++++++++++++ modules/lowvram.py | 51 +++++++++++++++++++++++++++----------- modules/paths.py | 9 ++++++- modules/processing.py | 7 ++++-- modules/prompt_parser.py | 23 ++++++++++++++--- modules/sd_hijack.py | 23 ++++++++++++++++- modules/sd_hijack_clip.py | 16 +++++++++--- modules/sd_hijack_open_clip.py | 38 +++++++++++++++++++++++++--- modules/sd_hijack_optimizations.py | 51 ++++++++++++++++++++++++++++++++------ modules/sd_models.py | 14 +++++++++-- modules/sd_models_config.py | 5 +++- modules/sd_models_xl.py | 27 +++++++++++++++++--- modules/sd_samplers_kdiffusion.py | 2 +- modules/shared.py | 2 ++ requirements.txt | 1 + requirements_versions.txt | 1 + 16 files changed, 242 insertions(+), 45 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/launch_utils.py b/modules/launch_utils.py index 3b740dbd..aa9d1880 100644 --- a/modules/launch_utils.py +++ b/modules/launch_utils.py @@ -224,6 +224,20 @@ def run_extensions_installers(settings_file): run_extension_installer(os.path.join(extensions_dir, dirname_extension)) +def mute_sdxl_imports(): + """create fake modules that SDXL wants to import but doesn't actually use for our purposes""" + + import importlib + + module = importlib.util.module_from_spec(importlib.machinery.ModuleSpec('taming.modules.losses.lpips', None)) + module.LPIPS = None + sys.modules['taming.modules.losses.lpips'] = module + + module = importlib.util.module_from_spec(importlib.machinery.ModuleSpec('sgm.data', None)) + module.StableDataModuleFromConfig = None + sys.modules['sgm.data'] = module + + def prepare_environment(): torch_index_url = os.environ.get('TORCH_INDEX_URL', "https://download.pytorch.org/whl/cu118") torch_command = os.environ.get('TORCH_COMMAND', f"pip install torch==2.0.1 torchvision==0.15.2 --extra-index-url {torch_index_url}") @@ -319,11 +333,14 @@ def prepare_environment(): if args.update_all_extensions: git_pull_recursive(extensions_dir) + mute_sdxl_imports() + if "--exit" in sys.argv: print("Exiting because of --exit argument") exit(0) + def configure_for_tests(): if "--api" not in sys.argv: sys.argv.append("--api") diff --git a/modules/lowvram.py b/modules/lowvram.py index d95bcfbf..da4f33a8 100644 --- a/modules/lowvram.py +++ b/modules/lowvram.py @@ -53,19 +53,46 @@ def setup_for_low_vram(sd_model, use_medvram): send_me_to_gpu(first_stage_model, None) return first_stage_model_decode(z) - # for SD1, cond_stage_model is CLIP and its NN is in the tranformer frield, but for SD2, it's open clip, and it's in model field - if hasattr(sd_model.cond_stage_model, 'model'): - sd_model.cond_stage_model.transformer = sd_model.cond_stage_model.model - - # remove several big modules: cond, first_stage, depth/embedder (if applicable), and unet from the model and then - # send the model to GPU. Then put modules back. the modules will be in CPU. - stored = sd_model.cond_stage_model.transformer, sd_model.first_stage_model, getattr(sd_model, 'depth_model', None), getattr(sd_model, 'embedder', None), sd_model.model - sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.depth_model, sd_model.embedder, sd_model.model = None, None, None, None, None + to_remain_in_cpu = [ + (sd_model, 'first_stage_model'), + (sd_model, 'depth_model'), + (sd_model, 'embedder'), + (sd_model, 'model'), + (sd_model, 'embedder'), + ] + + is_sdxl = hasattr(sd_model, 'conditioner') + is_sd2 = not is_sdxl and hasattr(sd_model.cond_stage_model, 'model') + + if is_sdxl: + to_remain_in_cpu.append((sd_model, 'conditioner')) + elif is_sd2: + to_remain_in_cpu.append((sd_model.cond_stage_model, 'model')) + else: + to_remain_in_cpu.append((sd_model.cond_stage_model, 'transformer')) + + # remove several big modules: cond, first_stage, depth/embedder (if applicable), and unet from the model + stored = [] + for obj, field in to_remain_in_cpu: + module = getattr(obj, field, None) + stored.append(module) + setattr(obj, field, None) + + # send the model to GPU. sd_model.to(devices.device) - sd_model.cond_stage_model.transformer, sd_model.first_stage_model, sd_model.depth_model, sd_model.embedder, sd_model.model = stored + + # put modules back. the modules will be in CPU. + for (obj, field), module in zip(to_remain_in_cpu, stored): + setattr(obj, field, module) # register hooks for those the first three models - sd_model.cond_stage_model.transformer.register_forward_pre_hook(send_me_to_gpu) + if is_sdxl: + sd_model.conditioner.register_forward_pre_hook(send_me_to_gpu) + elif is_sd2: + sd_model.cond_stage_model.model.register_forward_pre_hook(send_me_to_gpu) + else: + sd_model.cond_stage_model.transformer.register_forward_pre_hook(send_me_to_gpu) + sd_model.first_stage_model.register_forward_pre_hook(send_me_to_gpu) sd_model.first_stage_model.encode = first_stage_model_encode_wrap sd_model.first_stage_model.decode = first_stage_model_decode_wrap @@ -75,10 +102,6 @@ def setup_for_low_vram(sd_model, use_medvram): sd_model.embedder.register_forward_pre_hook(send_me_to_gpu) parents[sd_model.cond_stage_model.transformer] = sd_model.cond_stage_model - if hasattr(sd_model.cond_stage_model, 'model'): - sd_model.cond_stage_model.model = sd_model.cond_stage_model.transformer - del sd_model.cond_stage_model.transformer - if use_medvram: sd_model.model.register_forward_pre_hook(send_me_to_gpu) else: diff --git a/modules/paths.py b/modules/paths.py index f509a85f..1100a8dc 100644 --- a/modules/paths.py +++ b/modules/paths.py @@ -20,7 +20,7 @@ assert sd_path is not None, f"Couldn't find Stable Diffusion in any of: {possibl path_dirs = [ (sd_path, 'ldm', 'Stable Diffusion', []), - (os.path.join(sd_path, '../generative-models'), 'sgm', 'Stable Diffusion XL', []), + (os.path.join(sd_path, '../generative-models'), 'sgm', 'Stable Diffusion XL', ["sgm"]), (os.path.join(sd_path, '../CodeFormer'), 'inference_codeformer.py', 'CodeFormer', []), (os.path.join(sd_path, '../BLIP'), 'models/blip.py', 'BLIP', []), (os.path.join(sd_path, '../k-diffusion'), 'k_diffusion/sampling.py', 'k_diffusion', ["atstart"]), @@ -36,6 +36,13 @@ for d, must_exist, what, options in path_dirs: d = os.path.abspath(d) if "atstart" in options: sys.path.insert(0, d) + elif "sgm" in options: + # Stable Diffusion XL repo has scripts dir with __init__.py in it which ruins every extension's scripts dir, so we + # import sgm and remove it from sys.path so that when a script imports scripts.something, it doesbn't use sgm's scripts dir. + + sys.path.insert(0, d) + import sgm + sys.path.pop(0) else: sys.path.append(d) paths[what] = d diff --git a/modules/processing.py b/modules/processing.py index cd568a20..85d35423 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -343,10 +343,13 @@ class StableDiffusionProcessing: return cache[1] def setup_conds(self): + prompts = prompt_parser.SdConditioning(self.prompts, width=self.width, height=self.height) + negative_prompts = prompt_parser.SdConditioning(self.negative_prompts, width=self.width, height=self.height) + sampler_config = sd_samplers.find_sampler_config(self.sampler_name) self.step_multiplier = 2 if sampler_config and sampler_config.options.get("second_order", False) else 1 - self.uc = self.get_conds_with_caching(prompt_parser.get_learned_conditioning, self.negative_prompts, self.steps * self.step_multiplier, [self.cached_uc], self.extra_network_data) - self.c = self.get_conds_with_caching(prompt_parser.get_multicond_learned_conditioning, self.prompts, self.steps * self.step_multiplier, [self.cached_c], self.extra_network_data) + self.uc = self.get_conds_with_caching(prompt_parser.get_learned_conditioning, negative_prompts, self.steps * self.step_multiplier, [self.cached_uc], self.extra_network_data) + self.c = self.get_conds_with_caching(prompt_parser.get_multicond_learned_conditioning, prompts, self.steps * self.step_multiplier, [self.cached_c], self.extra_network_data) def parse_extra_network_prompts(self): self.prompts, self.extra_network_data = extra_networks.parse_prompts(self.prompts) diff --git a/modules/prompt_parser.py b/modules/prompt_parser.py index d7f9e9a9..33810669 100644 --- a/modules/prompt_parser.py +++ b/modules/prompt_parser.py @@ -1,3 +1,5 @@ +from __future__ import annotations + import re from collections import namedtuple from typing import List @@ -109,7 +111,19 @@ def get_learned_conditioning_prompt_schedules(prompts, steps): ScheduledPromptConditioning = namedtuple("ScheduledPromptConditioning", ["end_at_step", "cond"]) -def get_learned_conditioning(model, prompts, steps): +class SdConditioning(list): + """ + A list with prompts for stable diffusion's conditioner model. + Can also specify width and height of created image - SDXL needs it. + """ + def __init__(self, prompts, width=None, height=None): + super().__init__() + self.extend(prompts) + self.width = width or getattr(prompts, 'width', None) + self.height = height or getattr(prompts, 'height', None) + + +def get_learned_conditioning(model, prompts: SdConditioning | list[str], steps): """converts a list of prompts into a list of prompt schedules - each schedule is a list of ScheduledPromptConditioning, specifying the comdition (cond), and the sampling step at which this condition is to be replaced by the next one. @@ -160,11 +174,13 @@ def get_learned_conditioning(model, prompts, steps): re_AND = re.compile(r"\bAND\b") re_weight = re.compile(r"^(.*?)(?:\s*:\s*([-+]?(?:\d+\.?|\d*\.\d+)))?\s*$") -def get_multicond_prompt_list(prompts): + +def get_multicond_prompt_list(prompts: SdConditioning | list[str]): res_indexes = [] - prompt_flat_list = [] prompt_indexes = {} + prompt_flat_list = SdConditioning(prompts) + prompt_flat_list.clear() for prompt in prompts: subprompts = re_AND.split(prompt) @@ -201,6 +217,7 @@ class MulticondLearnedConditioning: self.shape: tuple = shape # the shape field is needed to send this object to DDIM/PLMS self.batch: List[List[ComposableScheduledPromptConditioning]] = batch + def get_multicond_learned_conditioning(model, prompts, steps) -> MulticondLearnedConditioning: """same as get_learned_conditioning, but returns a list of ScheduledPromptConditioning along with the weight objects for each prompt. For each prompt, the list is obtained by splitting the prompt using the AND separator. diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index c4b9211f..266811f9 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -15,6 +15,11 @@ import ldm.models.diffusion.ddim import ldm.models.diffusion.plms import ldm.modules.encoders.modules +import sgm.modules.attention +import sgm.modules.diffusionmodules.model +import sgm.modules.diffusionmodules.openaimodel +import sgm.modules.encoders.modules + attention_CrossAttention_forward = ldm.modules.attention.CrossAttention.forward diffusionmodules_model_nonlinearity = ldm.modules.diffusionmodules.model.nonlinearity diffusionmodules_model_AttnBlock_forward = ldm.modules.diffusionmodules.model.AttnBlock.forward @@ -56,6 +61,9 @@ def apply_optimizations(option=None): ldm.modules.diffusionmodules.model.nonlinearity = silu ldm.modules.diffusionmodules.openaimodel.th = sd_hijack_unet.th + sgm.modules.diffusionmodules.model.nonlinearity = silu + sgm.modules.diffusionmodules.openaimodel.th = sd_hijack_unet.th + if current_optimizer is not None: current_optimizer.undo() current_optimizer = None @@ -89,6 +97,10 @@ def undo_optimizations(): ldm.modules.attention.CrossAttention.forward = hypernetwork.attention_CrossAttention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = diffusionmodules_model_AttnBlock_forward + sgm.modules.diffusionmodules.model.nonlinearity = diffusionmodules_model_nonlinearity + sgm.modules.attention.CrossAttention.forward = hypernetwork.attention_CrossAttention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = diffusionmodules_model_AttnBlock_forward + def fix_checkpoint(): """checkpoints are now added and removed in embedding/hypernet code, since torch doesn't want @@ -170,10 +182,19 @@ class StableDiffusionModelHijack: if conditioner: for i in range(len(conditioner.embedders)): embedder = conditioner.embedders[i] - if type(embedder).__name__ == 'FrozenOpenCLIPEmbedder': + typename = type(embedder).__name__ + if typename == 'FrozenOpenCLIPEmbedder': embedder.model.token_embedding = EmbeddingsWithFixes(embedder.model.token_embedding, self) m.cond_stage_model = sd_hijack_open_clip.FrozenOpenCLIPEmbedderWithCustomWords(embedder, self) conditioner.embedders[i] = m.cond_stage_model + if typename == 'FrozenCLIPEmbedder': + model_embeddings = m.cond_stage_model.transformer.text_model.embeddings + model_embeddings.token_embedding = EmbeddingsWithFixes(model_embeddings.token_embedding, self) + m.cond_stage_model = sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords(embedder, self) + conditioner.embedders[i] = m.cond_stage_model + if typename == 'FrozenOpenCLIPEmbedder2': + embedder.model.token_embedding = EmbeddingsWithFixes(embedder.model.token_embedding, self) + conditioner.embedders[i] = sd_hijack_open_clip.FrozenOpenCLIPEmbedder2WithCustomWords(embedder, self) if type(m.cond_stage_model) == xlmr.BertSeriesModelWithTransformation: model_embeddings = m.cond_stage_model.roberta.embeddings diff --git a/modules/sd_hijack_clip.py b/modules/sd_hijack_clip.py index 3b5a7666..6c17a81d 100644 --- a/modules/sd_hijack_clip.py +++ b/modules/sd_hijack_clip.py @@ -42,6 +42,10 @@ class FrozenCLIPEmbedderWithCustomWordsBase(torch.nn.Module): self.hijack: sd_hijack.StableDiffusionModelHijack = hijack self.chunk_length = 75 + self.is_trainable = getattr(wrapped, 'is_trainable', False) + self.input_key = getattr(wrapped, 'input_key', 'txt') + self.legacy_ucg_val = None + def empty_chunk(self): """creates an empty PromptChunk and returns it""" @@ -199,8 +203,9 @@ class FrozenCLIPEmbedderWithCustomWordsBase(torch.nn.Module): """ Accepts an array of texts; Passes texts through transformers network to create a tensor with numerical representation of those texts. Returns a tensor with shape of (B, T, C), where B is length of the array; T is length, in tokens, of texts (including padding) - T will - be a multiple of 77; and C is dimensionality of each token - for SD1 it's 768, and for SD2 it's 1024. + be a multiple of 77; and C is dimensionality of each token - for SD1 it's 768, for SD2 it's 1024, and for SDXL it's 1280. An example shape returned by this function can be: (2, 77, 768). + For SDXL, instead of returning one tensor avobe, it returns a tuple with two: the other one with shape (B, 1280) with pooled values. Webui usually sends just one text at a time through this function - the only time when texts is an array with more than one elemenet is when you do prompt editing: "a picture of a [cat:dog:0.4] eating ice cream" """ @@ -233,7 +238,10 @@ class FrozenCLIPEmbedderWithCustomWordsBase(torch.nn.Module): embeddings_list = ", ".join([f'{name} [{embedding.checksum()}]' for name, embedding in used_embeddings.items()]) self.hijack.comments.append(f"Used embeddings: {embeddings_list}") - return torch.hstack(zs) + if getattr(self.wrapped, 'return_pooled', False): + return torch.hstack(zs), zs[0].pooled + else: + return torch.hstack(zs) def process_tokens(self, remade_batch_tokens, batch_multipliers): """ @@ -256,9 +264,9 @@ class FrozenCLIPEmbedderWithCustomWordsBase(torch.nn.Module): # restoring original mean is likely not correct, but it seems to work well to prevent artifacts that happen otherwise batch_multipliers = torch.asarray(batch_multipliers).to(devices.device) original_mean = z.mean() - z = z * batch_multipliers.reshape(batch_multipliers.shape + (1,)).expand(z.shape) + z *= batch_multipliers.reshape(batch_multipliers.shape + (1,)).expand(z.shape) new_mean = z.mean() - z = z * (original_mean / new_mean) + z *= (original_mean / new_mean) return z diff --git a/modules/sd_hijack_open_clip.py b/modules/sd_hijack_open_clip.py index 6ac5bda6..fcf5ad07 100644 --- a/modules/sd_hijack_open_clip.py +++ b/modules/sd_hijack_open_clip.py @@ -16,10 +16,6 @@ class FrozenOpenCLIPEmbedderWithCustomWords(sd_hijack_clip.FrozenCLIPEmbedderWit self.id_end = tokenizer.encoder[""] self.id_pad = 0 - self.is_trainable = getattr(wrapped, 'is_trainable', False) - self.input_key = getattr(wrapped, 'input_key', 'txt') - self.legacy_ucg_val = None - def tokenize(self, texts): assert not opts.use_old_emphasis_implementation, 'Old emphasis implementation not supported for Open Clip' @@ -39,3 +35,37 @@ class FrozenOpenCLIPEmbedderWithCustomWords(sd_hijack_clip.FrozenCLIPEmbedderWit embedded = self.wrapped.model.token_embedding.wrapped(ids).squeeze(0) return embedded + + +class FrozenOpenCLIPEmbedder2WithCustomWords(sd_hijack_clip.FrozenCLIPEmbedderWithCustomWordsBase): + def __init__(self, wrapped, hijack): + super().__init__(wrapped, hijack) + + self.comma_token = [v for k, v in tokenizer.encoder.items() if k == ','][0] + self.id_start = tokenizer.encoder[""] + self.id_end = tokenizer.encoder[""] + self.id_pad = 0 + + def tokenize(self, texts): + assert not opts.use_old_emphasis_implementation, 'Old emphasis implementation not supported for Open Clip' + + tokenized = [tokenizer.encode(text) for text in texts] + + return tokenized + + def encode_with_transformers(self, tokens): + d = self.wrapped.encode_with_transformer(tokens) + z = d[self.wrapped.layer] + + pooled = d.get("pooled") + if pooled is not None: + z.pooled = pooled + + return z + + def encode_embedding_init_text(self, init_text, nvpt): + ids = tokenizer.encode(init_text) + ids = torch.asarray([ids], device=devices.device, dtype=torch.int) + embedded = self.wrapped.model.token_embedding.wrapped(ids).squeeze(0) + + return embedded diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 53e27ade..e99c9ba5 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -14,7 +14,11 @@ from modules.hypernetworks import hypernetwork import ldm.modules.attention import ldm.modules.diffusionmodules.model +import sgm.modules.attention +import sgm.modules.diffusionmodules.model + diffusionmodules_model_AttnBlock_forward = ldm.modules.diffusionmodules.model.AttnBlock.forward +sgm_diffusionmodules_model_AttnBlock_forward = sgm.modules.diffusionmodules.model.AttnBlock.forward class SdOptimization: @@ -39,6 +43,9 @@ class SdOptimization: ldm.modules.attention.CrossAttention.forward = hypernetwork.attention_CrossAttention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = diffusionmodules_model_AttnBlock_forward + sgm.modules.attention.CrossAttention.forward = hypernetwork.attention_CrossAttention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = sgm_diffusionmodules_model_AttnBlock_forward + class SdOptimizationXformers(SdOptimization): name = "xformers" @@ -51,6 +58,8 @@ class SdOptimizationXformers(SdOptimization): def apply(self): ldm.modules.attention.CrossAttention.forward = xformers_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = xformers_attnblock_forward + sgm.modules.attention.CrossAttention.forward = xformers_attention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = xformers_attnblock_forward class SdOptimizationSdpNoMem(SdOptimization): @@ -65,6 +74,8 @@ class SdOptimizationSdpNoMem(SdOptimization): def apply(self): ldm.modules.attention.CrossAttention.forward = scaled_dot_product_no_mem_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = sdp_no_mem_attnblock_forward + sgm.modules.attention.CrossAttention.forward = scaled_dot_product_no_mem_attention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = sdp_no_mem_attnblock_forward class SdOptimizationSdp(SdOptimizationSdpNoMem): @@ -76,6 +87,8 @@ class SdOptimizationSdp(SdOptimizationSdpNoMem): def apply(self): ldm.modules.attention.CrossAttention.forward = scaled_dot_product_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = sdp_attnblock_forward + sgm.modules.attention.CrossAttention.forward = scaled_dot_product_attention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = sdp_attnblock_forward class SdOptimizationSubQuad(SdOptimization): @@ -86,6 +99,8 @@ class SdOptimizationSubQuad(SdOptimization): def apply(self): ldm.modules.attention.CrossAttention.forward = sub_quad_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = sub_quad_attnblock_forward + sgm.modules.attention.CrossAttention.forward = sub_quad_attention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = sub_quad_attnblock_forward class SdOptimizationV1(SdOptimization): @@ -94,9 +109,9 @@ class SdOptimizationV1(SdOptimization): cmd_opt = "opt_split_attention_v1" priority = 10 - def apply(self): ldm.modules.attention.CrossAttention.forward = split_cross_attention_forward_v1 + sgm.modules.attention.CrossAttention.forward = split_cross_attention_forward_v1 class SdOptimizationInvokeAI(SdOptimization): @@ -109,6 +124,7 @@ class SdOptimizationInvokeAI(SdOptimization): def apply(self): ldm.modules.attention.CrossAttention.forward = split_cross_attention_forward_invokeAI + sgm.modules.attention.CrossAttention.forward = split_cross_attention_forward_invokeAI class SdOptimizationDoggettx(SdOptimization): @@ -119,6 +135,8 @@ class SdOptimizationDoggettx(SdOptimization): def apply(self): ldm.modules.attention.CrossAttention.forward = split_cross_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = cross_attention_attnblock_forward + sgm.modules.attention.CrossAttention.forward = split_cross_attention_forward + sgm.modules.diffusionmodules.model.AttnBlock.forward = cross_attention_attnblock_forward def list_optimizers(res): @@ -155,7 +173,7 @@ def get_available_vram(): # see https://github.com/basujindal/stable-diffusion/pull/117 for discussion -def split_cross_attention_forward_v1(self, x, context=None, mask=None): +def split_cross_attention_forward_v1(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): h = self.heads q_in = self.to_q(x) @@ -196,7 +214,7 @@ def split_cross_attention_forward_v1(self, x, context=None, mask=None): # taken from https://github.com/Doggettx/stable-diffusion and modified -def split_cross_attention_forward(self, x, context=None, mask=None): +def split_cross_attention_forward(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): h = self.heads q_in = self.to_q(x) @@ -262,11 +280,13 @@ def split_cross_attention_forward(self, x, context=None, mask=None): # -- Taken from https://github.com/invoke-ai/InvokeAI and modified -- mem_total_gb = psutil.virtual_memory().total // (1 << 30) + def einsum_op_compvis(q, k, v): s = einsum('b i d, b j d -> b i j', q, k) s = s.softmax(dim=-1, dtype=s.dtype) return einsum('b i j, b j d -> b i d', s, v) + def einsum_op_slice_0(q, k, v, slice_size): r = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype) for i in range(0, q.shape[0], slice_size): @@ -274,6 +294,7 @@ def einsum_op_slice_0(q, k, v, slice_size): r[i:end] = einsum_op_compvis(q[i:end], k[i:end], v[i:end]) return r + def einsum_op_slice_1(q, k, v, slice_size): r = torch.zeros(q.shape[0], q.shape[1], v.shape[2], device=q.device, dtype=q.dtype) for i in range(0, q.shape[1], slice_size): @@ -281,6 +302,7 @@ def einsum_op_slice_1(q, k, v, slice_size): r[:, i:end] = einsum_op_compvis(q[:, i:end], k, v) return r + def einsum_op_mps_v1(q, k, v): if q.shape[0] * q.shape[1] <= 2**16: # (512x512) max q.shape[1]: 4096 return einsum_op_compvis(q, k, v) @@ -290,12 +312,14 @@ def einsum_op_mps_v1(q, k, v): slice_size -= 1 return einsum_op_slice_1(q, k, v, slice_size) + def einsum_op_mps_v2(q, k, v): if mem_total_gb > 8 and q.shape[0] * q.shape[1] <= 2**16: return einsum_op_compvis(q, k, v) else: return einsum_op_slice_0(q, k, v, 1) + def einsum_op_tensor_mem(q, k, v, max_tensor_mb): size_mb = q.shape[0] * q.shape[1] * k.shape[1] * q.element_size() // (1 << 20) if size_mb <= max_tensor_mb: @@ -305,6 +329,7 @@ def einsum_op_tensor_mem(q, k, v, max_tensor_mb): return einsum_op_slice_0(q, k, v, q.shape[0] // div) return einsum_op_slice_1(q, k, v, max(q.shape[1] // div, 1)) + def einsum_op_cuda(q, k, v): stats = torch.cuda.memory_stats(q.device) mem_active = stats['active_bytes.all.current'] @@ -315,6 +340,7 @@ def einsum_op_cuda(q, k, v): # Divide factor of safety as there's copying and fragmentation return einsum_op_tensor_mem(q, k, v, mem_free_total / 3.3 / (1 << 20)) + def einsum_op(q, k, v): if q.device.type == 'cuda': return einsum_op_cuda(q, k, v) @@ -328,7 +354,8 @@ def einsum_op(q, k, v): # Tested on i7 with 8MB L3 cache. return einsum_op_tensor_mem(q, k, v, 32) -def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None): + +def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): h = self.heads q = self.to_q(x) @@ -356,7 +383,7 @@ def split_cross_attention_forward_invokeAI(self, x, context=None, mask=None): # Based on Birch-san's modified implementation of sub-quadratic attention from https://github.com/Birch-san/diffusers/pull/1 # The sub_quad_attention_forward function is under the MIT License listed under Memory Efficient Attention in the Licenses section of the web UI interface -def sub_quad_attention_forward(self, x, context=None, mask=None): +def sub_quad_attention_forward(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): assert mask is None, "attention-mask not currently implemented for SubQuadraticCrossAttnProcessor." h = self.heads @@ -392,6 +419,7 @@ def sub_quad_attention_forward(self, x, context=None, mask=None): return x + def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_size_min=None, chunk_threshold=None, use_checkpoint=True): bytes_per_token = torch.finfo(q.dtype).bits//8 batch_x_heads, q_tokens, _ = q.shape @@ -442,7 +470,7 @@ def get_xformers_flash_attention_op(q, k, v): return None -def xformers_attention_forward(self, x, context=None, mask=None): +def xformers_attention_forward(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): h = self.heads q_in = self.to_q(x) context = default(context, x) @@ -465,9 +493,10 @@ def xformers_attention_forward(self, x, context=None, mask=None): out = rearrange(out, 'b n h d -> b n (h d)', h=h) return self.to_out(out) + # Based on Diffusers usage of scaled dot product attention from https://github.com/huggingface/diffusers/blob/c7da8fd23359a22d0df2741688b5b4f33c26df21/src/diffusers/models/cross_attention.py # The scaled_dot_product_attention_forward function contains parts of code under Apache-2.0 license listed under Scaled Dot Product Attention in the Licenses section of the web UI interface -def scaled_dot_product_attention_forward(self, x, context=None, mask=None): +def scaled_dot_product_attention_forward(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): batch_size, sequence_length, inner_dim = x.shape if mask is not None: @@ -507,10 +536,12 @@ def scaled_dot_product_attention_forward(self, x, context=None, mask=None): hidden_states = self.to_out[1](hidden_states) return hidden_states -def scaled_dot_product_no_mem_attention_forward(self, x, context=None, mask=None): + +def scaled_dot_product_no_mem_attention_forward(self, x, context=None, mask=None, additional_tokens=None, n_times_crossframe_attn_in_self=0): with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=True, enable_mem_efficient=False): return scaled_dot_product_attention_forward(self, x, context, mask) + def cross_attention_attnblock_forward(self, x): h_ = x h_ = self.norm(h_) @@ -569,6 +600,7 @@ def cross_attention_attnblock_forward(self, x): return h3 + def xformers_attnblock_forward(self, x): try: h_ = x @@ -592,6 +624,7 @@ def xformers_attnblock_forward(self, x): except NotImplementedError: return cross_attention_attnblock_forward(self, x) + def sdp_attnblock_forward(self, x): h_ = x h_ = self.norm(h_) @@ -612,10 +645,12 @@ def sdp_attnblock_forward(self, x): out = self.proj_out(out) return x + out + def sdp_no_mem_attnblock_forward(self, x): with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=True, enable_mem_efficient=False): return sdp_attnblock_forward(self, x) + def sub_quad_attnblock_forward(self, x): h_ = x h_ = self.norm(h_) diff --git a/modules/sd_models.py b/modules/sd_models.py index 8d639583..e4aae597 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -411,6 +411,7 @@ def repair_config(sd_config): sd1_clip_weight = 'cond_stage_model.transformer.text_model.embeddings.token_embedding.weight' sd2_clip_weight = 'cond_stage_model.model.transformer.resblocks.0.attn.in_proj_weight' +sdxl_clip_weight = 'conditioner.embedders.1.model.ln_final.weight' class SdModelData: @@ -445,6 +446,15 @@ class SdModelData: model_data = SdModelData() +def get_empty_cond(sd_model): + if hasattr(sd_model, 'conditioner'): + d = sd_model.get_learned_conditioning([""]) + return d['crossattn'] + else: + return sd_model.cond_stage_model([""]) + + + def load_model(checkpoint_info=None, already_loaded_state_dict=None): from modules import lowvram, sd_hijack checkpoint_info = checkpoint_info or select_checkpoint() @@ -465,7 +475,7 @@ def load_model(checkpoint_info=None, already_loaded_state_dict=None): state_dict = get_checkpoint_state_dict(checkpoint_info, timer) checkpoint_config = sd_models_config.find_checkpoint_config(state_dict, checkpoint_info) - clip_is_included_into_sd = sd1_clip_weight in state_dict or sd2_clip_weight in state_dict + clip_is_included_into_sd = sd1_clip_weight in state_dict or sd2_clip_weight in state_dict or sdxl_clip_weight in state_dict timer.record("find config") @@ -517,7 +527,7 @@ def load_model(checkpoint_info=None, already_loaded_state_dict=None): timer.record("scripts callbacks") with devices.autocast(), torch.no_grad(): - sd_model.cond_stage_model_empty_prompt = sd_model.cond_stage_model([""]) + sd_model.cond_stage_model_empty_prompt = get_empty_cond(sd_model) timer.record("calculate empty prompt") diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py index 96501569..2e92479a 100644 --- a/modules/sd_models_config.py +++ b/modules/sd_models_config.py @@ -14,6 +14,7 @@ config_sd2 = os.path.join(sd_repo_configs_path, "v2-inference.yaml") config_sd2v = os.path.join(sd_repo_configs_path, "v2-inference-v.yaml") config_sd2v = os.path.join(sd_xl_repo_configs_path, "sd_2_1_768.yaml") config_sd2_inpainting = os.path.join(sd_repo_configs_path, "v2-inpainting-inference.yaml") +config_sdxl = os.path.join(sd_xl_repo_configs_path, "sd_xl_base.yaml") config_depth_model = os.path.join(sd_repo_configs_path, "v2-midas-inference.yaml") config_unclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-l-inference.yaml") config_unopenclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-h-inference.yaml") @@ -70,7 +71,9 @@ def guess_model_config_from_state_dict(sd, filename): diffusion_model_input = sd.get('model.diffusion_model.input_blocks.0.0.weight', None) sd2_variations_weight = sd.get('embedder.model.ln_final.weight', None) - if sd.get('depth_model.model.pretrained.act_postprocess3.0.project.0.bias', None) is not None: + if sd.get('conditioner.embedders.1.model.ln_final.weight', None) is not None: + return config_sdxl + elif sd.get('depth_model.model.pretrained.act_postprocess3.0.project.0.bias', None) is not None: return config_depth_model elif sd2_variations_weight is not None and sd2_variations_weight.shape[0] == 768: return config_unclip diff --git a/modules/sd_models_xl.py b/modules/sd_models_xl.py index d43b8868..e8e270c3 100644 --- a/modules/sd_models_xl.py +++ b/modules/sd_models_xl.py @@ -1,18 +1,30 @@ from __future__ import annotations +import sys + import torch import sgm.models.diffusion import sgm.modules.diffusionmodules.denoiser_scaling import sgm.modules.diffusionmodules.discretizer -from modules import devices +from modules import devices, shared, prompt_parser -def get_learned_conditioning(self: sgm.models.diffusion.DiffusionEngine, batch: list[str]): +def get_learned_conditioning(self: sgm.models.diffusion.DiffusionEngine, batch: prompt_parser.SdConditioning | list[str]): for embedder in self.conditioner.embedders: embedder.ucg_rate = 0.0 - c = self.conditioner({'txt': batch}) + width = getattr(self, 'target_width', 1024) + height = getattr(self, 'target_height', 1024) + + sdxl_conds = { + "txt": batch, + "original_size_as_tuple": torch.tensor([height, width]).repeat(len(batch), 1).to(devices.device, devices.dtype), + "crop_coords_top_left": torch.tensor([shared.opts.sdxl_crop_top, shared.opts.sdxl_crop_left]).repeat(len(batch), 1).to(devices.device, devices.dtype), + "target_size_as_tuple": torch.tensor([height, width]).repeat(len(batch), 1).to(devices.device, devices.dtype), + } + + c = self.conditioner(sdxl_conds) return c @@ -26,7 +38,7 @@ def extend_sdxl(model): model.model.diffusion_model.dtype = dtype model.model.conditioning_key = 'crossattn' - model.cond_stage_model = [x for x in model.conditioner.embedders if type(x).__name__ == 'FrozenOpenCLIPEmbedder'][0] + model.cond_stage_model = [x for x in model.conditioner.embedders if 'CLIPEmbedder' in type(x).__name__][0] model.cond_stage_key = model.cond_stage_model.input_key model.parameterization = "v" if isinstance(model.denoiser.scaling, sgm.modules.diffusionmodules.denoiser_scaling.VScaling) else "eps" @@ -34,7 +46,14 @@ def extend_sdxl(model): discretization = sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization() model.alphas_cumprod = torch.asarray(discretization.alphas_cumprod, device=devices.device, dtype=dtype) + model.is_xl = True + sgm.models.diffusion.DiffusionEngine.get_learned_conditioning = get_learned_conditioning sgm.models.diffusion.DiffusionEngine.apply_model = apply_model +sgm.modules.attention.print = lambda *args: None +sgm.modules.diffusionmodules.model.print = lambda *args: None +sgm.modules.diffusionmodules.openaimodel.print = lambda *args: None +sgm.modules.encoders.modules.print = lambda *args: None + diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py index 73289ce4..5552a8dc 100644 --- a/modules/sd_samplers_kdiffusion.py +++ b/modules/sd_samplers_kdiffusion.py @@ -186,7 +186,7 @@ class CFGDenoiser(torch.nn.Module): for batch_offset in range(0, x_out.shape[0], batch_size): a = batch_offset b = a + batch_size - x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=make_condition_dict(cond_in[a:b], image_cond_in[a:b])) + x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=make_condition_dict(subscript_cond(cond_in, a, b), image_cond_in[a:b])) else: x_out = torch.zeros_like(x_in) batch_size = batch_size*2 if shared.batch_cond_uncond else batch_size diff --git a/modules/shared.py b/modules/shared.py index b7518de6..71afd94f 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -428,6 +428,8 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "CLIP_stop_at_last_layers": OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}).link("wiki", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip").info("ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer"), "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"), "randn_source": OptionInfo("GPU", "Random number generator source.", gr.Radio, {"choices": ["GPU", "CPU"]}).info("changes seeds drastically; use CPU to produce the same picture across different videocard vendors"), + "sdxl_crop_top": OptionInfo(0, "SDXL top coordinate of the crop"), + "sdxl_crop_left": OptionInfo(0, "SDXL left coordinate of the crop"), })) options_templates.update(options_section(('optimizations', "Optimizations"), { diff --git a/requirements.txt b/requirements.txt index 3142085e..b3f8a7f4 100644 --- a/requirements.txt +++ b/requirements.txt @@ -14,6 +14,7 @@ kornia lark numpy omegaconf +open-clip-torch piexif psutil diff --git a/requirements_versions.txt b/requirements_versions.txt index f71b9d6c..b826bf43 100644 --- a/requirements_versions.txt +++ b/requirements_versions.txt @@ -15,6 +15,7 @@ kornia==0.6.7 lark==1.1.2 numpy==1.23.5 omegaconf==2.2.3 +open-clip-torch==2.20.0 piexif==1.1.3 psutil~=5.9.5 pytorch_lightning==1.9.4 -- cgit v1.2.3 From 6d8dcdefa07d5f8f7e528046b0facdcc51185e60 Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Fri, 14 Jul 2023 09:16:01 +0300 Subject: initial SDXL refiner support --- modules/sd_hijack.py | 18 ++++++++++---- modules/sd_models.py | 3 ++- modules/sd_models_config.py | 3 +++ modules/sd_models_xl.py | 57 ++++++++++++++++++++++++++++++++++++--------- modules/shared.py | 9 +++++-- 5 files changed, 71 insertions(+), 19 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 647cdfbe..2b274c18 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -180,21 +180,29 @@ class StableDiffusionModelHijack: def hijack(self, m): conditioner = getattr(m, 'conditioner', None) if conditioner: + text_cond_models = [] + for i in range(len(conditioner.embedders)): embedder = conditioner.embedders[i] typename = type(embedder).__name__ if typename == 'FrozenOpenCLIPEmbedder': embedder.model.token_embedding = EmbeddingsWithFixes(embedder.model.token_embedding, self) - m.cond_stage_model = sd_hijack_open_clip.FrozenOpenCLIPEmbedderWithCustomWords(embedder, self) - conditioner.embedders[i] = m.cond_stage_model + conditioner.embedders[i] = sd_hijack_open_clip.FrozenOpenCLIPEmbedderWithCustomWords(embedder, self) + text_cond_models.append(conditioner.embedders[i]) if typename == 'FrozenCLIPEmbedder': - model_embeddings = m.cond_stage_model.transformer.text_model.embeddings + model_embeddings = embedder.transformer.text_model.embeddings model_embeddings.token_embedding = EmbeddingsWithFixes(model_embeddings.token_embedding, self) - m.cond_stage_model = sd_hijack_clip.FrozenCLIPEmbedderForSDXLWithCustomWords(embedder, self) - conditioner.embedders[i] = m.cond_stage_model + conditioner.embedders[i] = sd_hijack_clip.FrozenCLIPEmbedderForSDXLWithCustomWords(embedder, self) + text_cond_models.append(conditioner.embedders[i]) if typename == 'FrozenOpenCLIPEmbedder2': embedder.model.token_embedding = EmbeddingsWithFixes(embedder.model.token_embedding, self) conditioner.embedders[i] = sd_hijack_open_clip.FrozenOpenCLIPEmbedder2WithCustomWords(embedder, self) + text_cond_models.append(conditioner.embedders[i]) + + if len(text_cond_models) == 1: + m.cond_stage_model = text_cond_models[0] + else: + m.cond_stage_model = conditioner if type(m.cond_stage_model) == xlmr.BertSeriesModelWithTransformation: model_embeddings = m.cond_stage_model.roberta.embeddings diff --git a/modules/sd_models.py b/modules/sd_models.py index 07702175..267f4d8e 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -414,6 +414,7 @@ def repair_config(sd_config): sd1_clip_weight = 'cond_stage_model.transformer.text_model.embeddings.token_embedding.weight' sd2_clip_weight = 'cond_stage_model.model.transformer.resblocks.0.attn.in_proj_weight' sdxl_clip_weight = 'conditioner.embedders.1.model.ln_final.weight' +sdxl_refiner_clip_weight = 'conditioner.embedders.0.model.ln_final.weight' class SdModelData: @@ -477,7 +478,7 @@ def load_model(checkpoint_info=None, already_loaded_state_dict=None): state_dict = get_checkpoint_state_dict(checkpoint_info, timer) checkpoint_config = sd_models_config.find_checkpoint_config(state_dict, checkpoint_info) - clip_is_included_into_sd = sd1_clip_weight in state_dict or sd2_clip_weight in state_dict or sdxl_clip_weight in state_dict + clip_is_included_into_sd = any([x for x in [sd1_clip_weight, sd2_clip_weight, sdxl_clip_weight, sdxl_refiner_clip_weight] if x in state_dict]) timer.record("find config") diff --git a/modules/sd_models_config.py b/modules/sd_models_config.py index 04c09ab0..8266fa39 100644 --- a/modules/sd_models_config.py +++ b/modules/sd_models_config.py @@ -14,6 +14,7 @@ config_sd2 = os.path.join(sd_repo_configs_path, "v2-inference.yaml") config_sd2v = os.path.join(sd_repo_configs_path, "v2-inference-v.yaml") config_sd2_inpainting = os.path.join(sd_repo_configs_path, "v2-inpainting-inference.yaml") config_sdxl = os.path.join(sd_xl_repo_configs_path, "sd_xl_base.yaml") +config_sdxl_refiner = os.path.join(sd_xl_repo_configs_path, "sd_xl_refiner.yaml") config_depth_model = os.path.join(sd_repo_configs_path, "v2-midas-inference.yaml") config_unclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-l-inference.yaml") config_unopenclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-h-inference.yaml") @@ -72,6 +73,8 @@ def guess_model_config_from_state_dict(sd, filename): if sd.get('conditioner.embedders.1.model.ln_final.weight', None) is not None: return config_sdxl + if sd.get('conditioner.embedders.0.model.ln_final.weight', None) is not None: + return config_sdxl_refiner elif sd.get('depth_model.model.pretrained.act_postprocess3.0.project.0.bias', None) is not None: return config_depth_model elif sd2_variations_weight is not None and sd2_variations_weight.shape[0] == 768: diff --git a/modules/sd_models_xl.py b/modules/sd_models_xl.py index a7240dc0..01320c7a 100644 --- a/modules/sd_models_xl.py +++ b/modules/sd_models_xl.py @@ -14,15 +14,20 @@ def get_learned_conditioning(self: sgm.models.diffusion.DiffusionEngine, batch: width = getattr(self, 'target_width', 1024) height = getattr(self, 'target_height', 1024) + is_negative_prompt = getattr(batch, 'is_negative_prompt', False) + aesthetic_score = shared.opts.sdxl_refiner_low_aesthetic_score if is_negative_prompt else shared.opts.sdxl_refiner_high_aesthetic_score + + devices_args = dict(device=devices.device, dtype=devices.dtype) sdxl_conds = { "txt": batch, - "original_size_as_tuple": torch.tensor([height, width]).repeat(len(batch), 1).to(devices.device, devices.dtype), - "crop_coords_top_left": torch.tensor([shared.opts.sdxl_crop_top, shared.opts.sdxl_crop_left]).repeat(len(batch), 1).to(devices.device, devices.dtype), - "target_size_as_tuple": torch.tensor([height, width]).repeat(len(batch), 1).to(devices.device, devices.dtype), + "original_size_as_tuple": torch.tensor([height, width], **devices_args).repeat(len(batch), 1), + "crop_coords_top_left": torch.tensor([shared.opts.sdxl_crop_top, shared.opts.sdxl_crop_left], **devices_args).repeat(len(batch), 1), + "target_size_as_tuple": torch.tensor([height, width], **devices_args).repeat(len(batch), 1), + "aesthetic_score": torch.tensor([aesthetic_score], **devices_args).repeat(len(batch), 1), } - force_zero_negative_prompt = getattr(batch, 'is_negative_prompt', False) and all(x == '' for x in batch) + force_zero_negative_prompt = is_negative_prompt and all(x == '' for x in batch) c = self.conditioner(sdxl_conds, force_zero_embeddings=['txt'] if force_zero_negative_prompt else []) return c @@ -35,25 +40,55 @@ def apply_model(self: sgm.models.diffusion.DiffusionEngine, x, t, cond): def get_first_stage_encoding(self, x): # SDXL's encode_first_stage does everything so get_first_stage_encoding is just there for compatibility return x + +sgm.models.diffusion.DiffusionEngine.get_learned_conditioning = get_learned_conditioning +sgm.models.diffusion.DiffusionEngine.apply_model = apply_model +sgm.models.diffusion.DiffusionEngine.get_first_stage_encoding = get_first_stage_encoding + + +def encode_embedding_init_text(self: sgm.modules.GeneralConditioner, init_text, nvpt): + res = [] + + for embedder in [embedder for embedder in self.embedders if hasattr(embedder, 'encode_embedding_init_text')]: + encoded = embedder.encode_embedding_init_text(init_text, nvpt) + res.append(encoded) + + return torch.cat(res, dim=1) + + +def process_texts(self, texts): + for embedder in [embedder for embedder in self.embedders if hasattr(embedder, 'process_texts')]: + return embedder.process_texts(texts) + + +def get_target_prompt_token_count(self, token_count): + for embedder in [embedder for embedder in self.embedders if hasattr(embedder, 'get_target_prompt_token_count')]: + return embedder.get_target_prompt_token_count(token_count) + + +# those additions to GeneralConditioner make it possible to use it as model.cond_stage_model from SD1.5 in exist +sgm.modules.GeneralConditioner.encode_embedding_init_text = encode_embedding_init_text +sgm.modules.GeneralConditioner.process_texts = process_texts +sgm.modules.GeneralConditioner.get_target_prompt_token_count = get_target_prompt_token_count + + def extend_sdxl(model): + """this adds a bunch of parameters to make SDXL model look a bit more like SD1.5 to the rest of the codebase.""" + dtype = next(model.model.diffusion_model.parameters()).dtype model.model.diffusion_model.dtype = dtype model.model.conditioning_key = 'crossattn' - - model.cond_stage_model = [x for x in model.conditioner.embedders if 'CLIPEmbedder' in type(x).__name__][0] - model.cond_stage_key = model.cond_stage_model.input_key + model.cond_stage_key = 'txt' + # model.cond_stage_model will be set in sd_hijack model.parameterization = "v" if isinstance(model.denoiser.scaling, sgm.modules.diffusionmodules.denoiser_scaling.VScaling) else "eps" discretization = sgm.modules.diffusionmodules.discretizer.LegacyDDPMDiscretization() model.alphas_cumprod = torch.asarray(discretization.alphas_cumprod, device=devices.device, dtype=dtype) + model.conditioner.wrapped = torch.nn.Module() -sgm.models.diffusion.DiffusionEngine.get_learned_conditioning = get_learned_conditioning -sgm.models.diffusion.DiffusionEngine.apply_model = apply_model -sgm.models.diffusion.DiffusionEngine.get_first_stage_encoding = get_first_stage_encoding - sgm.modules.attention.print = lambda *args: None sgm.modules.diffusionmodules.model.print = lambda *args: None sgm.modules.diffusionmodules.openaimodel.print = lambda *args: None diff --git a/modules/shared.py b/modules/shared.py index 71afd94f..234ede0d 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -428,8 +428,13 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "CLIP_stop_at_last_layers": OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}).link("wiki", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip").info("ignore last layers of CLIP network; 1 ignores none, 2 ignores one layer"), "upcast_attn": OptionInfo(False, "Upcast cross attention layer to float32"), "randn_source": OptionInfo("GPU", "Random number generator source.", gr.Radio, {"choices": ["GPU", "CPU"]}).info("changes seeds drastically; use CPU to produce the same picture across different videocard vendors"), - "sdxl_crop_top": OptionInfo(0, "SDXL top coordinate of the crop"), - "sdxl_crop_left": OptionInfo(0, "SDXL left coordinate of the crop"), +})) + +options_templates.update(options_section(('sdxl', "Stable Diffusion XL"), { + "sdxl_crop_top": OptionInfo(0, "crop top coordinate"), + "sdxl_crop_left": OptionInfo(0, "crop left coordinate"), + "sdxl_refiner_low_aesthetic_score": OptionInfo(2.5, "SDXL low aesthetic score", gr.Number).info("used for refiner model negative prompt"), + "sdxl_refiner_high_aesthetic_score": OptionInfo(6.0, "SDXL high aesthetic score", gr.Number).info("used for refiner model prompt"), })) options_templates.update(options_section(('optimizations', "Optimizations"), { -- cgit v1.2.3 From abb948dab09841571dd24c6be9ff9d6b212778ea Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Fri, 14 Jul 2023 09:28:01 +0300 Subject: raise maximum Negative Guidance minimum sigma due to request in PR discussion --- modules/shared.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index 234ede0d..89b7132e 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -439,7 +439,7 @@ options_templates.update(options_section(('sdxl', "Stable Diffusion XL"), { options_templates.update(options_section(('optimizations', "Optimizations"), { "cross_attention_optimization": OptionInfo("Automatic", "Cross attention optimization", gr.Dropdown, lambda: {"choices": shared_items.cross_attention_optimizations()}), - "s_min_uncond": OptionInfo(0.0, "Negative Guidance minimum sigma", gr.Slider, {"minimum": 0.0, "maximum": 4.0, "step": 0.01}).link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/9177").info("skip negative prompt for some steps when the image is almost ready; 0=disable, higher=faster"), + "s_min_uncond": OptionInfo(0.0, "Negative Guidance minimum sigma", gr.Slider, {"minimum": 0.0, "maximum": 15.0, "step": 0.01}).link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/9177").info("skip negative prompt for some steps when the image is almost ready; 0=disable, higher=faster"), "token_merging_ratio": OptionInfo(0.0, "Token merging ratio", gr.Slider, {"minimum": 0.0, "maximum": 0.9, "step": 0.1}).link("PR", "https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/9256").info("0=disable, higher=faster"), "token_merging_ratio_img2img": OptionInfo(0.0, "Token merging ratio for img2img", gr.Slider, {"minimum": 0.0, "maximum": 0.9, "step": 0.1}).info("only applies if non-zero and overrides above"), "token_merging_ratio_hr": OptionInfo(0.0, "Token merging ratio for high-res pass", gr.Slider, {"minimum": 0.0, "maximum": 0.9, "step": 0.1}).info("only applies if non-zero and overrides above"), -- cgit v1.2.3 From 2b1bae0d755c2d5201f6a6aadeadb5588208d43f Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Sat, 15 Jul 2023 08:41:22 +0300 Subject: add textual inversion hashes to infotext --- modules/processing.py | 7 ++++--- modules/sd_hijack.py | 5 ++++- modules/sd_hijack_clip.py | 15 ++++++++++++--- modules/shared.py | 1 + modules/textual_inversion/textual_inversion.py | 9 ++++++++- style.css | 4 ++++ 6 files changed, 33 insertions(+), 8 deletions(-) (limited to 'modules/shared.py') diff --git a/modules/processing.py b/modules/processing.py index cd568a20..49441e77 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -732,9 +732,10 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed: p.setup_conds() - if len(model_hijack.comments) > 0: - for comment in model_hijack.comments: - comments[comment] = 1 + for comment in model_hijack.comments: + comments[comment] = 1 + + p.extra_generation_params.update(model_hijack.extra_generation_params) if p.n_iter > 1: shared.state.job = f"Batch {n+1} out of {p.n_iter}" diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 3b6f95ce..6b5aae4b 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -147,7 +147,6 @@ def undo_weighted_forward(sd_model): class StableDiffusionModelHijack: fixes = None - comments = [] layers = None circular_enabled = False clip = None @@ -156,6 +155,9 @@ class StableDiffusionModelHijack: embedding_db = modules.textual_inversion.textual_inversion.EmbeddingDatabase() def __init__(self): + self.extra_generation_params = {} + self.comments = [] + self.embedding_db.add_embedding_dir(cmd_opts.embeddings_dir) def apply_optimizations(self, option=None): @@ -236,6 +238,7 @@ class StableDiffusionModelHijack: def clear_comments(self): self.comments = [] + self.extra_generation_params = {} def get_prompt_lengths(self, text): if self.clip is None: diff --git a/modules/sd_hijack_clip.py b/modules/sd_hijack_clip.py index 3b5a7666..c1d780a3 100644 --- a/modules/sd_hijack_clip.py +++ b/modules/sd_hijack_clip.py @@ -229,9 +229,18 @@ class FrozenCLIPEmbedderWithCustomWordsBase(torch.nn.Module): z = self.process_tokens(tokens, multipliers) zs.append(z) - if len(used_embeddings) > 0: - embeddings_list = ", ".join([f'{name} [{embedding.checksum()}]' for name, embedding in used_embeddings.items()]) - self.hijack.comments.append(f"Used embeddings: {embeddings_list}") + if opts.textual_inversion_add_hashes_to_infotext and used_embeddings: + hashes = [] + for name, embedding in used_embeddings.items(): + shorthash = embedding.shorthash + if not shorthash: + continue + + name = name.replace(":", "").replace(",", "") + hashes.append(f"{name}: {shorthash}") + + if hashes: + self.hijack.extra_generation_params["TI hashes"] = ", ".join(hashes) return torch.hstack(zs) diff --git a/modules/shared.py b/modules/shared.py index 48478a68..a32fd4ed 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -472,6 +472,7 @@ options_templates.update(options_section(('extra_networks', "Extra Networks"), { "extra_networks_card_height": OptionInfo(0, "Card height for Extra Networks").info("in pixels"), "extra_networks_add_text_separator": OptionInfo(" ", "Extra networks separator").info("extra text to add before <...> when adding extra network to prompt"), "ui_extra_networks_tab_reorder": OptionInfo("", "Extra networks tab order").needs_restart(), + "textual_inversion_add_hashes_to_infotext": OptionInfo(True, "Add Textual Inversion hashes to infotext"), "sd_hypernetwork": OptionInfo("None", "Add hypernetwork to prompt", gr.Dropdown, lambda: {"choices": ["None", *hypernetworks]}, refresh=reload_hypernetworks), })) diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index cbe975b7..38e072a8 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -13,7 +13,7 @@ import numpy as np from PIL import Image, PngImagePlugin from torch.utils.tensorboard import SummaryWriter -from modules import shared, devices, sd_hijack, processing, sd_models, images, sd_samplers, sd_hijack_checkpoint, errors +from modules import shared, devices, sd_hijack, processing, sd_models, images, sd_samplers, sd_hijack_checkpoint, errors, hashes import modules.textual_inversion.dataset from modules.textual_inversion.learn_schedule import LearnRateScheduler @@ -49,6 +49,8 @@ class Embedding: self.sd_checkpoint_name = None self.optimizer_state_dict = None self.filename = None + self.hash = None + self.shorthash = None def save(self, filename): embedding_data = { @@ -82,6 +84,10 @@ class Embedding: self.cached_checksum = f'{const_hash(self.vec.reshape(-1) * 100) & 0xffff:04x}' return self.cached_checksum + def set_hash(self, v): + self.hash = v + self.shorthash = self.hash[0:12] + class DirWithTextualInversionEmbeddings: def __init__(self, path): @@ -199,6 +205,7 @@ class EmbeddingDatabase: embedding.vectors = vec.shape[0] embedding.shape = vec.shape[-1] embedding.filename = path + embedding.set_hash(hashes.sha256(embedding.filename, "textual_inversion/" + name) or '') if self.expected_shape == -1 or self.expected_shape == embedding.shape: self.register_embedding(embedding, shared.sd_model) diff --git a/style.css b/style.css index a424067f..9e13d7fd 100644 --- a/style.css +++ b/style.css @@ -231,6 +231,10 @@ button.custom-button{ padding-top: 0.5em; } +.html-log .comments:empty{ + padding-top: 0; +} + .html-log .performance { font-size: 0.85em; color: #444; -- cgit v1.2.3 From 0aa8d538e147ba87df36d8196845807c8fa3f4e1 Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Sat, 15 Jul 2023 09:24:22 +0300 Subject: suppress printing TI embedding into console by default --- modules/shared.py | 1 + modules/textual_inversion/textual_inversion.py | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'modules/shared.py') diff --git a/modules/shared.py b/modules/shared.py index a32fd4ed..427dcc50 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -472,6 +472,7 @@ options_templates.update(options_section(('extra_networks', "Extra Networks"), { "extra_networks_card_height": OptionInfo(0, "Card height for Extra Networks").info("in pixels"), "extra_networks_add_text_separator": OptionInfo(" ", "Extra networks separator").info("extra text to add before <...> when adding extra network to prompt"), "ui_extra_networks_tab_reorder": OptionInfo("", "Extra networks tab order").needs_restart(), + "textual_inversion_print_at_load": OptionInfo(False, "Print a list of Textual Inversion embeddings when loading model"), "textual_inversion_add_hashes_to_infotext": OptionInfo(True, "Add Textual Inversion hashes to infotext"), "sd_hypernetwork": OptionInfo("None", "Add hypernetwork to prompt", gr.Dropdown, lambda: {"choices": ["None", *hypernetworks]}, refresh=reload_hypernetworks), })) diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index 38e072a8..6166c76f 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -256,7 +256,7 @@ class EmbeddingDatabase: self.word_embeddings.update(sorted_word_embeddings) displayed_embeddings = (tuple(self.word_embeddings.keys()), tuple(self.skipped_embeddings.keys())) - if self.previously_displayed_embeddings != displayed_embeddings: + if shared.opts.textual_inversion_print_at_load and self.previously_displayed_embeddings != displayed_embeddings: self.previously_displayed_embeddings = displayed_embeddings print(f"Textual inversion embeddings loaded({len(self.word_embeddings)}): {', '.join(self.word_embeddings.keys())}") if self.skipped_embeddings: -- cgit v1.2.3 From 690d56f3c10e5359e15eeba9c68e56b2eb193ac3 Mon Sep 17 00:00:00 2001 From: AUTOMATIC1111 <16777216c@gmail.com> Date: Sun, 16 Jul 2023 10:25:34 +0300 Subject: nuke thumbs extra networks view mode (use settings tab to change width/height/scale to get thumbs) --- html/image-update.svg | 7 ---- javascript/hints.js | 1 - modules/shared.py | 5 ++- modules/ui_extra_networks.py | 9 ++--- style.css | 94 +++++++------------------------------------- 5 files changed, 22 insertions(+), 94 deletions(-) delete mode 100644 html/image-update.svg (limited to 'modules/shared.py') diff --git a/html/image-update.svg b/html/image-update.svg deleted file mode 100644 index 3abf12df..00000000 --- a/html/image-update.svg +++ /dev/null @@ -1,7 +0,0 @@ - - - - - - - diff --git a/javascript/hints.js b/javascript/hints.js index 41201b2f..4167cb28 100644 --- a/javascript/hints.js +++ b/javascript/hints.js @@ -108,7 +108,6 @@ var titles = { "Upscale by": "Adjusts the size of the image by multiplying the original width and height by the selected value. Ignored if either Resize width to or Resize height to are non-zero.", "Resize width to": "Resizes image to this width. If 0, width is inferred from either of two nearby sliders.", "Resize height to": "Resizes image to this height. If 0, height is inferred from either of two nearby sliders.", - "Multiplier for extra networks": "When adding extra network such as Hypernetwork or Lora to prompt, use this multiplier for it.", "Discard weights with matching name": "Regular expression; if weights's name matches it, the weights is not written to the resulting checkpoint. Use ^model_ema to discard EMA weights.", "Extra networks tab order": "Comma-separated list of tab names; tabs listed here will appear in the extra networks UI first and in order listed.", "Negative Guidance minimum sigma": "Skip negative prompt for steps where image is already mostly denoised; the higher this value, the more skips there will be; provides increased performance in exchange for minor quality reduction." diff --git a/modules/shared.py b/modules/shared.py index 427dcc50..f6604ef9 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -466,10 +466,11 @@ options_templates.update(options_section(('interrogate', "Interrogate Options"), options_templates.update(options_section(('extra_networks', "Extra Networks"), { "extra_networks_show_hidden_directories": OptionInfo(True, "Show hidden directories").info("directory is hidden if its name starts with \".\"."), "extra_networks_hidden_models": OptionInfo("When searched", "Show cards for models in hidden directories", gr.Radio, {"choices": ["Always", "When searched", "Never"]}).info('"When searched" option will only show the item when the search string has 4 characters or more'), - "extra_networks_default_view": OptionInfo("cards", "Default view for Extra Networks", gr.Dropdown, {"choices": ["cards", "thumbs"]}), - "extra_networks_default_multiplier": OptionInfo(1.0, "Multiplier for extra networks", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.01}), + "extra_networks_default_multiplier": OptionInfo(1.0, "Default multiplier for extra networks", gr.Slider, {"minimum": 0.0, "maximum": 2.0, "step": 0.01}), "extra_networks_card_width": OptionInfo(0, "Card width for Extra Networks").info("in pixels"), "extra_networks_card_height": OptionInfo(0, "Card height for Extra Networks").info("in pixels"), + "extra_networks_card_text_scale": OptionInfo(1.0, "Card text scale", gr.Slider, {"minimum": 0.0, "maximum": 2.0, "step": 0.01}).info("1 = original size"), + "extra_networks_card_show_desc": OptionInfo(True, "Show description on card"), "extra_networks_add_text_separator": OptionInfo(" ", "Extra networks separator").info("extra text to add before <...> when adding extra network to prompt"), "ui_extra_networks_tab_reorder": OptionInfo("", "Extra networks tab order").needs_restart(), "textual_inversion_print_at_load": OptionInfo(False, "Print a list of Textual Inversion embeddings when loading model"), diff --git a/modules/ui_extra_networks.py b/modules/ui_extra_networks.py index a4927c11..d9deccb2 100644 --- a/modules/ui_extra_networks.py +++ b/modules/ui_extra_networks.py @@ -132,7 +132,6 @@ class ExtraNetworksPage: return "" def create_html(self, tabname): - view = shared.opts.extra_networks_default_view items_html = '' self.metadata = {} @@ -186,10 +185,10 @@ class ExtraNetworksPage: self_name_id = self.name.replace(" ", "_") res = f""" -
+
{subdirs_html}
-
+
{items_html}
""" @@ -248,12 +247,12 @@ class ExtraNetworksPage: args = { "background_image": background_image, - "style": f"'display: none; {height}{width}'", + "style": f"'display: none; {height}{width}; font-size: {shared.opts.extra_networks_card_text_scale*100}%'", "prompt": item.get("prompt", None), "tabname": quote_js(tabname), "local_preview": quote_js(item["local_preview"]), "name": item["name"], - "description": (item.get("description") or ""), + "description": (item.get("description") or "" if shared.opts.extra_networks_card_show_desc else ""), "card_clicked": onclick, "save_card_preview": '"' + html.escape(f"""return saveCardPreview(event, {quote_js(tabname)}, {quote_js(item["local_preview"])})""") + '"', "search_term": item.get("search_term", ""), diff --git a/style.css b/style.css index 4e22cfd6..f7a8cb17 100644 --- a/style.css +++ b/style.css @@ -804,127 +804,67 @@ footer { width: auto; } -.extra-network-cards .nocards, .extra-network-thumbs .nocards{ +.extra-network-cards .nocards{ margin: 1.25em 0.5em 0.5em 0.5em; } -.extra-network-cards .nocards h1, .extra-network-thumbs .nocards h1{ +.extra-network-cards .nocards h1{ font-size: 1.5em; margin-bottom: 1em; } -.extra-network-cards .nocards li, .extra-network-thumbs .nocards li{ +.extra-network-cards .nocards li{ margin-left: 0.5em; } -.extra-network-cards .card .button-row, .extra-network-thumbs .card .button-row{ +.extra-network-cards .card .button-row{ display: none; position: absolute; color: white; right: 0; } -.extra-network-cards .card:hover .button-row, .extra-network-thumbs .card:hover .button-row{ +.extra-network-cards .card:hover .button-row{ display: flex; } -.extra-network-cards .card .card-button, .extra-network-thumbs .card .card-button{ +.extra-network-cards .card .card-button{ color: white; } -.extra-network-cards .card .metadata-button:before, .extra-network-thumbs .card .metadata-button:before{ +.extra-network-cards .card .metadata-button:before{ content: "🛈"; } -.extra-network-cards .card .edit-button:before, .extra-network-thumbs .card .edit-button:before{ +.extra-network-cards .card .edit-button:before{ content: "🛠"; } .extra-network-cards .card .card-button { text-shadow: 2px 2px 3px black; padding: 0.25em; - font-size: 22pt; + font-size: 200%; width: 1.5em; } -.extra-network-thumbs .card .card-button { - text-shadow: 1px 1px 2px black; - padding: 0; - font-size: 16pt; - width: 1em; - top: -0.25em; -} -.extra-network-cards .card .card-button:hover, .extra-network-thumbs .card .card-button:hover{ +.extra-network-cards .card .card-button:hover{ color: red; } -.extra-network-thumbs { - display: flex; - flex-flow: row wrap; - gap: 10px; -} - -.extra-network-thumbs .card { - height: 6em; - width: 6em; - cursor: pointer; - background-image: url('./file=html/card-no-preview.png'); - background-size: cover; - background-position: center center; - position: relative; -} - -.extra-network-thumbs .card .preview, .standalone-card-preview.card .preview{ +.standalone-card-preview.card .preview{ position: absolute; object-fit: cover; width: 100%; height:100%; } -.extra-network-thumbs .card:hover .additional a { - display: inline-block; -} - -.extra-network-thumbs .actions .additional a { - background-image: url('./file=html/image-update.svg'); - background-repeat: no-repeat; - background-size: cover; - background-position: center center; - position: absolute; - top: 0; - left: 0; - width: 24px; - height: 24px; - display: none; - font-size: 0; - text-align: -9999; -} - -.extra-network-thumbs .actions .name { - position: absolute; - bottom: 0; - font-size: 10px; - padding: 3px; - width: 100%; - overflow: hidden; - white-space: nowrap; - text-overflow: ellipsis; - background: rgba(0,0,0,.5); - color: white; -} - -.extra-network-thumbs .card:hover .actions .name { - white-space: normal; - word-break: break-all; -} - .extra-network-cards .card, .standalone-card-preview.card{ display: inline-block; - margin: 0.5em; - width: 16em; - height: 24em; + margin: 0.5rem; + width: 16rem; + height: 24rem; box-shadow: 0 0 5px rgba(128, 128, 128, 0.5); - border-radius: 0.2em; + border-radius: 0.2rem; position: relative; background-size: auto 100%; @@ -958,10 +898,6 @@ footer { color: white; } -.extra-network-cards .card .actions:hover{ - box-shadow: 0 0 0.75em 0.75em rgba(0,0,0,0.5) !important; -} - .extra-network-cards .card .actions .name{ font-size: 1.7em; font-weight: bold; -- cgit v1.2.3