From bb57f30c2de46cfca5419ad01738a41705f96cc3 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Fri, 14 Oct 2022 10:56:41 +0200 Subject: init --- modules/processing.py | 17 +++++- modules/sd_hijack.py | 80 +++++++++++++++++++++++++- modules/shared.py | 5 ++ modules/textual_inversion/dataset.py | 2 +- modules/textual_inversion/textual_inversion.py | 35 +++++++---- modules/txt2img.py | 11 +++- modules/ui.py | 59 ++++++++++++------- 7 files changed, 171 insertions(+), 38 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index d5172f00..9a033759 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -316,11 +316,16 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration return f"{all_prompts[index]}{negative_prompt_text}\n{generation_params_text}".strip() -def process_images(p: StableDiffusionProcessing) -> Processed: +def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, + aesthetic_imgs=None,aesthetic_slerp=False) -> Processed: """this is the main loop that both txt2img and img2img use; it calls func_init once inside all the scopes and func_sample once per batch""" + aesthetic_lr = float(aesthetic_lr) + aesthetic_weight = float(aesthetic_weight) + aesthetic_steps = int(aesthetic_steps) + if type(p.prompt) == list: - assert(len(p.prompt) > 0) + assert (len(p.prompt) > 0) else: assert p.prompt is not None @@ -394,7 +399,13 @@ def process_images(p: StableDiffusionProcessing) -> Processed: #uc = p.sd_model.get_learned_conditioning(len(prompts) * [p.negative_prompt]) #c = p.sd_model.get_learned_conditioning(prompts) with devices.autocast(): - uc = prompt_parser.get_learned_conditioning(shared.sd_model, len(prompts) * [p.negative_prompt], p.steps) + if hasattr(shared.sd_model.cond_stage_model, "set_aesthetic_params"): + shared.sd_model.cond_stage_model.set_aesthetic_params(0, 0, 0) + uc = prompt_parser.get_learned_conditioning(shared.sd_model, len(prompts) * [p.negative_prompt], + p.steps) + if hasattr(shared.sd_model.cond_stage_model, "set_aesthetic_params"): + shared.sd_model.cond_stage_model.set_aesthetic_params(aesthetic_lr, aesthetic_weight, + aesthetic_steps, aesthetic_imgs,aesthetic_slerp) c = prompt_parser.get_multicond_learned_conditioning(shared.sd_model, prompts, p.steps) if len(model_hijack.comments) > 0: diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index c81722a0..6d5196fe 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -9,11 +9,14 @@ from torch.nn.functional import silu import modules.textual_inversion.textual_inversion from modules import prompt_parser, devices, sd_hijack_optimizations, shared -from modules.shared import opts, device, cmd_opts +from modules.shared import opts, device, cmd_opts, aesthetic_embeddings from modules.sd_hijack_optimizations import invokeAI_mps_available import ldm.modules.attention import ldm.modules.diffusionmodules.model +from transformers import CLIPVisionModel, CLIPModel +import torch.optim as optim +import copy attention_CrossAttention_forward = ldm.modules.attention.CrossAttention.forward diffusionmodules_model_nonlinearity = ldm.modules.diffusionmodules.model.nonlinearity @@ -109,13 +112,29 @@ class StableDiffusionModelHijack: _, remade_batch_tokens, _, _, _, token_count = self.clip.process_text([text]) return remade_batch_tokens[0], token_count, get_target_prompt_token_count(token_count) +def slerp(low, high, val): + low_norm = low/torch.norm(low, dim=1, keepdim=True) + high_norm = high/torch.norm(high, dim=1, keepdim=True) + omega = torch.acos((low_norm*high_norm).sum(1)) + so = torch.sin(omega) + res = (torch.sin((1.0-val)*omega)/so).unsqueeze(1)*low + (torch.sin(val*omega)/so).unsqueeze(1) * high + return res class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): def __init__(self, wrapped, hijack): super().__init__() self.wrapped = wrapped + self.clipModel = CLIPModel.from_pretrained( + self.wrapped.transformer.name_or_path + ) + del self.clipModel.vision_model self.hijack: StableDiffusionModelHijack = hijack self.tokenizer = wrapped.tokenizer + # self.vision = CLIPVisionModel.from_pretrained(self.wrapped.transformer.name_or_path).eval() + self.image_embs_name = None + self.image_embs = None + self.load_image_embs(None) + self.token_mults = {} self.comma_token = [v for k, v in self.tokenizer.get_vocab().items() if k == ','][0] @@ -136,6 +155,23 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): if mult != 1.0: self.token_mults[ident] = mult + def set_aesthetic_params(self, aesthetic_lr, aesthetic_weight, aesthetic_steps, image_embs_name=None, + aesthetic_slerp=True): + self.slerp = aesthetic_slerp + self.aesthetic_lr = aesthetic_lr + self.aesthetic_weight = aesthetic_weight + self.aesthetic_steps = aesthetic_steps + self.load_image_embs(image_embs_name) + + def load_image_embs(self, image_embs_name): + if image_embs_name is None or len(image_embs_name) == 0: + image_embs_name = None + if image_embs_name is not None and self.image_embs_name != image_embs_name: + self.image_embs_name = image_embs_name + self.image_embs = torch.load(aesthetic_embeddings[self.image_embs_name], map_location=device) + self.image_embs /= self.image_embs.norm(dim=-1, keepdim=True) + self.image_embs.requires_grad_(False) + def tokenize_line(self, line, used_custom_terms, hijack_comments): id_end = self.wrapped.tokenizer.eos_token_id @@ -333,7 +369,47 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): z1 = self.process_tokens(tokens, multipliers) z = z1 if z is None else torch.cat((z, z1), axis=-2) - + + if len(text[ + 0]) != 0 and self.aesthetic_steps != 0 and self.aesthetic_lr != 0 and self.aesthetic_weight != 0 and self.image_embs_name != None: + if not opts.use_old_emphasis_implementation: + remade_batch_tokens = [ + [self.wrapped.tokenizer.bos_token_id] + x[:75] + [self.wrapped.tokenizer.eos_token_id] for x in + remade_batch_tokens] + + tokens = torch.asarray(remade_batch_tokens).to(device) + with torch.enable_grad(): + model = copy.deepcopy(self.clipModel).to(device) + model.requires_grad_(True) + + # We optimize the model to maximize the similarity + optimizer = optim.Adam( + model.text_model.parameters(), lr=self.aesthetic_lr + ) + + for i in range(self.aesthetic_steps): + text_embs = model.get_text_features(input_ids=tokens) + text_embs = text_embs / text_embs.norm(dim=-1, keepdim=True) + sim = text_embs @ self.image_embs.T + loss = -sim + optimizer.zero_grad() + loss.mean().backward() + optimizer.step() + + zn = model.text_model(input_ids=tokens, output_hidden_states=-opts.CLIP_stop_at_last_layers) + if opts.CLIP_stop_at_last_layers > 1: + zn = zn.hidden_states[-opts.CLIP_stop_at_last_layers] + zn = model.text_model.final_layer_norm(zn) + else: + zn = zn.last_hidden_state + model.cpu() + del model + + if self.slerp: + z = slerp(z, zn, self.aesthetic_weight) + else: + z = z * (1 - self.aesthetic_weight) + zn * self.aesthetic_weight + remade_batch_tokens = rem_tokens batch_multipliers = rem_multipliers i += 1 diff --git a/modules/shared.py b/modules/shared.py index 5901e605..cf13a10d 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -30,6 +30,8 @@ parser.add_argument("--no-half-vae", action='store_true', help="do not switch th parser.add_argument("--no-progressbar-hiding", action='store_true', help="do not hide progressbar in gradio UI (we hide it because it slows down ML if you have hardware acceleration in browser)") parser.add_argument("--max-batch-count", type=int, default=16, help="maximum batch count value for the UI") parser.add_argument("--embeddings-dir", type=str, default=os.path.join(script_path, 'embeddings'), help="embeddings directory for textual inversion (default: embeddings)") +parser.add_argument("--aesthetic_embeddings-dir", type=str, default=os.path.join(script_path, 'aesthetic_embeddings'), + help="aesthetic_embeddings directory(default: aesthetic_embeddings)") parser.add_argument("--hypernetwork-dir", type=str, default=os.path.join(models_path, 'hypernetworks'), help="hypernetwork directory") parser.add_argument("--allow-code", action='store_true', help="allow custom script execution from webui") parser.add_argument("--medvram", action='store_true', help="enable stable diffusion model optimizations for sacrificing a little speed for low VRM usage") @@ -90,6 +92,9 @@ os.makedirs(cmd_opts.hypernetwork_dir, exist_ok=True) hypernetworks = hypernetwork.list_hypernetworks(cmd_opts.hypernetwork_dir) loaded_hypernetwork = None +aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in + os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} + def reload_hypernetworks(): global hypernetworks diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py index 67e90afe..59b2b021 100644 --- a/modules/textual_inversion/dataset.py +++ b/modules/textual_inversion/dataset.py @@ -48,7 +48,7 @@ class PersonalizedBase(Dataset): print("Preparing dataset...") for path in tqdm.tqdm(self.image_paths): try: - image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.BICUBIC) + image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.Resampling.BICUBIC) except Exception: continue diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index fa0e33a2..b12a8e6d 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -172,7 +172,15 @@ def create_embedding(name, num_vectors_per_token, init_text='*'): return fn -def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_width, training_height, steps, create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, preview_image_prompt): +def batched(dataset, total, n=1): + for ndx in range(0, total, n): + yield [dataset.__getitem__(i) for i in range(ndx, min(ndx + n, total))] + + +def train_embedding(embedding_name, learn_rate, data_root, log_directory, training_width, training_height, steps, + create_image_every, save_embedding_every, template_file, save_image_with_stored_embedding, + preview_image_prompt, batch_size=1, + gradient_accumulation=1): assert embedding_name, 'embedding not selected' shared.state.textinfo = "Initializing textual inversion training..." @@ -204,7 +212,11 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..." with torch.autocast("cuda"): - ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=embedding_name, model=shared.sd_model, device=devices.device, template_file=template_file) + ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, + height=training_height, + repeats=shared.opts.training_image_repeats_per_epoch, + placeholder_token=embedding_name, model=shared.sd_model, + device=devices.device, template_file=template_file) hijack = sd_hijack.model_hijack @@ -223,7 +235,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini scheduler = LearnRateScheduler(learn_rate, steps, ititial_step) optimizer = torch.optim.AdamW([embedding.vec], lr=scheduler.learn_rate) - pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step) + pbar = tqdm.tqdm(enumerate(batched(ds, steps - ititial_step, batch_size)), total=steps - ititial_step) for i, entry in pbar: embedding.step = i + ititial_step @@ -235,17 +247,20 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini break with torch.autocast("cuda"): - c = cond_model([entry.cond_text]) + c = cond_model([e.cond_text for e in entry]) + + x = torch.stack([e.latent for e in entry]).to(devices.device) + loss = shared.sd_model(x, c)[0] - x = entry.latent.to(devices.device) - loss = shared.sd_model(x.unsqueeze(0), c)[0] del x losses[embedding.step % losses.shape[0]] = loss.item() - optimizer.zero_grad() loss.backward() - optimizer.step() + if ((i + 1) % gradient_accumulation == 0) or (i + 1 == steps - ititial_step): + optimizer.step() + optimizer.zero_grad() + epoch_num = embedding.step // len(ds) epoch_step = embedding.step - (epoch_num * len(ds)) + 1 @@ -259,7 +274,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini if embedding.step > 0 and images_dir is not None and embedding.step % create_image_every == 0: last_saved_image = os.path.join(images_dir, f'{embedding_name}-{embedding.step}.png') - preview_text = entry.cond_text if preview_image_prompt == "" else preview_image_prompt + preview_text = entry[0].cond_text if preview_image_prompt == "" else preview_image_prompt p = processing.StableDiffusionProcessingTxt2Img( sd_model=shared.sd_model, @@ -305,7 +320,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini

Loss: {losses.mean():.7f}
Step: {embedding.step}
-Last prompt: {html.escape(entry.cond_text)}
+Last prompt: {html.escape(entry[-1].cond_text)}
Last saved embedding: {html.escape(last_saved_file)}
Last saved image: {html.escape(last_saved_image)}

diff --git a/modules/txt2img.py b/modules/txt2img.py index e985242b..78342024 100644 --- a/modules/txt2img.py +++ b/modules/txt2img.py @@ -6,7 +6,14 @@ import modules.processing as processing from modules.ui import plaintext_to_html -def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: str, steps: int, sampler_index: int, restore_faces: bool, tiling: bool, n_iter: int, batch_size: int, cfg_scale: float, seed: int, subseed: int, subseed_strength: float, seed_resize_from_h: int, seed_resize_from_w: int, seed_enable_extras: bool, height: int, width: int, enable_hr: bool, scale_latent: bool, denoising_strength: float, *args): +def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: str, steps: int, sampler_index: int, + restore_faces: bool, tiling: bool, n_iter: int, batch_size: int, cfg_scale: float, seed: int, subseed: int, + subseed_strength: float, seed_resize_from_h: int, seed_resize_from_w: int, seed_enable_extras: bool, + height: int, width: int, enable_hr: bool, scale_latent: bool, denoising_strength: float, + aesthetic_lr=0, + aesthetic_weight=0, aesthetic_steps=0, + aesthetic_imgs=None, + aesthetic_slerp=False, *args): p = StableDiffusionProcessingTxt2Img( sd_model=shared.sd_model, outpath_samples=opts.outdir_samples or opts.outdir_txt2img_samples, @@ -40,7 +47,7 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: processed = modules.scripts.scripts_txt2img.run(p, *args) if processed is None: - processed = process_images(p) + processed = process_images(p, aesthetic_lr, aesthetic_weight, aesthetic_steps, aesthetic_imgs, aesthetic_slerp) shared.total_tqdm.clear() diff --git a/modules/ui.py b/modules/ui.py index 220fb80b..d961d126 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -24,7 +24,8 @@ import gradio.routes from modules import sd_hijack from modules.paths import script_path -from modules.shared import opts, cmd_opts +from modules.shared import opts, cmd_opts,aesthetic_embeddings + if cmd_opts.deepdanbooru: from modules.deepbooru import get_deepbooru_tags import modules.shared as shared @@ -534,6 +535,14 @@ def create_ui(wrap_gradio_gpu_call): width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512) height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) + with gr.Group(): + aesthetic_lr = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.005") + aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.7) + aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=50) + + aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Imgs embedding", value=sorted(aesthetic_embeddings.keys())[0] if len(aesthetic_embeddings) > 0 else None) + aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) + with gr.Row(): restore_faces = gr.Checkbox(label='Restore faces', value=False, visible=len(shared.face_restorers) > 1) tiling = gr.Checkbox(label='Tiling', value=False) @@ -586,25 +595,30 @@ def create_ui(wrap_gradio_gpu_call): fn=wrap_gradio_gpu_call(modules.txt2img.txt2img), _js="submit", inputs=[ - txt2img_prompt, - txt2img_negative_prompt, - txt2img_prompt_style, - txt2img_prompt_style2, - steps, - sampler_index, - restore_faces, - tiling, - batch_count, - batch_size, - cfg_scale, - seed, - subseed, subseed_strength, seed_resize_from_h, seed_resize_from_w, seed_checkbox, - height, - width, - enable_hr, - scale_latent, - denoising_strength, - ] + custom_inputs, + txt2img_prompt, + txt2img_negative_prompt, + txt2img_prompt_style, + txt2img_prompt_style2, + steps, + sampler_index, + restore_faces, + tiling, + batch_count, + batch_size, + cfg_scale, + seed, + subseed, subseed_strength, seed_resize_from_h, seed_resize_from_w, seed_checkbox, + height, + width, + enable_hr, + scale_latent, + denoising_strength, + aesthetic_lr, + aesthetic_weight, + aesthetic_steps, + aesthetic_imgs, + aesthetic_slerp + ] + custom_inputs, outputs=[ txt2img_gallery, generation_info, @@ -1097,6 +1111,9 @@ def create_ui(wrap_gradio_gpu_call): template_file = gr.Textbox(label='Prompt template file', value=os.path.join(script_path, "textual_inversion_templates", "style_filewords.txt")) training_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512) training_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) + batch_size = gr.Slider(minimum=1, maximum=64, step=1, label="Batch Size", value=4) + gradient_accumulation = gr.Slider(minimum=1, maximum=256, step=1, label="Gradient accumulation", + value=1) steps = gr.Number(label='Max steps', value=100000, precision=0) create_image_every = gr.Number(label='Save an image to log directory every N steps, 0 to disable', value=500, precision=0) save_embedding_every = gr.Number(label='Save a copy of embedding to log directory every N steps, 0 to disable', value=500, precision=0) @@ -1180,6 +1197,8 @@ def create_ui(wrap_gradio_gpu_call): template_file, save_image_with_stored_embedding, preview_image_prompt, + batch_size, + gradient_accumulation ], outputs=[ ti_output, -- cgit v1.2.3 From 37d7ffb415cd8c69b3c0bb5f61844dde0b169f78 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sat, 15 Oct 2022 15:59:37 +0200 Subject: fix to tokens lenght, addend embs generator, add new features to edit the embedding before the generation using text --- modules/aesthetic_clip.py | 78 ++++++++++++++++++++++++ modules/processing.py | 148 +++++++++++++++++++++++++++++++--------------- modules/sd_hijack.py | 111 ++++++++++++++++++++++------------ modules/shared.py | 4 ++ modules/txt2img.py | 10 +++- modules/ui.py | 47 ++++++++++++--- 6 files changed, 302 insertions(+), 96 deletions(-) create mode 100644 modules/aesthetic_clip.py (limited to 'modules') diff --git a/modules/aesthetic_clip.py b/modules/aesthetic_clip.py new file mode 100644 index 00000000..f15cfd47 --- /dev/null +++ b/modules/aesthetic_clip.py @@ -0,0 +1,78 @@ +import itertools +import os +from pathlib import Path +import html +import gc + +import gradio as gr +import torch +from PIL import Image +from modules import shared +from modules.shared import device, aesthetic_embeddings +from transformers import CLIPModel, CLIPProcessor + +from tqdm.auto import tqdm + + +def get_all_images_in_folder(folder): + return [os.path.join(folder, f) for f in os.listdir(folder) if + os.path.isfile(os.path.join(folder, f)) and check_is_valid_image_file(f)] + + +def check_is_valid_image_file(filename): + return filename.lower().endswith(('.png', '.jpg', '.jpeg')) + + +def batched(dataset, total, n=1): + for ndx in range(0, total, n): + yield [dataset.__getitem__(i) for i in range(ndx, min(ndx + n, total))] + + +def iter_to_batched(iterable, n=1): + it = iter(iterable) + while True: + chunk = tuple(itertools.islice(it, n)) + if not chunk: + return + yield chunk + + +def generate_imgs_embd(name, folder, batch_size): + # clipModel = CLIPModel.from_pretrained( + # shared.sd_model.cond_stage_model.clipModel.name_or_path + # ) + model = CLIPModel.from_pretrained(shared.sd_model.cond_stage_model.clipModel.name_or_path).to(device) + processor = CLIPProcessor.from_pretrained(shared.sd_model.cond_stage_model.clipModel.name_or_path) + + with torch.no_grad(): + embs = [] + for paths in tqdm(iter_to_batched(get_all_images_in_folder(folder), batch_size), + desc=f"Generating embeddings for {name}"): + if shared.state.interrupted: + break + inputs = processor(images=[Image.open(path) for path in paths], return_tensors="pt").to(device) + outputs = model.get_image_features(**inputs).cpu() + embs.append(torch.clone(outputs)) + inputs.to("cpu") + del inputs, outputs + + embs = torch.cat(embs, dim=0).mean(dim=0, keepdim=True) + + # The generated embedding will be located here + path = str(Path(shared.cmd_opts.aesthetic_embeddings_dir) / f"{name}.pt") + torch.save(embs, path) + + model = model.cpu() + del model + del processor + del embs + gc.collect() + torch.cuda.empty_cache() + res = f""" + Done generating embedding for {name}! + Hypernetwork saved to {html.escape(path)} + """ + shared.update_aesthetic_embeddings() + return gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Imgs embedding", + value=sorted(aesthetic_embeddings.keys())[0] if len( + aesthetic_embeddings) > 0 else None), res, "" diff --git a/modules/processing.py b/modules/processing.py index 9a033759..ab68d63a 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -20,7 +20,6 @@ import modules.images as images import modules.styles import logging - # some of those options should not be changed at all because they would break the model, so I removed them from options. opt_C = 4 opt_f = 8 @@ -52,8 +51,13 @@ def get_correct_sampler(p): elif isinstance(p, modules.processing.StableDiffusionProcessingImg2Img): return sd_samplers.samplers_for_img2img + class StableDiffusionProcessing: - def __init__(self, sd_model=None, outpath_samples=None, outpath_grids=None, prompt="", styles=None, seed=-1, subseed=-1, subseed_strength=0, seed_resize_from_h=-1, seed_resize_from_w=-1, seed_enable_extras=True, sampler_index=0, batch_size=1, n_iter=1, steps=50, cfg_scale=7.0, width=512, height=512, restore_faces=False, tiling=False, do_not_save_samples=False, do_not_save_grid=False, extra_generation_params=None, overlay_images=None, negative_prompt=None, eta=None): + def __init__(self, sd_model=None, outpath_samples=None, outpath_grids=None, prompt="", styles=None, seed=-1, + subseed=-1, subseed_strength=0, seed_resize_from_h=-1, seed_resize_from_w=-1, seed_enable_extras=True, + sampler_index=0, batch_size=1, n_iter=1, steps=50, cfg_scale=7.0, width=512, height=512, + restore_faces=False, tiling=False, do_not_save_samples=False, do_not_save_grid=False, + extra_generation_params=None, overlay_images=None, negative_prompt=None, eta=None): self.sd_model = sd_model self.outpath_samples: str = outpath_samples self.outpath_grids: str = outpath_grids @@ -104,7 +108,8 @@ class StableDiffusionProcessing: class Processed: - def __init__(self, p: StableDiffusionProcessing, images_list, seed=-1, info="", subseed=None, all_prompts=None, all_seeds=None, all_subseeds=None, index_of_first_image=0, infotexts=None): + def __init__(self, p: StableDiffusionProcessing, images_list, seed=-1, info="", subseed=None, all_prompts=None, + all_seeds=None, all_subseeds=None, index_of_first_image=0, infotexts=None): self.images = images_list self.prompt = p.prompt self.negative_prompt = p.negative_prompt @@ -141,7 +146,8 @@ class Processed: self.prompt = self.prompt if type(self.prompt) != list else self.prompt[0] self.negative_prompt = self.negative_prompt if type(self.negative_prompt) != list else self.negative_prompt[0] self.seed = int(self.seed if type(self.seed) != list else self.seed[0]) - self.subseed = int(self.subseed if type(self.subseed) != list else self.subseed[0]) if self.subseed is not None else -1 + self.subseed = int( + self.subseed if type(self.subseed) != list else self.subseed[0]) if self.subseed is not None else -1 self.all_prompts = all_prompts or [self.prompt] self.all_seeds = all_seeds or [self.seed] @@ -181,39 +187,43 @@ class Processed: return json.dumps(obj) - def infotext(self, p: StableDiffusionProcessing, index): - return create_infotext(p, self.all_prompts, self.all_seeds, self.all_subseeds, comments=[], position_in_batch=index % self.batch_size, iteration=index // self.batch_size) + def infotext(self, p: StableDiffusionProcessing, index): + return create_infotext(p, self.all_prompts, self.all_seeds, self.all_subseeds, comments=[], + position_in_batch=index % self.batch_size, iteration=index // self.batch_size) # from https://discuss.pytorch.org/t/help-regarding-slerp-function-for-generative-model-sampling/32475/3 def slerp(val, low, high): - low_norm = low/torch.norm(low, dim=1, keepdim=True) - high_norm = high/torch.norm(high, dim=1, keepdim=True) - dot = (low_norm*high_norm).sum(1) + low_norm = low / torch.norm(low, dim=1, keepdim=True) + high_norm = high / torch.norm(high, dim=1, keepdim=True) + dot = (low_norm * high_norm).sum(1) if dot.mean() > 0.9995: return low * val + high * (1 - val) omega = torch.acos(dot) so = torch.sin(omega) - res = (torch.sin((1.0-val)*omega)/so).unsqueeze(1)*low + (torch.sin(val*omega)/so).unsqueeze(1) * high + res = (torch.sin((1.0 - val) * omega) / so).unsqueeze(1) * low + (torch.sin(val * omega) / so).unsqueeze(1) * high return res -def create_random_tensors(shape, seeds, subseeds=None, subseed_strength=0.0, seed_resize_from_h=0, seed_resize_from_w=0, p=None): +def create_random_tensors(shape, seeds, subseeds=None, subseed_strength=0.0, seed_resize_from_h=0, seed_resize_from_w=0, + p=None): xs = [] # if we have multiple seeds, this means we are working with batch size>1; this then # enables the generation of additional tensors with noise that the sampler will use during its processing. # Using those pre-generated tensors instead of simple torch.randn allows a batch with seeds [100, 101] to # produce the same images as with two batches [100], [101]. - if p is not None and p.sampler is not None and (len(seeds) > 1 and opts.enable_batch_seeds or opts.eta_noise_seed_delta > 0): + if p is not None and p.sampler is not None and ( + len(seeds) > 1 and opts.enable_batch_seeds or opts.eta_noise_seed_delta > 0): sampler_noises = [[] for _ in range(p.sampler.number_of_needed_noises(p))] else: sampler_noises = None for i, seed in enumerate(seeds): - noise_shape = shape if seed_resize_from_h <= 0 or seed_resize_from_w <= 0 else (shape[0], seed_resize_from_h//8, seed_resize_from_w//8) + noise_shape = shape if seed_resize_from_h <= 0 or seed_resize_from_w <= 0 else ( + shape[0], seed_resize_from_h // 8, seed_resize_from_w // 8) subnoise = None if subseeds is not None: @@ -241,7 +251,7 @@ def create_random_tensors(shape, seeds, subseeds=None, subseed_strength=0.0, see dx = max(-dx, 0) dy = max(-dy, 0) - x[:, ty:ty+h, tx:tx+w] = noise[:, dy:dy+h, dx:dx+w] + x[:, ty:ty + h, tx:tx + w] = noise[:, dy:dy + h, dx:dx + w] noise = x if sampler_noises is not None: @@ -293,14 +303,20 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration "Seed": all_seeds[index], "Face restoration": (opts.face_restoration_model if p.restore_faces else None), "Size": f"{p.width}x{p.height}", - "Model hash": getattr(p, 'sd_model_hash', None if not opts.add_model_hash_to_info or not shared.sd_model.sd_model_hash else shared.sd_model.sd_model_hash), - "Model": (None if not opts.add_model_name_to_info or not shared.sd_model.sd_checkpoint_info.model_name else shared.sd_model.sd_checkpoint_info.model_name.replace(',', '').replace(':', '')), - "Hypernet": (None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.name.replace(',', '').replace(':', '')), + "Model hash": getattr(p, 'sd_model_hash', + None if not opts.add_model_hash_to_info or not shared.sd_model.sd_model_hash else shared.sd_model.sd_model_hash), + "Model": ( + None if not opts.add_model_name_to_info or not shared.sd_model.sd_checkpoint_info.model_name else shared.sd_model.sd_checkpoint_info.model_name.replace( + ',', '').replace(':', '')), + "Hypernet": ( + None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.name.replace(',', '').replace( + ':', '')), "Batch size": (None if p.batch_size < 2 else p.batch_size), "Batch pos": (None if p.batch_size < 2 else position_in_batch), "Variation seed": (None if p.subseed_strength == 0 else all_subseeds[index]), "Variation seed strength": (None if p.subseed_strength == 0 else p.subseed_strength), - "Seed resize from": (None if p.seed_resize_from_w == 0 or p.seed_resize_from_h == 0 else f"{p.seed_resize_from_w}x{p.seed_resize_from_h}"), + "Seed resize from": ( + None if p.seed_resize_from_w == 0 or p.seed_resize_from_h == 0 else f"{p.seed_resize_from_w}x{p.seed_resize_from_h}"), "Denoising strength": getattr(p, 'denoising_strength', None), "Eta": (None if p.sampler is None or p.sampler.eta == p.sampler.default_eta else p.sampler.eta), "Clip skip": None if clip_skip <= 1 else clip_skip, @@ -309,7 +325,8 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration generation_params.update(p.extra_generation_params) - generation_params_text = ", ".join([k if k == v else f'{k}: {v}' for k, v in generation_params.items() if v is not None]) + generation_params_text = ", ".join( + [k if k == v else f'{k}: {v}' for k, v in generation_params.items() if v is not None]) negative_prompt_text = "\nNegative prompt: " + p.negative_prompt if p.negative_prompt else "" @@ -317,7 +334,9 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, - aesthetic_imgs=None,aesthetic_slerp=False) -> Processed: + aesthetic_imgs=None, aesthetic_slerp=False, aesthetic_imgs_text="", + aesthetic_slerp_angle=0.15, + aesthetic_text_negative=False) -> Processed: """this is the main loop that both txt2img and img2img use; it calls func_init once inside all the scopes and func_sample once per batch""" aesthetic_lr = float(aesthetic_lr) @@ -385,7 +404,7 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh for n in range(p.n_iter): if state.skipped: state.skipped = False - + if state.interrupted: break @@ -396,16 +415,19 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh if (len(prompts) == 0): break - #uc = p.sd_model.get_learned_conditioning(len(prompts) * [p.negative_prompt]) - #c = p.sd_model.get_learned_conditioning(prompts) + # uc = p.sd_model.get_learned_conditioning(len(prompts) * [p.negative_prompt]) + # c = p.sd_model.get_learned_conditioning(prompts) with devices.autocast(): if hasattr(shared.sd_model.cond_stage_model, "set_aesthetic_params"): - shared.sd_model.cond_stage_model.set_aesthetic_params(0, 0, 0) + shared.sd_model.cond_stage_model.set_aesthetic_params() uc = prompt_parser.get_learned_conditioning(shared.sd_model, len(prompts) * [p.negative_prompt], p.steps) if hasattr(shared.sd_model.cond_stage_model, "set_aesthetic_params"): shared.sd_model.cond_stage_model.set_aesthetic_params(aesthetic_lr, aesthetic_weight, - aesthetic_steps, aesthetic_imgs,aesthetic_slerp) + aesthetic_steps, aesthetic_imgs, + aesthetic_slerp, aesthetic_imgs_text, + aesthetic_slerp_angle, + aesthetic_text_negative) c = prompt_parser.get_multicond_learned_conditioning(shared.sd_model, prompts, p.steps) if len(model_hijack.comments) > 0: @@ -413,13 +435,13 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh comments[comment] = 1 if p.n_iter > 1: - shared.state.job = f"Batch {n+1} out of {p.n_iter}" + shared.state.job = f"Batch {n + 1} out of {p.n_iter}" with devices.autocast(): - samples_ddim = p.sample(conditioning=c, unconditional_conditioning=uc, seeds=seeds, subseeds=subseeds, subseed_strength=p.subseed_strength) + samples_ddim = p.sample(conditioning=c, unconditional_conditioning=uc, seeds=seeds, subseeds=subseeds, + subseed_strength=p.subseed_strength) if state.interrupted or state.skipped: - # if we are interrupted, sample returns just noise # use the image collected previously in sampler loop samples_ddim = shared.state.current_latent @@ -445,7 +467,9 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh if p.restore_faces: if opts.save and not p.do_not_save_samples and opts.save_images_before_face_restoration: - images.save_image(Image.fromarray(x_sample), p.outpath_samples, "", seeds[i], prompts[i], opts.samples_format, info=infotext(n, i), p=p, suffix="-before-face-restoration") + images.save_image(Image.fromarray(x_sample), p.outpath_samples, "", seeds[i], prompts[i], + opts.samples_format, info=infotext(n, i), p=p, + suffix="-before-face-restoration") devices.torch_gc() @@ -456,7 +480,8 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh if p.color_corrections is not None and i < len(p.color_corrections): if opts.save and not p.do_not_save_samples and opts.save_images_before_color_correction: - images.save_image(image, p.outpath_samples, "", seeds[i], prompts[i], opts.samples_format, info=infotext(n, i), p=p, suffix="-before-color-correction") + images.save_image(image, p.outpath_samples, "", seeds[i], prompts[i], opts.samples_format, + info=infotext(n, i), p=p, suffix="-before-color-correction") image = apply_color_correction(p.color_corrections[i], image) if p.overlay_images is not None and i < len(p.overlay_images): @@ -474,7 +499,8 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh image = image.convert('RGB') if opts.samples_save and not p.do_not_save_samples: - images.save_image(image, p.outpath_samples, "", seeds[i], prompts[i], opts.samples_format, info=infotext(n, i), p=p) + images.save_image(image, p.outpath_samples, "", seeds[i], prompts[i], opts.samples_format, + info=infotext(n, i), p=p) text = infotext(n, i) infotexts.append(text) @@ -482,7 +508,7 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh image.info["parameters"] = text output_images.append(image) - del x_samples_ddim + del x_samples_ddim devices.torch_gc() @@ -504,10 +530,13 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh index_of_first_image = 1 if opts.grid_save: - images.save_image(grid, p.outpath_grids, "grid", all_seeds[0], all_prompts[0], opts.grid_format, info=infotext(), short_filename=not opts.grid_extended_filename, p=p, grid=True) + images.save_image(grid, p.outpath_grids, "grid", all_seeds[0], all_prompts[0], opts.grid_format, + info=infotext(), short_filename=not opts.grid_extended_filename, p=p, grid=True) devices.torch_gc() - return Processed(p, output_images, all_seeds[0], infotext() + "".join(["\n\n" + x for x in comments]), subseed=all_subseeds[0], all_prompts=all_prompts, all_seeds=all_seeds, all_subseeds=all_subseeds, index_of_first_image=index_of_first_image, infotexts=infotexts) + return Processed(p, output_images, all_seeds[0], infotext() + "".join(["\n\n" + x for x in comments]), + subseed=all_subseeds[0], all_prompts=all_prompts, all_seeds=all_seeds, all_subseeds=all_subseeds, + index_of_first_image=index_of_first_image, infotexts=infotexts) class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): @@ -543,25 +572,34 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model) if not self.enable_hr: - x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) + x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, + subseeds=subseeds, subseed_strength=self.subseed_strength, + seed_resize_from_h=self.seed_resize_from_h, + seed_resize_from_w=self.seed_resize_from_w, p=self) samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning) return samples - x = create_random_tensors([opt_C, self.firstphase_height // opt_f, self.firstphase_width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) + x = create_random_tensors([opt_C, self.firstphase_height // opt_f, self.firstphase_width // opt_f], seeds=seeds, + subseeds=subseeds, subseed_strength=self.subseed_strength, + seed_resize_from_h=self.seed_resize_from_h, + seed_resize_from_w=self.seed_resize_from_w, p=self) samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning) truncate_x = (self.firstphase_width - self.firstphase_width_truncated) // opt_f truncate_y = (self.firstphase_height - self.firstphase_height_truncated) // opt_f - samples = samples[:, :, truncate_y//2:samples.shape[2]-truncate_y//2, truncate_x//2:samples.shape[3]-truncate_x//2] + samples = samples[:, :, truncate_y // 2:samples.shape[2] - truncate_y // 2, + truncate_x // 2:samples.shape[3] - truncate_x // 2] if self.scale_latent: - samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), mode="bilinear") + samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), + mode="bilinear") else: decoded_samples = decode_first_stage(self.sd_model, samples) if opts.upscaler_for_img2img is None or opts.upscaler_for_img2img == "None": - decoded_samples = torch.nn.functional.interpolate(decoded_samples, size=(self.height, self.width), mode="bilinear") + decoded_samples = torch.nn.functional.interpolate(decoded_samples, size=(self.height, self.width), + mode="bilinear") else: lowres_samples = torch.clamp((decoded_samples + 1.0) / 2.0, min=0.0, max=1.0) @@ -585,13 +623,16 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model) - noise = create_random_tensors(samples.shape[1:], seeds=seeds, subseeds=subseeds, subseed_strength=subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) + noise = create_random_tensors(samples.shape[1:], seeds=seeds, subseeds=subseeds, + subseed_strength=subseed_strength, seed_resize_from_h=self.seed_resize_from_h, + seed_resize_from_w=self.seed_resize_from_w, p=self) # GC now before running the next img2img to prevent running out of memory x = None devices.torch_gc() - samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps) + samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, + steps=self.steps) return samples @@ -599,7 +640,9 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): sampler = None - def __init__(self, init_images=None, resize_mode=0, denoising_strength=0.75, mask=None, mask_blur=4, inpainting_fill=0, inpaint_full_res=True, inpaint_full_res_padding=0, inpainting_mask_invert=0, **kwargs): + def __init__(self, init_images=None, resize_mode=0, denoising_strength=0.75, mask=None, mask_blur=4, + inpainting_fill=0, inpaint_full_res=True, inpaint_full_res_padding=0, inpainting_mask_invert=0, + **kwargs): super().__init__(**kwargs) self.init_images = init_images @@ -607,7 +650,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): self.denoising_strength: float = denoising_strength self.init_latent = None self.image_mask = mask - #self.image_unblurred_mask = None + # self.image_unblurred_mask = None self.latent_mask = None self.mask_for_overlay = None self.mask_blur = mask_blur @@ -619,7 +662,8 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): self.nmask = None def init(self, all_prompts, all_seeds, all_subseeds): - self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers_for_img2img, self.sampler_index, self.sd_model) + self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers_for_img2img, self.sampler_index, + self.sd_model) crop_region = None if self.image_mask is not None: @@ -628,7 +672,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): if self.inpainting_mask_invert: self.image_mask = ImageOps.invert(self.image_mask) - #self.image_unblurred_mask = self.image_mask + # self.image_unblurred_mask = self.image_mask if self.mask_blur > 0: self.image_mask = self.image_mask.filter(ImageFilter.GaussianBlur(self.mask_blur)) @@ -642,7 +686,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): mask = mask.crop(crop_region) self.image_mask = images.resize_image(2, mask, self.width, self.height) - self.paste_to = (x1, y1, x2-x1, y2-y1) + self.paste_to = (x1, y1, x2 - x1, y2 - y1) else: self.image_mask = images.resize_image(self.resize_mode, self.image_mask, self.width, self.height) np_mask = np.array(self.image_mask) @@ -665,7 +709,8 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): if self.image_mask is not None: image_masked = Image.new('RGBa', (image.width, image.height)) - image_masked.paste(image.convert("RGBA").convert("RGBa"), mask=ImageOps.invert(self.mask_for_overlay.convert('L'))) + image_masked.paste(image.convert("RGBA").convert("RGBa"), + mask=ImageOps.invert(self.mask_for_overlay.convert('L'))) self.overlay_images.append(image_masked.convert('RGBA')) @@ -714,12 +759,17 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): # this needs to be fixed to be done in sample() using actual seeds for batches if self.inpainting_fill == 2: - self.init_latent = self.init_latent * self.mask + create_random_tensors(self.init_latent.shape[1:], all_seeds[0:self.init_latent.shape[0]]) * self.nmask + self.init_latent = self.init_latent * self.mask + create_random_tensors(self.init_latent.shape[1:], + all_seeds[ + 0:self.init_latent.shape[ + 0]]) * self.nmask elif self.inpainting_fill == 3: self.init_latent = self.init_latent * self.mask def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): - x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) + x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, + subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, + seed_resize_from_w=self.seed_resize_from_w, p=self) samples = self.sampler.sample_img2img(self, self.init_latent, x, conditioning, unconditional_conditioning) diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 6d5196fe..192883b2 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -14,7 +14,8 @@ from modules.sd_hijack_optimizations import invokeAI_mps_available import ldm.modules.attention import ldm.modules.diffusionmodules.model -from transformers import CLIPVisionModel, CLIPModel +from tqdm import trange +from transformers import CLIPVisionModel, CLIPModel, CLIPTokenizer import torch.optim as optim import copy @@ -22,21 +23,25 @@ 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 + def apply_optimizations(): undo_optimizations() ldm.modules.diffusionmodules.model.nonlinearity = silu - if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and (6, 0) <= torch.cuda.get_device_capability(shared.device) <= (8, 6)): + if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and ( + 6, 0) <= torch.cuda.get_device_capability(shared.device) <= (8, 6)): print("Applying xformers cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.xformers_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.xformers_attnblock_forward elif cmd_opts.opt_split_attention_v1: print("Applying v1 cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.split_cross_attention_forward_v1 - elif not cmd_opts.disable_opt_split_attention and (cmd_opts.opt_split_attention_invokeai or not torch.cuda.is_available()): + elif not cmd_opts.disable_opt_split_attention and ( + cmd_opts.opt_split_attention_invokeai or not torch.cuda.is_available()): if not invokeAI_mps_available and shared.device.type == 'mps': - print("The InvokeAI cross attention optimization for MPS requires the psutil package which is not installed.") + print( + "The InvokeAI cross attention optimization for MPS requires the psutil package which is not installed.") print("Applying v1 cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.split_cross_attention_forward_v1 else: @@ -112,14 +117,16 @@ class StableDiffusionModelHijack: _, remade_batch_tokens, _, _, _, token_count = self.clip.process_text([text]) return remade_batch_tokens[0], token_count, get_target_prompt_token_count(token_count) + def slerp(low, high, val): - low_norm = low/torch.norm(low, dim=1, keepdim=True) - high_norm = high/torch.norm(high, dim=1, keepdim=True) - omega = torch.acos((low_norm*high_norm).sum(1)) + low_norm = low / torch.norm(low, dim=1, keepdim=True) + high_norm = high / torch.norm(high, dim=1, keepdim=True) + omega = torch.acos((low_norm * high_norm).sum(1)) so = torch.sin(omega) - res = (torch.sin((1.0-val)*omega)/so).unsqueeze(1)*low + (torch.sin(val*omega)/so).unsqueeze(1) * high + res = (torch.sin((1.0 - val) * omega) / so).unsqueeze(1) * low + (torch.sin(val * omega) / so).unsqueeze(1) * high return res + class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): def __init__(self, wrapped, hijack): super().__init__() @@ -128,6 +135,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): self.wrapped.transformer.name_or_path ) del self.clipModel.vision_model + self.tokenizer = CLIPTokenizer.from_pretrained(self.wrapped.transformer.name_or_path) self.hijack: StableDiffusionModelHijack = hijack self.tokenizer = wrapped.tokenizer # self.vision = CLIPVisionModel.from_pretrained(self.wrapped.transformer.name_or_path).eval() @@ -139,7 +147,8 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): self.comma_token = [v for k, v in self.tokenizer.get_vocab().items() if k == ','][0] - tokens_with_parens = [(k, v) for k, v in self.tokenizer.get_vocab().items() if '(' in k or ')' in k or '[' in k or ']' in k] + tokens_with_parens = [(k, v) for k, v in self.tokenizer.get_vocab().items() if + '(' in k or ')' in k or '[' in k or ']' in k] for text, ident in tokens_with_parens: mult = 1.0 for c in text: @@ -155,8 +164,13 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): if mult != 1.0: self.token_mults[ident] = mult - def set_aesthetic_params(self, aesthetic_lr, aesthetic_weight, aesthetic_steps, image_embs_name=None, - aesthetic_slerp=True): + def set_aesthetic_params(self, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, image_embs_name=None, + aesthetic_slerp=True, aesthetic_imgs_text="", + aesthetic_slerp_angle=0.15, + aesthetic_text_negative=False): + self.aesthetic_imgs_text = aesthetic_imgs_text + self.aesthetic_slerp_angle = aesthetic_slerp_angle + self.aesthetic_text_negative = aesthetic_text_negative self.slerp = aesthetic_slerp self.aesthetic_lr = aesthetic_lr self.aesthetic_weight = aesthetic_weight @@ -180,7 +194,8 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): else: parsed = [[line, 1.0]] - tokenized = self.wrapped.tokenizer([text for text, _ in parsed], truncation=False, add_special_tokens=False)["input_ids"] + tokenized = self.wrapped.tokenizer([text for text, _ in parsed], truncation=False, add_special_tokens=False)[ + "input_ids"] fixes = [] remade_tokens = [] @@ -196,18 +211,20 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): if token == self.comma_token: last_comma = len(remade_tokens) - elif opts.comma_padding_backtrack != 0 and max(len(remade_tokens), 1) % 75 == 0 and last_comma != -1 and len(remade_tokens) - last_comma <= opts.comma_padding_backtrack: + elif opts.comma_padding_backtrack != 0 and max(len(remade_tokens), + 1) % 75 == 0 and last_comma != -1 and len( + remade_tokens) - last_comma <= opts.comma_padding_backtrack: last_comma += 1 reloc_tokens = remade_tokens[last_comma:] reloc_mults = multipliers[last_comma:] remade_tokens = remade_tokens[:last_comma] length = len(remade_tokens) - + rem = int(math.ceil(length / 75)) * 75 - length remade_tokens += [id_end] * rem + reloc_tokens multipliers = multipliers[:last_comma] + [1.0] * rem + reloc_mults - + if embedding is None: remade_tokens.append(token) multipliers.append(weight) @@ -248,7 +265,8 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): if line in cache: remade_tokens, fixes, multipliers = cache[line] else: - remade_tokens, fixes, multipliers, current_token_count = self.tokenize_line(line, used_custom_terms, hijack_comments) + remade_tokens, fixes, multipliers, current_token_count = self.tokenize_line(line, used_custom_terms, + hijack_comments) token_count = max(current_token_count, token_count) cache[line] = (remade_tokens, fixes, multipliers) @@ -259,7 +277,6 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): return batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count - def process_text_old(self, text): id_start = self.wrapped.tokenizer.bos_token_id id_end = self.wrapped.tokenizer.eos_token_id @@ -289,7 +306,8 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): while i < len(tokens): token = tokens[i] - embedding, embedding_length_in_tokens = self.hijack.embedding_db.find_embedding_at_position(tokens, i) + embedding, embedding_length_in_tokens = self.hijack.embedding_db.find_embedding_at_position(tokens, + i) mult_change = self.token_mults.get(token) if opts.enable_emphasis else None if mult_change is not None: @@ -312,11 +330,12 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): ovf = remade_tokens[maxlen - 2:] overflowing_words = [vocab.get(int(x), "") for x in ovf] overflowing_text = self.wrapped.tokenizer.convert_tokens_to_string(''.join(overflowing_words)) - hijack_comments.append(f"Warning: too many input tokens; some ({len(overflowing_words)}) have been truncated:\n{overflowing_text}\n") + hijack_comments.append( + f"Warning: too many input tokens; some ({len(overflowing_words)}) have been truncated:\n{overflowing_text}\n") token_count = len(remade_tokens) remade_tokens = remade_tokens + [id_end] * (maxlen - 2 - len(remade_tokens)) - remade_tokens = [id_start] + remade_tokens[0:maxlen-2] + [id_end] + remade_tokens = [id_start] + remade_tokens[0:maxlen - 2] + [id_end] cache[tuple_tokens] = (remade_tokens, fixes, multipliers) multipliers = multipliers + [1.0] * (maxlen - 2 - len(multipliers)) @@ -326,23 +345,26 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): hijack_fixes.append(fixes) batch_multipliers.append(multipliers) return batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count - + def forward(self, text): use_old = opts.use_old_emphasis_implementation if use_old: - batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count = self.process_text_old(text) + batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count = self.process_text_old( + text) else: - batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count = self.process_text(text) + batch_multipliers, remade_batch_tokens, used_custom_terms, hijack_comments, hijack_fixes, token_count = self.process_text( + text) self.hijack.comments += hijack_comments if len(used_custom_terms) > 0: - self.hijack.comments.append("Used embeddings: " + ", ".join([f'{word} [{checksum}]' for word, checksum in used_custom_terms])) - + self.hijack.comments.append( + "Used embeddings: " + ", ".join([f'{word} [{checksum}]' for word, checksum in used_custom_terms])) + if use_old: self.hijack.fixes = hijack_fixes return self.process_tokens(remade_batch_tokens, batch_multipliers) - + z = None i = 0 while max(map(len, remade_batch_tokens)) != 0: @@ -356,7 +378,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): if fix[0] == i: fixes.append(fix[1]) self.hijack.fixes.append(fixes) - + tokens = [] multipliers = [] for j in range(len(remade_batch_tokens)): @@ -378,19 +400,30 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): remade_batch_tokens] tokens = torch.asarray(remade_batch_tokens).to(device) + + model = copy.deepcopy(self.clipModel).to(device) + model.requires_grad_(True) + if self.aesthetic_imgs_text is not None and len(self.aesthetic_imgs_text) > 0: + text_embs_2 = model.get_text_features( + **self.tokenizer([self.aesthetic_imgs_text], padding=True, return_tensors="pt").to(device)) + if self.aesthetic_text_negative: + text_embs_2 = self.image_embs - text_embs_2 + text_embs_2 /= text_embs_2.norm(dim=-1, keepdim=True) + img_embs = slerp(self.image_embs, text_embs_2, self.aesthetic_slerp_angle) + else: + img_embs = self.image_embs + with torch.enable_grad(): - model = copy.deepcopy(self.clipModel).to(device) - model.requires_grad_(True) # We optimize the model to maximize the similarity optimizer = optim.Adam( model.text_model.parameters(), lr=self.aesthetic_lr ) - for i in range(self.aesthetic_steps): + for i in trange(self.aesthetic_steps, desc="Aesthetic optimization"): text_embs = model.get_text_features(input_ids=tokens) text_embs = text_embs / text_embs.norm(dim=-1, keepdim=True) - sim = text_embs @ self.image_embs.T + sim = text_embs @ img_embs.T loss = -sim optimizer.zero_grad() loss.mean().backward() @@ -405,6 +438,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): model.cpu() del model + zn = torch.concat([zn for i in range(z.shape[1] // 77)], 1) if self.slerp: z = slerp(z, zn, self.aesthetic_weight) else: @@ -413,15 +447,16 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): remade_batch_tokens = rem_tokens batch_multipliers = rem_multipliers i += 1 - + return z - - + def process_tokens(self, remade_batch_tokens, batch_multipliers): if not opts.use_old_emphasis_implementation: - remade_batch_tokens = [[self.wrapped.tokenizer.bos_token_id] + x[:75] + [self.wrapped.tokenizer.eos_token_id] for x in remade_batch_tokens] + remade_batch_tokens = [ + [self.wrapped.tokenizer.bos_token_id] + x[:75] + [self.wrapped.tokenizer.eos_token_id] for x in + remade_batch_tokens] batch_multipliers = [[1.0] + x[:75] + [1.0] for x in batch_multipliers] - + tokens = torch.asarray(remade_batch_tokens).to(device) outputs = self.wrapped.transformer(input_ids=tokens, output_hidden_states=-opts.CLIP_stop_at_last_layers) @@ -461,8 +496,8 @@ class EmbeddingsWithFixes(torch.nn.Module): for fixes, tensor in zip(batch_fixes, inputs_embeds): for offset, embedding in fixes: emb = embedding.vec - emb_len = min(tensor.shape[0]-offset-1, emb.shape[0]) - tensor = torch.cat([tensor[0:offset+1], emb[0:emb_len], tensor[offset+1+emb_len:]]) + emb_len = min(tensor.shape[0] - offset - 1, emb.shape[0]) + tensor = torch.cat([tensor[0:offset + 1], emb[0:emb_len], tensor[offset + 1 + emb_len:]]) vecs.append(tensor) diff --git a/modules/shared.py b/modules/shared.py index cf13a10d..7cd608ca 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -95,6 +95,10 @@ loaded_hypernetwork = None aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} +def update_aesthetic_embeddings(): + global aesthetic_embeddings + aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in + os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} def reload_hypernetworks(): global hypernetworks diff --git a/modules/txt2img.py b/modules/txt2img.py index 78342024..eedcdfe0 100644 --- a/modules/txt2img.py +++ b/modules/txt2img.py @@ -13,7 +13,11 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, aesthetic_imgs=None, - aesthetic_slerp=False, *args): + aesthetic_slerp=False, + aesthetic_imgs_text="", + aesthetic_slerp_angle=0.15, + aesthetic_text_negative=False, + *args): p = StableDiffusionProcessingTxt2Img( sd_model=shared.sd_model, outpath_samples=opts.outdir_samples or opts.outdir_txt2img_samples, @@ -47,7 +51,9 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: processed = modules.scripts.scripts_txt2img.run(p, *args) if processed is None: - processed = process_images(p, aesthetic_lr, aesthetic_weight, aesthetic_steps, aesthetic_imgs, aesthetic_slerp) + processed = process_images(p, aesthetic_lr, aesthetic_weight, aesthetic_steps, aesthetic_imgs, aesthetic_slerp,aesthetic_imgs_text, + aesthetic_slerp_angle, + aesthetic_text_negative) shared.total_tqdm.clear() diff --git a/modules/ui.py b/modules/ui.py index d961d126..e98e2113 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -41,6 +41,7 @@ from modules import prompt_parser from modules.images import save_image import modules.textual_inversion.ui import modules.hypernetworks.ui +import modules.aesthetic_clip # this is a fix for Windows users. Without it, javascript files will be served with text/html content-type and the browser will not show any UI mimetypes.init() @@ -449,7 +450,7 @@ def create_toprow(is_img2img): with gr.Row(): negative_prompt = gr.Textbox(label="Negative prompt", elem_id="negative_prompt", show_label=False, placeholder="Negative prompt", lines=2) with gr.Column(scale=1, elem_id="roll_col"): - sh = gr.Button(elem_id="sh", visible=True) + sh = gr.Button(elem_id="sh", visible=True) with gr.Column(scale=1, elem_id="style_neg_col"): prompt_style2 = gr.Dropdown(label="Style 2", elem_id=f"{id_part}_style2_index", choices=[k for k, v in shared.prompt_styles.styles.items()], value=next(iter(shared.prompt_styles.styles.keys())), visible=len(shared.prompt_styles.styles) > 1) @@ -536,9 +537,13 @@ def create_ui(wrap_gradio_gpu_call): height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) with gr.Group(): - aesthetic_lr = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.005") - aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.7) - aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=50) + aesthetic_lr = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.0001") + aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) + aesthetic_steps = gr.Slider(minimum=0, maximum=256, step=1, label="Aesthetic steps", value=5) + with gr.Row(): + aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", value="") + aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) + aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Imgs embedding", value=sorted(aesthetic_embeddings.keys())[0] if len(aesthetic_embeddings) > 0 else None) aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) @@ -617,7 +622,10 @@ def create_ui(wrap_gradio_gpu_call): aesthetic_weight, aesthetic_steps, aesthetic_imgs, - aesthetic_slerp + aesthetic_slerp, + aesthetic_imgs_text, + aesthetic_slerp_angle, + aesthetic_text_negative ] + custom_inputs, outputs=[ txt2img_gallery, @@ -721,7 +729,7 @@ def create_ui(wrap_gradio_gpu_call): with gr.Row(): inpaint_full_res = gr.Checkbox(label='Inpaint at full resolution', value=False) - inpaint_full_res_padding = gr.Slider(label='Inpaint at full resolution padding, pixels', minimum=0, maximum=256, step=4, value=32) + inpaint_full_res_padding = gr.Slider(label='Inpaint at full resolution padding, pixels', minimum=0, maximum=1024, step=4, value=32) with gr.TabItem('Batch img2img', id='batch'): hidden = '
Disabled when launched with --hide-ui-dir-config.' if shared.cmd_opts.hide_ui_dir_config else '' @@ -1071,6 +1079,17 @@ def create_ui(wrap_gradio_gpu_call): with gr.Column(): create_embedding = gr.Button(value="Create embedding", variant='primary') + with gr.Tab(label="Create images embedding"): + new_embedding_name_ae = gr.Textbox(label="Name") + process_src_ae = gr.Textbox(label='Source directory') + batch_ae = gr.Slider(minimum=1, maximum=1024, step=1, label="Batch size", value=256) + with gr.Row(): + with gr.Column(scale=3): + gr.HTML(value="") + + with gr.Column(): + create_embedding_ae = gr.Button(value="Create images embedding", variant='primary') + with gr.Tab(label="Create hypernetwork"): new_hypernetwork_name = gr.Textbox(label="Name") new_hypernetwork_sizes = gr.CheckboxGroup(label="Modules", value=["768", "320", "640", "1280"], choices=["768", "320", "640", "1280"]) @@ -1139,7 +1158,7 @@ def create_ui(wrap_gradio_gpu_call): fn=modules.textual_inversion.ui.create_embedding, inputs=[ new_embedding_name, - initialization_text, + process_src, nvpt, ], outputs=[ @@ -1149,6 +1168,20 @@ def create_ui(wrap_gradio_gpu_call): ] ) + create_embedding_ae.click( + fn=modules.aesthetic_clip.generate_imgs_embd, + inputs=[ + new_embedding_name_ae, + process_src_ae, + batch_ae + ], + outputs=[ + aesthetic_imgs, + ti_output, + ti_outcome, + ] + ) + create_hypernetwork.click( fn=modules.hypernetworks.ui.create_hypernetwork, inputs=[ -- cgit v1.2.3 From 4387e4fe6479c08f7bc7e42924c3a1093e3a1872 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sat, 15 Oct 2022 18:39:29 +0200 Subject: Update modules/ui.py MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: Víctor Gallego --- modules/ui.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index d0696101..5bb961b2 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -599,7 +599,8 @@ def create_ui(wrap_gradio_gpu_call): with gr.Group(): aesthetic_lr = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.0001") aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) - aesthetic_steps = gr.Slider(minimum=0, maximum=256, step=1, label="Aesthetic steps", value=5) + aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) + with gr.Row(): aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", value="") aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) -- cgit v1.2.3 From 9b7705e0573bddde26df4575c71f994d73a4d519 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sat, 15 Oct 2022 18:40:34 +0200 Subject: Update modules/aesthetic_clip.py MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: Víctor Gallego --- modules/aesthetic_clip.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/aesthetic_clip.py b/modules/aesthetic_clip.py index f15cfd47..bcf2b073 100644 --- a/modules/aesthetic_clip.py +++ b/modules/aesthetic_clip.py @@ -70,7 +70,7 @@ def generate_imgs_embd(name, folder, batch_size): torch.cuda.empty_cache() res = f""" Done generating embedding for {name}! - Hypernetwork saved to {html.escape(path)} + Aesthetic embedding saved to {html.escape(path)} """ shared.update_aesthetic_embeddings() return gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Imgs embedding", -- cgit v1.2.3 From 0d4f5db235357aeb4c7a8738179ba33aaf5a6b75 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sat, 15 Oct 2022 18:40:58 +0200 Subject: Update modules/ui.py MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: Víctor Gallego --- modules/ui.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 5bb961b2..25eba548 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -597,7 +597,8 @@ def create_ui(wrap_gradio_gpu_call): height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) with gr.Group(): - aesthetic_lr = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.0001") + aesthetic_lr = gr.Textbox(label='Aesthetic learning rate', placeholder="Aesthetic learning rate", value="0.0001") + aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) -- cgit v1.2.3 From ad9bc604a8fadcfebe72be37f66cec51e7e87fb5 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sat, 15 Oct 2022 18:41:18 +0200 Subject: Update modules/ui.py MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: Víctor Gallego --- modules/ui.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 25eba548..3b28b69c 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -607,7 +607,8 @@ def create_ui(wrap_gradio_gpu_call): aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) - aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Imgs embedding", value=sorted(aesthetic_embeddings.keys())[0] if len(aesthetic_embeddings) > 0 else None) + aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Aesthetic imgs embedding", value=sorted(aesthetic_embeddings.keys())[0] if len(aesthetic_embeddings) > 0 else None) + aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) with gr.Row(): -- cgit v1.2.3 From 3f5c3b981e46c16bb10948d012575b25170efb3b Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sat, 15 Oct 2022 18:41:46 +0200 Subject: Update modules/ui.py MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: Víctor Gallego --- modules/ui.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 3b28b69c..1f6fcdc9 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1190,7 +1190,8 @@ def create_ui(wrap_gradio_gpu_call): with gr.Column(): create_embedding = gr.Button(value="Create embedding", variant='primary') - with gr.Tab(label="Create images embedding"): + with gr.Tab(label="Create aesthetic images embedding"): + new_embedding_name_ae = gr.Textbox(label="Name") process_src_ae = gr.Textbox(label='Source directory') batch_ae = gr.Slider(minimum=1, maximum=1024, step=1, label="Batch size", value=256) -- cgit v1.2.3 From 9a33292ce41b01252cdb8ab6214a11d274e32fa0 Mon Sep 17 00:00:00 2001 From: zhengxiaoyao0716 <1499383852@qq.com> Date: Sat, 15 Oct 2022 01:04:47 +0800 Subject: reload javascript files when custom script bodies --- modules/ui.py | 28 ++++++++++++++++------------ 1 file changed, 16 insertions(+), 12 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index b867d40f..90b8646b 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -12,7 +12,7 @@ import time import traceback import platform import subprocess as sp -from functools import reduce +from functools import partial, reduce import numpy as np import torch @@ -1491,6 +1491,7 @@ Requested path was: {f} def reload_scripts(): modules.scripts.reload_script_body_only() + reload_javascript() # need to refresh the html page reload_script_bodies.click( fn=reload_scripts, @@ -1738,22 +1739,25 @@ Requested path was: {f} return demo -with open(os.path.join(script_path, "script.js"), "r", encoding="utf8") as jsfile: - javascript = f'' +def load_javascript(raw_response): + with open(os.path.join(script_path, "script.js"), "r", encoding="utf8") as jsfile: + javascript = f'' -jsdir = os.path.join(script_path, "javascript") -for filename in sorted(os.listdir(jsdir)): - with open(os.path.join(jsdir, filename), "r", encoding="utf8") as jsfile: - javascript += f"\n" + jsdir = os.path.join(script_path, "javascript") + for filename in sorted(os.listdir(jsdir)): + with open(os.path.join(jsdir, filename), "r", encoding="utf8") as jsfile: + javascript += f"\n" - -if 'gradio_routes_templates_response' not in globals(): def template_response(*args, **kwargs): - res = gradio_routes_templates_response(*args, **kwargs) - res.body = res.body.replace(b'', f'{javascript}'.encode("utf8")) + res = raw_response(*args, **kwargs) + res.body = res.body.replace( + b'', f'{javascript}'.encode("utf8")) res.init_headers() return res - gradio_routes_templates_response = gradio.routes.templates.TemplateResponse gradio.routes.templates.TemplateResponse = template_response + +reload_javascript = partial(load_javascript, + gradio.routes.templates.TemplateResponse) +reload_javascript() -- cgit v1.2.3 From 3d21684ee30ca5734126b8d08c05b3a0f513fe75 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sun, 16 Oct 2022 00:01:00 +0200 Subject: Add support to other img format, fixed dropbox update --- modules/aesthetic_clip.py | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'modules') diff --git a/modules/aesthetic_clip.py b/modules/aesthetic_clip.py index bcf2b073..68264284 100644 --- a/modules/aesthetic_clip.py +++ b/modules/aesthetic_clip.py @@ -8,7 +8,7 @@ import gradio as gr import torch from PIL import Image from modules import shared -from modules.shared import device, aesthetic_embeddings +from modules.shared import device from transformers import CLIPModel, CLIPProcessor from tqdm.auto import tqdm @@ -20,7 +20,7 @@ def get_all_images_in_folder(folder): def check_is_valid_image_file(filename): - return filename.lower().endswith(('.png', '.jpg', '.jpeg')) + return filename.lower().endswith(('.png', '.jpg', '.jpeg', ".gif", ".tiff", ".webp")) def batched(dataset, total, n=1): @@ -73,6 +73,6 @@ def generate_imgs_embd(name, folder, batch_size): Aesthetic embedding saved to {html.escape(path)} """ shared.update_aesthetic_embeddings() - return gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Imgs embedding", - value=sorted(aesthetic_embeddings.keys())[0] if len( - aesthetic_embeddings) > 0 else None), res, "" + return gr.Dropdown.update(choices=sorted(shared.aesthetic_embeddings.keys()), label="Imgs embedding", + value=sorted(shared.aesthetic_embeddings.keys())[0] if len( + shared.aesthetic_embeddings) > 0 else None), res, "" -- cgit v1.2.3 From 9325c85f780c569d1823e422eaf51b2e497e0d3e Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sun, 16 Oct 2022 00:23:47 +0200 Subject: fixed dropbox update --- modules/sd_hijack.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 192883b2..491312b4 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -9,7 +9,7 @@ from torch.nn.functional import silu import modules.textual_inversion.textual_inversion from modules import prompt_parser, devices, sd_hijack_optimizations, shared -from modules.shared import opts, device, cmd_opts, aesthetic_embeddings +from modules.shared import opts, device, cmd_opts from modules.sd_hijack_optimizations import invokeAI_mps_available import ldm.modules.attention @@ -182,7 +182,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): image_embs_name = None if image_embs_name is not None and self.image_embs_name != image_embs_name: self.image_embs_name = image_embs_name - self.image_embs = torch.load(aesthetic_embeddings[self.image_embs_name], map_location=device) + self.image_embs = torch.load(shared.aesthetic_embeddings[self.image_embs_name], map_location=device) self.image_embs /= self.image_embs.norm(dim=-1, keepdim=True) self.image_embs.requires_grad_(False) -- cgit v1.2.3 From 523140d7805c644700009b8a2483ff4eb4a22304 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sun, 16 Oct 2022 10:23:30 +0200 Subject: ui fix --- modules/aesthetic_clip.py | 3 +-- modules/sd_hijack.py | 3 +-- modules/shared.py | 2 ++ modules/ui.py | 24 ++++++++++++++---------- 4 files changed, 18 insertions(+), 14 deletions(-) (limited to 'modules') diff --git a/modules/aesthetic_clip.py b/modules/aesthetic_clip.py index 68264284..ccb35c73 100644 --- a/modules/aesthetic_clip.py +++ b/modules/aesthetic_clip.py @@ -74,5 +74,4 @@ def generate_imgs_embd(name, folder, batch_size): """ shared.update_aesthetic_embeddings() return gr.Dropdown.update(choices=sorted(shared.aesthetic_embeddings.keys()), label="Imgs embedding", - value=sorted(shared.aesthetic_embeddings.keys())[0] if len( - shared.aesthetic_embeddings) > 0 else None), res, "" + value="None"), res, "" diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 01fcb78f..2de2eed5 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -392,8 +392,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): z1 = self.process_tokens(tokens, multipliers) z = z1 if z is None else torch.cat((z, z1), axis=-2) - if len(text[ - 0]) != 0 and self.aesthetic_steps != 0 and self.aesthetic_lr != 0 and self.aesthetic_weight != 0 and self.image_embs_name != None: + if self.aesthetic_steps != 0 and self.aesthetic_lr != 0 and self.aesthetic_weight != 0 and self.image_embs_name != None: if not opts.use_old_emphasis_implementation: remade_batch_tokens = [ [self.wrapped.tokenizer.bos_token_id] + x[:75] + [self.wrapped.tokenizer.eos_token_id] for x in diff --git a/modules/shared.py b/modules/shared.py index 3c5ffef1..e2c98b2d 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -96,11 +96,13 @@ loaded_hypernetwork = None aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} +aesthetic_embeddings = aesthetic_embeddings | {"None": None} def update_aesthetic_embeddings(): global aesthetic_embeddings aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} + aesthetic_embeddings = aesthetic_embeddings | {"None": None} def reload_hypernetworks(): global hypernetworks diff --git a/modules/ui.py b/modules/ui.py index 13ba3142..4069f0d2 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -594,19 +594,23 @@ def create_ui(wrap_gradio_gpu_call): height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) with gr.Group(): - aesthetic_lr = gr.Textbox(label='Aesthetic learning rate', placeholder="Aesthetic learning rate", value="0.0001") - - aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) - aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) + with gr.Accordion("Open for Clip Aesthetic!",open=False): + with gr.Row(): + aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) + aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) - with gr.Row(): - aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", value="") - aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) - aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) + with gr.Row(): + aesthetic_lr = gr.Textbox(label='Aesthetic learning rate', placeholder="Aesthetic learning rate", value="0.0001") + aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) + aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), + label="Aesthetic imgs embedding", + value="None") - aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), label="Aesthetic imgs embedding", value=sorted(aesthetic_embeddings.keys())[0] if len(aesthetic_embeddings) > 0 else None) + with gr.Row(): + aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", value="") + aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) + aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) - aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) with gr.Row(): restore_faces = gr.Checkbox(label='Restore faces', value=False, visible=len(shared.face_restorers) > 1) -- cgit v1.2.3 From e4f8b5f00dd33b7547cc6b76fbed26bb83b37a64 Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sun, 16 Oct 2022 10:28:21 +0200 Subject: ui fix --- modules/sd_hijack.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 2de2eed5..5d0590af 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -178,7 +178,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): self.load_image_embs(image_embs_name) def load_image_embs(self, image_embs_name): - if image_embs_name is None or len(image_embs_name) == 0: + if image_embs_name is None or len(image_embs_name) == 0 or image_embs_name == "None": image_embs_name = None if image_embs_name is not None and self.image_embs_name != image_embs_name: self.image_embs_name = image_embs_name -- cgit v1.2.3 From 9324cdaa3199d65c182858785dd1eca42b192b8e Mon Sep 17 00:00:00 2001 From: MalumaDev Date: Sun, 16 Oct 2022 17:53:56 +0200 Subject: ui fix, re organization of the code --- modules/aesthetic_clip.py | 154 +++++++++++++++++++++++++++++++++-- modules/img2img.py | 14 +++- modules/processing.py | 29 ++----- modules/sd_hijack.py | 102 ++--------------------- modules/sd_models.py | 5 +- modules/shared.py | 14 +++- modules/textual_inversion/dataset.py | 2 +- modules/txt2img.py | 18 ++-- modules/ui.py | 52 +++++++----- 9 files changed, 233 insertions(+), 157 deletions(-) (limited to 'modules') diff --git a/modules/aesthetic_clip.py b/modules/aesthetic_clip.py index ccb35c73..34efa931 100644 --- a/modules/aesthetic_clip.py +++ b/modules/aesthetic_clip.py @@ -1,3 +1,4 @@ +import copy import itertools import os from pathlib import Path @@ -7,11 +8,12 @@ import gc import gradio as gr import torch from PIL import Image -from modules import shared -from modules.shared import device -from transformers import CLIPModel, CLIPProcessor +from torch import optim -from tqdm.auto import tqdm +from modules import shared +from transformers import CLIPModel, CLIPProcessor, CLIPTokenizer +from tqdm.auto import tqdm, trange +from modules.shared import opts, device def get_all_images_in_folder(folder): @@ -37,12 +39,39 @@ def iter_to_batched(iterable, n=1): yield chunk +def create_ui(): + with gr.Group(): + with gr.Accordion("Open for Clip Aesthetic!", open=False): + with gr.Row(): + aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", + value=0.9) + aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) + + with gr.Row(): + aesthetic_lr = gr.Textbox(label='Aesthetic learning rate', + placeholder="Aesthetic learning rate", value="0.0001") + aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) + aesthetic_imgs = gr.Dropdown(sorted(shared.aesthetic_embeddings.keys()), + label="Aesthetic imgs embedding", + value="None") + + with gr.Row(): + aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', + placeholder="This text is used to rotate the feature space of the imgs embs", + value="") + aesthetic_slerp_angle = gr.Slider(label='Slerp angle', minimum=0, maximum=1, step=0.01, + value=0.1) + aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) + + return aesthetic_weight, aesthetic_steps, aesthetic_lr, aesthetic_slerp, aesthetic_imgs, aesthetic_imgs_text, aesthetic_slerp_angle, aesthetic_text_negative + + def generate_imgs_embd(name, folder, batch_size): # clipModel = CLIPModel.from_pretrained( # shared.sd_model.cond_stage_model.clipModel.name_or_path # ) - model = CLIPModel.from_pretrained(shared.sd_model.cond_stage_model.clipModel.name_or_path).to(device) - processor = CLIPProcessor.from_pretrained(shared.sd_model.cond_stage_model.clipModel.name_or_path) + model = shared.clip_model.to(device) + processor = CLIPProcessor.from_pretrained(model.name_or_path) with torch.no_grad(): embs = [] @@ -63,7 +92,6 @@ def generate_imgs_embd(name, folder, batch_size): torch.save(embs, path) model = model.cpu() - del model del processor del embs gc.collect() @@ -74,4 +102,114 @@ def generate_imgs_embd(name, folder, batch_size): """ shared.update_aesthetic_embeddings() return gr.Dropdown.update(choices=sorted(shared.aesthetic_embeddings.keys()), label="Imgs embedding", - value="None"), res, "" + value="None"), \ + gr.Dropdown.update(choices=sorted(shared.aesthetic_embeddings.keys()), + label="Imgs embedding", + value="None"), res, "" + + +def slerp(low, high, val): + low_norm = low / torch.norm(low, dim=1, keepdim=True) + high_norm = high / torch.norm(high, dim=1, keepdim=True) + omega = torch.acos((low_norm * high_norm).sum(1)) + so = torch.sin(omega) + res = (torch.sin((1.0 - val) * omega) / so).unsqueeze(1) * low + (torch.sin(val * omega) / so).unsqueeze(1) * high + return res + + +class AestheticCLIP: + def __init__(self): + self.skip = False + self.aesthetic_steps = 0 + self.aesthetic_weight = 0 + self.aesthetic_lr = 0 + self.slerp = False + self.aesthetic_text_negative = "" + self.aesthetic_slerp_angle = 0 + self.aesthetic_imgs_text = "" + + self.image_embs_name = None + self.image_embs = None + self.load_image_embs(None) + + def set_aesthetic_params(self, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, image_embs_name=None, + aesthetic_slerp=True, aesthetic_imgs_text="", + aesthetic_slerp_angle=0.15, + aesthetic_text_negative=False): + self.aesthetic_imgs_text = aesthetic_imgs_text + self.aesthetic_slerp_angle = aesthetic_slerp_angle + self.aesthetic_text_negative = aesthetic_text_negative + self.slerp = aesthetic_slerp + self.aesthetic_lr = aesthetic_lr + self.aesthetic_weight = aesthetic_weight + self.aesthetic_steps = aesthetic_steps + self.load_image_embs(image_embs_name) + + def set_skip(self, skip): + self.skip = skip + + def load_image_embs(self, image_embs_name): + if image_embs_name is None or len(image_embs_name) == 0 or image_embs_name == "None": + image_embs_name = None + self.image_embs_name = None + if image_embs_name is not None and self.image_embs_name != image_embs_name: + self.image_embs_name = image_embs_name + self.image_embs = torch.load(shared.aesthetic_embeddings[self.image_embs_name], map_location=device) + self.image_embs /= self.image_embs.norm(dim=-1, keepdim=True) + self.image_embs.requires_grad_(False) + + def __call__(self, z, remade_batch_tokens): + if not self.skip and self.aesthetic_steps != 0 and self.aesthetic_lr != 0 and self.aesthetic_weight != 0 and self.image_embs_name is not None: + tokenizer = shared.sd_model.cond_stage_model.tokenizer + if not opts.use_old_emphasis_implementation: + remade_batch_tokens = [ + [tokenizer.bos_token_id] + x[:75] + [tokenizer.eos_token_id] for x in + remade_batch_tokens] + + tokens = torch.asarray(remade_batch_tokens).to(device) + + model = copy.deepcopy(shared.clip_model).to(device) + model.requires_grad_(True) + if self.aesthetic_imgs_text is not None and len(self.aesthetic_imgs_text) > 0: + text_embs_2 = model.get_text_features( + **tokenizer([self.aesthetic_imgs_text], padding=True, return_tensors="pt").to(device)) + if self.aesthetic_text_negative: + text_embs_2 = self.image_embs - text_embs_2 + text_embs_2 /= text_embs_2.norm(dim=-1, keepdim=True) + img_embs = slerp(self.image_embs, text_embs_2, self.aesthetic_slerp_angle) + else: + img_embs = self.image_embs + + with torch.enable_grad(): + + # We optimize the model to maximize the similarity + optimizer = optim.Adam( + model.text_model.parameters(), lr=self.aesthetic_lr + ) + + for _ in trange(self.aesthetic_steps, desc="Aesthetic optimization"): + text_embs = model.get_text_features(input_ids=tokens) + text_embs = text_embs / text_embs.norm(dim=-1, keepdim=True) + sim = text_embs @ img_embs.T + loss = -sim + optimizer.zero_grad() + loss.mean().backward() + optimizer.step() + + zn = model.text_model(input_ids=tokens, output_hidden_states=-opts.CLIP_stop_at_last_layers) + if opts.CLIP_stop_at_last_layers > 1: + zn = zn.hidden_states[-opts.CLIP_stop_at_last_layers] + zn = model.text_model.final_layer_norm(zn) + else: + zn = zn.last_hidden_state + model.cpu() + del model + gc.collect() + torch.cuda.empty_cache() + zn = torch.concat([zn[77 * i:77 * (i + 1)] for i in range(max(z.shape[1] // 77, 1))], 1) + if self.slerp: + z = slerp(z, zn, self.aesthetic_weight) + else: + z = z * (1 - self.aesthetic_weight) + zn * self.aesthetic_weight + + return z diff --git a/modules/img2img.py b/modules/img2img.py index 24126774..4ed80c4b 100644 --- a/modules/img2img.py +++ b/modules/img2img.py @@ -56,7 +56,14 @@ def process_batch(p, input_dir, output_dir, args): processed_image.save(os.path.join(output_dir, filename)) -def img2img(mode: int, prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: str, init_img, init_img_with_mask, init_img_inpaint, init_mask_inpaint, mask_mode, steps: int, sampler_index: int, mask_blur: int, inpainting_fill: int, restore_faces: bool, tiling: bool, n_iter: int, batch_size: int, cfg_scale: float, denoising_strength: float, seed: int, subseed: int, subseed_strength: float, seed_resize_from_h: int, seed_resize_from_w: int, seed_enable_extras: bool, height: int, width: int, resize_mode: int, inpaint_full_res: bool, inpaint_full_res_padding: int, inpainting_mask_invert: int, img2img_batch_input_dir: str, img2img_batch_output_dir: str, *args): +def img2img(mode: int, prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: str, init_img, init_img_with_mask, init_img_inpaint, init_mask_inpaint, mask_mode, steps: int, sampler_index: int, mask_blur: int, inpainting_fill: int, restore_faces: bool, tiling: bool, n_iter: int, batch_size: int, cfg_scale: float, denoising_strength: float, seed: int, subseed: int, subseed_strength: float, seed_resize_from_h: int, seed_resize_from_w: int, seed_enable_extras: bool, height: int, width: int, resize_mode: int, inpaint_full_res: bool, inpaint_full_res_padding: int, inpainting_mask_invert: int, img2img_batch_input_dir: str, img2img_batch_output_dir: str, + aesthetic_lr=0, + aesthetic_weight=0, aesthetic_steps=0, + aesthetic_imgs=None, + aesthetic_slerp=False, + aesthetic_imgs_text="", + aesthetic_slerp_angle=0.15, + aesthetic_text_negative=False, *args): is_inpaint = mode == 1 is_batch = mode == 2 @@ -109,6 +116,11 @@ def img2img(mode: int, prompt: str, negative_prompt: str, prompt_style: str, pro inpainting_mask_invert=inpainting_mask_invert, ) + shared.aesthetic_clip.set_aesthetic_params(float(aesthetic_lr), float(aesthetic_weight), int(aesthetic_steps), + aesthetic_imgs, aesthetic_slerp, aesthetic_imgs_text, + aesthetic_slerp_angle, + aesthetic_text_negative) + if shared.cmd_opts.enable_console_prompts: print(f"\nimg2img: {prompt}", file=shared.progress_print_out) diff --git a/modules/processing.py b/modules/processing.py index 1db26c3e..685f9fcd 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -146,7 +146,8 @@ class Processed: self.prompt = self.prompt if type(self.prompt) != list else self.prompt[0] self.negative_prompt = self.negative_prompt if type(self.negative_prompt) != list else self.negative_prompt[0] self.seed = int(self.seed if type(self.seed) != list else self.seed[0]) if self.seed is not None else -1 - self.subseed = int(self.subseed if type(self.subseed) != list else self.subseed[0]) if self.subseed is not None else -1 + self.subseed = int( + self.subseed if type(self.subseed) != list else self.subseed[0]) if self.subseed is not None else -1 self.all_prompts = all_prompts or [self.prompt] self.all_seeds = all_seeds or [self.seed] @@ -332,16 +333,9 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration return f"{all_prompts[index]}{negative_prompt_text}\n{generation_params_text}".strip() -def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, - aesthetic_imgs=None, aesthetic_slerp=False, aesthetic_imgs_text="", - aesthetic_slerp_angle=0.15, - aesthetic_text_negative=False) -> Processed: +def process_images(p: StableDiffusionProcessing) -> Processed: """this is the main loop that both txt2img and img2img use; it calls func_init once inside all the scopes and func_sample once per batch""" - aesthetic_lr = float(aesthetic_lr) - aesthetic_weight = float(aesthetic_weight) - aesthetic_steps = int(aesthetic_steps) - if type(p.prompt) == list: assert (len(p.prompt) > 0) else: @@ -417,16 +411,10 @@ def process_images(p: StableDiffusionProcessing, aesthetic_lr=0, aesthetic_weigh # uc = p.sd_model.get_learned_conditioning(len(prompts) * [p.negative_prompt]) # c = p.sd_model.get_learned_conditioning(prompts) with devices.autocast(): - if hasattr(shared.sd_model.cond_stage_model, "set_aesthetic_params"): - shared.sd_model.cond_stage_model.set_aesthetic_params() + shared.aesthetic_clip.set_skip(True) uc = prompt_parser.get_learned_conditioning(shared.sd_model, len(prompts) * [p.negative_prompt], p.steps) - if hasattr(shared.sd_model.cond_stage_model, "set_aesthetic_params"): - shared.sd_model.cond_stage_model.set_aesthetic_params(aesthetic_lr, aesthetic_weight, - aesthetic_steps, aesthetic_imgs, - aesthetic_slerp, aesthetic_imgs_text, - aesthetic_slerp_angle, - aesthetic_text_negative) + shared.aesthetic_clip.set_skip(False) c = prompt_parser.get_multicond_learned_conditioning(shared.sd_model, prompts, p.steps) if len(model_hijack.comments) > 0: @@ -582,7 +570,6 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.truncate_x = int(self.firstphase_width - firstphase_width_truncated) // opt_f self.truncate_y = int(self.firstphase_height - firstphase_height_truncated) // opt_f - def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model) @@ -600,10 +587,12 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): seed_resize_from_w=self.seed_resize_from_w, p=self) samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning) - samples = samples[:, :, self.truncate_y//2:samples.shape[2]-self.truncate_y//2, self.truncate_x//2:samples.shape[3]-self.truncate_x//2] + samples = samples[:, :, self.truncate_y // 2:samples.shape[2] - self.truncate_y // 2, + self.truncate_x // 2:samples.shape[3] - self.truncate_x // 2] if opts.use_scale_latent_for_hires_fix: - samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), mode="bilinear") + samples = torch.nn.functional.interpolate(samples, size=(self.height // opt_f, self.width // opt_f), + mode="bilinear") else: decoded_samples = decode_first_stage(self.sd_model, samples) lowres_samples = torch.clamp((decoded_samples + 1.0) / 2.0, min=0.0, max=1.0) diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 5d0590af..227e7670 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -29,8 +29,8 @@ def apply_optimizations(): ldm.modules.diffusionmodules.model.nonlinearity = silu - - if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and (6, 0) <= torch.cuda.get_device_capability(shared.device) <= (9, 0)): + if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and ( + 6, 0) <= torch.cuda.get_device_capability(shared.device) <= (9, 0)): print("Applying xformers cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.xformers_attention_forward ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.xformers_attnblock_forward @@ -118,33 +118,14 @@ class StableDiffusionModelHijack: return remade_batch_tokens[0], token_count, get_target_prompt_token_count(token_count) -def slerp(low, high, val): - low_norm = low / torch.norm(low, dim=1, keepdim=True) - high_norm = high / torch.norm(high, dim=1, keepdim=True) - omega = torch.acos((low_norm * high_norm).sum(1)) - so = torch.sin(omega) - res = (torch.sin((1.0 - val) * omega) / so).unsqueeze(1) * low + (torch.sin(val * omega) / so).unsqueeze(1) * high - return res - - class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): def __init__(self, wrapped, hijack): super().__init__() self.wrapped = wrapped - self.clipModel = CLIPModel.from_pretrained( - self.wrapped.transformer.name_or_path - ) - del self.clipModel.vision_model - self.tokenizer = CLIPTokenizer.from_pretrained(self.wrapped.transformer.name_or_path) - self.hijack: StableDiffusionModelHijack = hijack - self.tokenizer = wrapped.tokenizer - # self.vision = CLIPVisionModel.from_pretrained(self.wrapped.transformer.name_or_path).eval() - self.image_embs_name = None - self.image_embs = None - self.load_image_embs(None) self.token_mults = {} - + self.hijack: StableDiffusionModelHijack = hijack + self.tokenizer = wrapped.tokenizer self.comma_token = [v for k, v in self.tokenizer.get_vocab().items() if k == ','][0] tokens_with_parens = [(k, v) for k, v in self.tokenizer.get_vocab().items() if @@ -164,28 +145,6 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): if mult != 1.0: self.token_mults[ident] = mult - def set_aesthetic_params(self, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, image_embs_name=None, - aesthetic_slerp=True, aesthetic_imgs_text="", - aesthetic_slerp_angle=0.15, - aesthetic_text_negative=False): - self.aesthetic_imgs_text = aesthetic_imgs_text - self.aesthetic_slerp_angle = aesthetic_slerp_angle - self.aesthetic_text_negative = aesthetic_text_negative - self.slerp = aesthetic_slerp - self.aesthetic_lr = aesthetic_lr - self.aesthetic_weight = aesthetic_weight - self.aesthetic_steps = aesthetic_steps - self.load_image_embs(image_embs_name) - - def load_image_embs(self, image_embs_name): - if image_embs_name is None or len(image_embs_name) == 0 or image_embs_name == "None": - image_embs_name = None - if image_embs_name is not None and self.image_embs_name != image_embs_name: - self.image_embs_name = image_embs_name - self.image_embs = torch.load(shared.aesthetic_embeddings[self.image_embs_name], map_location=device) - self.image_embs /= self.image_embs.norm(dim=-1, keepdim=True) - self.image_embs.requires_grad_(False) - def tokenize_line(self, line, used_custom_terms, hijack_comments): id_end = self.wrapped.tokenizer.eos_token_id @@ -391,58 +350,7 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): z1 = self.process_tokens(tokens, multipliers) z = z1 if z is None else torch.cat((z, z1), axis=-2) - - if self.aesthetic_steps != 0 and self.aesthetic_lr != 0 and self.aesthetic_weight != 0 and self.image_embs_name != None: - if not opts.use_old_emphasis_implementation: - remade_batch_tokens = [ - [self.wrapped.tokenizer.bos_token_id] + x[:75] + [self.wrapped.tokenizer.eos_token_id] for x in - remade_batch_tokens] - - tokens = torch.asarray(remade_batch_tokens).to(device) - - model = copy.deepcopy(self.clipModel).to(device) - model.requires_grad_(True) - if self.aesthetic_imgs_text is not None and len(self.aesthetic_imgs_text) > 0: - text_embs_2 = model.get_text_features( - **self.tokenizer([self.aesthetic_imgs_text], padding=True, return_tensors="pt").to(device)) - if self.aesthetic_text_negative: - text_embs_2 = self.image_embs - text_embs_2 - text_embs_2 /= text_embs_2.norm(dim=-1, keepdim=True) - img_embs = slerp(self.image_embs, text_embs_2, self.aesthetic_slerp_angle) - else: - img_embs = self.image_embs - - with torch.enable_grad(): - - # We optimize the model to maximize the similarity - optimizer = optim.Adam( - model.text_model.parameters(), lr=self.aesthetic_lr - ) - - for i in trange(self.aesthetic_steps, desc="Aesthetic optimization"): - text_embs = model.get_text_features(input_ids=tokens) - text_embs = text_embs / text_embs.norm(dim=-1, keepdim=True) - sim = text_embs @ img_embs.T - loss = -sim - optimizer.zero_grad() - loss.mean().backward() - optimizer.step() - - zn = model.text_model(input_ids=tokens, output_hidden_states=-opts.CLIP_stop_at_last_layers) - if opts.CLIP_stop_at_last_layers > 1: - zn = zn.hidden_states[-opts.CLIP_stop_at_last_layers] - zn = model.text_model.final_layer_norm(zn) - else: - zn = zn.last_hidden_state - model.cpu() - del model - - zn = torch.concat([zn for i in range(z.shape[1] // 77)], 1) - if self.slerp: - z = slerp(z, zn, self.aesthetic_weight) - else: - z = z * (1 - self.aesthetic_weight) + zn * self.aesthetic_weight - + z = shared.aesthetic_clip(z, remade_batch_tokens) remade_batch_tokens = rem_tokens batch_multipliers = rem_multipliers i += 1 diff --git a/modules/sd_models.py b/modules/sd_models.py index 3aa21ec1..8e4ee435 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -20,7 +20,7 @@ checkpoints_loaded = collections.OrderedDict() try: # this silences the annoying "Some weights of the model checkpoint were not used when initializing..." message at start. - from transformers import logging + from transformers import logging, CLIPModel logging.set_verbosity_error() except Exception: @@ -196,6 +196,9 @@ def load_model(): sd_hijack.model_hijack.hijack(sd_model) + if shared.clip_model is None or shared.clip_model.transformer.name_or_path != sd_model.cond_stage_model.wrapped.transformer.name_or_path: + shared.clip_model = CLIPModel.from_pretrained(sd_model.cond_stage_model.wrapped.transformer.name_or_path) + sd_model.eval() print(f"Model loaded.") diff --git a/modules/shared.py b/modules/shared.py index e2c98b2d..e19ca779 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -3,6 +3,7 @@ import datetime import json import os import sys +from collections import OrderedDict import gradio as gr import tqdm @@ -94,15 +95,15 @@ os.makedirs(cmd_opts.hypernetwork_dir, exist_ok=True) hypernetworks = hypernetwork.list_hypernetworks(cmd_opts.hypernetwork_dir) loaded_hypernetwork = None -aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in - os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} -aesthetic_embeddings = aesthetic_embeddings | {"None": None} +aesthetic_embeddings = {} def update_aesthetic_embeddings(): global aesthetic_embeddings aesthetic_embeddings = {f.replace(".pt",""): os.path.join(cmd_opts.aesthetic_embeddings_dir, f) for f in os.listdir(cmd_opts.aesthetic_embeddings_dir) if f.endswith(".pt")} - aesthetic_embeddings = aesthetic_embeddings | {"None": None} + aesthetic_embeddings = OrderedDict(**{"None": None}, **aesthetic_embeddings) + +update_aesthetic_embeddings() def reload_hypernetworks(): global hypernetworks @@ -381,6 +382,11 @@ sd_upscalers = [] sd_model = None +clip_model = None + +from modules.aesthetic_clip import AestheticCLIP +aesthetic_clip = AestheticCLIP() + progress_print_out = sys.stdout diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py index 68ceffe3..23bb4b6a 100644 --- a/modules/textual_inversion/dataset.py +++ b/modules/textual_inversion/dataset.py @@ -49,7 +49,7 @@ class PersonalizedBase(Dataset): print("Preparing dataset...") for path in tqdm.tqdm(self.image_paths): try: - image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.Resampling.BICUBIC) + image = Image.open(path).convert('RGB').resize((self.width, self.height), PIL.Image.BICUBIC) except Exception: continue diff --git a/modules/txt2img.py b/modules/txt2img.py index 8f394d05..6cbc50fc 100644 --- a/modules/txt2img.py +++ b/modules/txt2img.py @@ -1,12 +1,17 @@ import modules.scripts -from modules.processing import StableDiffusionProcessing, Processed, StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img, process_images +from modules.processing import StableDiffusionProcessing, Processed, StableDiffusionProcessingTxt2Img, \ + StableDiffusionProcessingImg2Img, process_images from modules.shared import opts, cmd_opts import modules.shared as shared import modules.processing as processing from modules.ui import plaintext_to_html -def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: str, steps: int, sampler_index: int, restore_faces: bool, tiling: bool, n_iter: int, batch_size: int, cfg_scale: float, seed: int, subseed: int, subseed_strength: float, seed_resize_from_h: int, seed_resize_from_w: int, seed_enable_extras: bool, height: int, width: int, enable_hr: bool, denoising_strength: float, firstphase_width: int, firstphase_height: int,aesthetic_lr=0, +def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: str, steps: int, sampler_index: int, + restore_faces: bool, tiling: bool, n_iter: int, batch_size: int, cfg_scale: float, seed: int, subseed: int, + subseed_strength: float, seed_resize_from_h: int, seed_resize_from_w: int, seed_enable_extras: bool, + height: int, width: int, enable_hr: bool, denoising_strength: float, firstphase_width: int, + firstphase_height: int, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, aesthetic_imgs=None, aesthetic_slerp=False, @@ -41,15 +46,17 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: firstphase_height=firstphase_height if enable_hr else None, ) + shared.aesthetic_clip.set_aesthetic_params(float(aesthetic_lr), float(aesthetic_weight), int(aesthetic_steps), + aesthetic_imgs, aesthetic_slerp, aesthetic_imgs_text, aesthetic_slerp_angle, + aesthetic_text_negative) + if cmd_opts.enable_console_prompts: print(f"\ntxt2img: {prompt}", file=shared.progress_print_out) processed = modules.scripts.scripts_txt2img.run(p, *args) if processed is None: - processed = process_images(p, aesthetic_lr, aesthetic_weight, aesthetic_steps, aesthetic_imgs, aesthetic_slerp,aesthetic_imgs_text, - aesthetic_slerp_angle, - aesthetic_text_negative) + processed = process_images(p) shared.total_tqdm.clear() @@ -61,4 +68,3 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: processed.images = [] return processed.images, generation_info_js, plaintext_to_html(processed.info) - diff --git a/modules/ui.py b/modules/ui.py index 4069f0d2..0e5d73f0 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -43,7 +43,7 @@ from modules.images import save_image import modules.textual_inversion.ui import modules.hypernetworks.ui -import modules.aesthetic_clip +import modules.aesthetic_clip as aesthetic_clip import modules.images_history as img_his @@ -593,23 +593,25 @@ def create_ui(wrap_gradio_gpu_call): width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512) height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) - with gr.Group(): - with gr.Accordion("Open for Clip Aesthetic!",open=False): - with gr.Row(): - aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) - aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) - - with gr.Row(): - aesthetic_lr = gr.Textbox(label='Aesthetic learning rate', placeholder="Aesthetic learning rate", value="0.0001") - aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) - aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), - label="Aesthetic imgs embedding", - value="None") - - with gr.Row(): - aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", value="") - aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) - aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) + # with gr.Group(): + # with gr.Accordion("Open for Clip Aesthetic!",open=False): + # with gr.Row(): + # aesthetic_weight = gr.Slider(minimum=0, maximum=1, step=0.01, label="Aesthetic weight", value=0.9) + # aesthetic_steps = gr.Slider(minimum=0, maximum=50, step=1, label="Aesthetic steps", value=5) + # + # with gr.Row(): + # aesthetic_lr = gr.Textbox(label='Aesthetic learning rate', placeholder="Aesthetic learning rate", value="0.0001") + # aesthetic_slerp = gr.Checkbox(label="Slerp interpolation", value=False) + # aesthetic_imgs = gr.Dropdown(sorted(aesthetic_embeddings.keys()), + # label="Aesthetic imgs embedding", + # value="None") + # + # with gr.Row(): + # aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", value="") + # aesthetic_slerp_angle = gr.Slider(label='Slerp angle',minimum=0, maximum=1, step=0.01, value=0.1) + # aesthetic_text_negative = gr.Checkbox(label="Is negative text", value=False) + + aesthetic_weight, aesthetic_steps, aesthetic_lr, aesthetic_slerp, aesthetic_imgs, aesthetic_imgs_text, aesthetic_slerp_angle, aesthetic_text_negative = aesthetic_clip.create_ui() with gr.Row(): @@ -840,6 +842,9 @@ def create_ui(wrap_gradio_gpu_call): width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512) height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) + aesthetic_weight_im, aesthetic_steps_im, aesthetic_lr_im, aesthetic_slerp_im, aesthetic_imgs_im, aesthetic_imgs_text_im, aesthetic_slerp_angle_im, aesthetic_text_negative_im = aesthetic_clip.create_ui() + + with gr.Row(): restore_faces = gr.Checkbox(label='Restore faces', value=False, visible=len(shared.face_restorers) > 1) tiling = gr.Checkbox(label='Tiling', value=False) @@ -944,6 +949,14 @@ def create_ui(wrap_gradio_gpu_call): inpainting_mask_invert, img2img_batch_input_dir, img2img_batch_output_dir, + aesthetic_lr_im, + aesthetic_weight_im, + aesthetic_steps_im, + aesthetic_imgs_im, + aesthetic_slerp_im, + aesthetic_imgs_text_im, + aesthetic_slerp_angle_im, + aesthetic_text_negative_im, ] + custom_inputs, outputs=[ img2img_gallery, @@ -1283,7 +1296,7 @@ def create_ui(wrap_gradio_gpu_call): ) create_embedding_ae.click( - fn=modules.aesthetic_clip.generate_imgs_embd, + fn=aesthetic_clip.generate_imgs_embd, inputs=[ new_embedding_name_ae, process_src_ae, @@ -1291,6 +1304,7 @@ def create_ui(wrap_gradio_gpu_call): ], outputs=[ aesthetic_imgs, + aesthetic_imgs_im, ti_output, ti_outcome, ] -- cgit v1.2.3 From 60251c9456f5472784862896c2f97e38feb42482 Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Mon, 17 Oct 2022 06:58:42 +0000 Subject: initial prototype by borrowing contracts --- modules/api/api.py | 60 +++++++++++++++++++++++++++++++++++++++++++++++++++ modules/processing.py | 2 +- modules/shared.py | 2 +- 3 files changed, 62 insertions(+), 2 deletions(-) create mode 100644 modules/api/api.py (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py new file mode 100644 index 00000000..9d7c699d --- /dev/null +++ b/modules/api/api.py @@ -0,0 +1,60 @@ +from modules.api.processing import StableDiffusionProcessingAPI +from modules.processing import StableDiffusionProcessingTxt2Img, process_images +import modules.shared as shared +import uvicorn +from fastapi import FastAPI, Body, APIRouter +from fastapi.responses import JSONResponse +from pydantic import BaseModel, Field, Json +import json +import io +import base64 + +app = FastAPI() + +class TextToImageResponse(BaseModel): + images: list[str] = Field(default=None, title="Image", description="The generated image in base64 format.") + parameters: Json + info: Json + + +class Api: + def __init__(self, txt2img, img2img, run_extras, run_pnginfo): + self.router = APIRouter() + app.add_api_route("/v1/txt2img", self.text2imgapi, methods=["POST"]) + + def text2imgapi(self, txt2imgreq: StableDiffusionProcessingAPI ): + print(txt2imgreq) + p = StableDiffusionProcessingTxt2Img(**vars(txt2imgreq)) + p.sd_model = shared.sd_model + print(p) + processed = process_images(p) + + b64images = [] + for i in processed.images: + buffer = io.BytesIO() + i.save(buffer, format="png") + b64images.append(base64.b64encode(buffer.getvalue())) + + response = { + "images": b64images, + "info": processed.js(), + "parameters": json.dumps(vars(txt2imgreq)) + } + + + return TextToImageResponse(images=b64images, parameters=json.dumps(vars(txt2imgreq)), info=json.dumps(processed.info)) + + + + def img2imgendoint(self): + raise NotImplementedError + + def extrasendoint(self): + raise NotImplementedError + + def pnginfoendoint(self): + raise NotImplementedError + + def launch(self, server_name, port): + app.include_router(self.router) + uvicorn.run(app, host=server_name, port=port) \ No newline at end of file diff --git a/modules/processing.py b/modules/processing.py index deb6125e..4a7c6ccc 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -723,4 +723,4 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): del x devices.torch_gc() - return samples + return samples \ No newline at end of file diff --git a/modules/shared.py b/modules/shared.py index c2775603..6c6405fd 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -74,7 +74,7 @@ parser.add_argument("--disable-console-progressbars", action='store_true', help= parser.add_argument("--enable-console-prompts", action='store_true', help="print prompts to console when generating with txt2img and img2img", default=False) parser.add_argument('--vae-path', type=str, help='Path to Variational Autoencoders model', default=None) parser.add_argument("--disable-safe-unpickle", action='store_true', help="disable checking pytorch models for malicious code", default=False) - +parser.add_argument("--api", action='store_true', help="use api=True to launch the api instead of the webui") cmd_opts = parser.parse_args() restricted_opts = [ -- cgit v1.2.3 From 9e02812afd10582f00a7fbbfa63c8f9188678e26 Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Mon, 17 Oct 2022 07:02:08 +0000 Subject: pydantic instrumentation --- modules/api/processing.py | 99 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 99 insertions(+) create mode 100644 modules/api/processing.py (limited to 'modules') diff --git a/modules/api/processing.py b/modules/api/processing.py new file mode 100644 index 00000000..459a8f49 --- /dev/null +++ b/modules/api/processing.py @@ -0,0 +1,99 @@ +from inflection import underscore +from typing import Any, Dict, Optional +from pydantic import BaseModel, Field, create_model +from modules.processing import StableDiffusionProcessing, Processed, StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img, process_images +import inspect + + +class ModelDef(BaseModel): + """Assistance Class for Pydantic Dynamic Model Generation""" + + field: str + field_alias: str + field_type: Any + field_value: Any + + +class pydanticModelGenerator: + """ + Takes source_data:Dict ( a single instance example of something like a JSON node) and self generates a pythonic data model with Alias to original source field names. This makes it easy to popuate or export to other systems yet handle the data in a pythonic way. + Being a pydantic datamodel all the richness of pydantic data validation is available and these models can easily be used in FastAPI and or a ORM + + It does not process full JSON data structures but takes simple JSON document with basic elements + + Provide a model_name, an example of JSON data and a dict of type overrides + + Example: + + source_data = {'Name': '48 Rainbow Rd', + 'GroupAddressStyle': 'ThreeLevel', + 'LastModified': '2020-12-21T07:02:51.2400232Z', + 'ProjectStart': '2020-12-03T07:36:03.324856Z', + 'Comment': '', + 'CompletionStatus': 'Editing', + 'LastUsedPuid': '955', + 'Guid': '0c85957b-c2ae-4985-9752-b300ab385b36'} + + source_overrides = {'Guid':{'type':uuid.UUID}, + 'LastModified':{'type':datetime }, + 'ProjectStart':{'type':datetime }, + } + source_optionals = {"Comment":True} + + #create Model + model_Project=pydanticModelGenerator( + model_name="Project", + source_data=source_data, + overrides=source_overrides, + optionals=source_optionals).generate_model() + + #create instance using DynamicModel + project_instance=model_Project(**project_info) + + """ + + def __init__( + self, + model_name: str = None, + source_data: str = None, + params: Dict = {}, + overrides: Dict = {}, + optionals: Dict = {}, + ): + def field_type_generator(k, v, overrides, optionals): + print(k, v) + field_type = str if not overrides.get(k) else overrides[k]["type"] + if v is None: + field_type = Any + else: + field_type = type(v) + + return Optional[field_type] + + self._model_name = model_name + self._json_data = source_data + self._model_def = [ + ModelDef( + field=underscore(k), + field_alias=k, + field_type=field_type_generator(k, v, overrides, optionals), + field_value=v + ) + for (k,v) in source_data.items() if k in params + ] + + def generate_model(self): + """ + Creates a pydantic BaseModel + from the json and overrides provided at initialization + """ + fields = { + d.field: (d.field_type, Field(default=d.field_value, alias=d.field_alias)) for d in self._model_def + } + DynamicModel = create_model(self._model_name, **fields) + DynamicModel.__config__.allow_population_by_field_name = True + return DynamicModel + +StableDiffusionProcessingAPI = pydanticModelGenerator("StableDiffusionProcessing", + StableDiffusionProcessing().__dict__, + inspect.signature(StableDiffusionProcessing.__init__).parameters).generate_model() \ No newline at end of file -- cgit v1.2.3 From 832b490e5173f78c4d3aa7ca9ca9ac794d140664 Mon Sep 17 00:00:00 2001 From: Jonathan Date: Mon, 17 Oct 2022 03:18:41 -0400 Subject: Update processing.py --- modules/api/processing.py | 41 +++++------------------------------------ 1 file changed, 5 insertions(+), 36 deletions(-) (limited to 'modules') diff --git a/modules/api/processing.py b/modules/api/processing.py index 459a8f49..4c3d0bd0 100644 --- a/modules/api/processing.py +++ b/modules/api/processing.py @@ -16,46 +16,15 @@ class ModelDef(BaseModel): class pydanticModelGenerator: """ - Takes source_data:Dict ( a single instance example of something like a JSON node) and self generates a pythonic data model with Alias to original source field names. This makes it easy to popuate or export to other systems yet handle the data in a pythonic way. - Being a pydantic datamodel all the richness of pydantic data validation is available and these models can easily be used in FastAPI and or a ORM - - It does not process full JSON data structures but takes simple JSON document with basic elements - - Provide a model_name, an example of JSON data and a dict of type overrides - - Example: - - source_data = {'Name': '48 Rainbow Rd', - 'GroupAddressStyle': 'ThreeLevel', - 'LastModified': '2020-12-21T07:02:51.2400232Z', - 'ProjectStart': '2020-12-03T07:36:03.324856Z', - 'Comment': '', - 'CompletionStatus': 'Editing', - 'LastUsedPuid': '955', - 'Guid': '0c85957b-c2ae-4985-9752-b300ab385b36'} - - source_overrides = {'Guid':{'type':uuid.UUID}, - 'LastModified':{'type':datetime }, - 'ProjectStart':{'type':datetime }, - } - source_optionals = {"Comment":True} - - #create Model - model_Project=pydanticModelGenerator( - model_name="Project", - source_data=source_data, - overrides=source_overrides, - optionals=source_optionals).generate_model() - - #create instance using DynamicModel - project_instance=model_Project(**project_info) - + Takes in created classes and stubs them out in a way FastAPI/Pydantic is happy about: + source_data is a snapshot of the default values produced by the class + params are the names of the actual keys required by __init__ """ def __init__( self, model_name: str = None, - source_data: str = None, + source_data: {} = {}, params: Dict = {}, overrides: Dict = {}, optionals: Dict = {}, @@ -96,4 +65,4 @@ class pydanticModelGenerator: StableDiffusionProcessingAPI = pydanticModelGenerator("StableDiffusionProcessing", StableDiffusionProcessing().__dict__, - inspect.signature(StableDiffusionProcessing.__init__).parameters).generate_model() \ No newline at end of file + inspect.signature(StableDiffusionProcessing.__init__).parameters).generate_model() -- cgit v1.2.3 From 99013ba68a5fe1bde3621632e5539c03562a3ae8 Mon Sep 17 00:00:00 2001 From: Jonathan Date: Mon, 17 Oct 2022 03:20:17 -0400 Subject: Update processing.py --- modules/api/processing.py | 1 - 1 file changed, 1 deletion(-) (limited to 'modules') diff --git a/modules/api/processing.py b/modules/api/processing.py index 4c3d0bd0..e4df93c5 100644 --- a/modules/api/processing.py +++ b/modules/api/processing.py @@ -30,7 +30,6 @@ class pydanticModelGenerator: optionals: Dict = {}, ): def field_type_generator(k, v, overrides, optionals): - print(k, v) field_type = str if not overrides.get(k) else overrides[k]["type"] if v is None: field_type = Any -- cgit v1.2.3 From 71d42bb44b257f3fb274c3ad5075a195281ff915 Mon Sep 17 00:00:00 2001 From: Jonathan Date: Mon, 17 Oct 2022 03:22:19 -0400 Subject: Update api.py --- modules/api/api.py | 11 +---------- 1 file changed, 1 insertion(+), 10 deletions(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index 9d7c699d..4d9619a8 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -23,10 +23,8 @@ class Api: app.add_api_route("/v1/txt2img", self.text2imgapi, methods=["POST"]) def text2imgapi(self, txt2imgreq: StableDiffusionProcessingAPI ): - print(txt2imgreq) p = StableDiffusionProcessingTxt2Img(**vars(txt2imgreq)) p.sd_model = shared.sd_model - print(p) processed = process_images(p) b64images = [] @@ -34,13 +32,6 @@ class Api: buffer = io.BytesIO() i.save(buffer, format="png") b64images.append(base64.b64encode(buffer.getvalue())) - - response = { - "images": b64images, - "info": processed.js(), - "parameters": json.dumps(vars(txt2imgreq)) - } - return TextToImageResponse(images=b64images, parameters=json.dumps(vars(txt2imgreq)), info=json.dumps(processed.info)) @@ -57,4 +48,4 @@ class Api: def launch(self, server_name, port): app.include_router(self.router) - uvicorn.run(app, host=server_name, port=port) \ No newline at end of file + uvicorn.run(app, host=server_name, port=port) -- cgit v1.2.3 From d42125baf62880854ad06af06c15c23e7e50cca6 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Mon, 17 Oct 2022 11:50:20 +0300 Subject: add missing requirement for api and fix some typos --- modules/api/api.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index 4d9619a8..fd09d352 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -18,7 +18,7 @@ class TextToImageResponse(BaseModel): class Api: - def __init__(self, txt2img, img2img, run_extras, run_pnginfo): + def __init__(self): self.router = APIRouter() app.add_api_route("/v1/txt2img", self.text2imgapi, methods=["POST"]) -- cgit v1.2.3 From 8c6a981d5d9ef30381ac2327460285111550acbc Mon Sep 17 00:00:00 2001 From: Michoko Date: Mon, 17 Oct 2022 11:05:05 +0200 Subject: Added dark mode switch Launch the UI in dark mode with the --dark-mode switch --- modules/shared.py | 2 +- modules/ui.py | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index c2775603..cbf158e4 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -69,13 +69,13 @@ parser.add_argument("--gradio-img2img-tool", type=str, help='gradio image upload parser.add_argument("--opt-channelslast", action='store_true', help="change memory type for stable diffusion to channels last") parser.add_argument("--styles-file", type=str, help="filename to use for styles", default=os.path.join(script_path, 'styles.csv')) parser.add_argument("--autolaunch", action='store_true', help="open the webui URL in the system's default browser upon launch", default=False) +parser.add_argument("--dark-mode", action='store_true', help="launches the UI in dark mode", default=False) parser.add_argument("--use-textbox-seed", action='store_true', help="use textbox for seeds in UI (no up/down, but possible to input long seeds)", default=False) parser.add_argument("--disable-console-progressbars", action='store_true', help="do not output progressbars to console", default=False) parser.add_argument("--enable-console-prompts", action='store_true', help="print prompts to console when generating with txt2img and img2img", default=False) parser.add_argument('--vae-path', type=str, help='Path to Variational Autoencoders model', default=None) parser.add_argument("--disable-safe-unpickle", action='store_true', help="disable checking pytorch models for malicious code", default=False) - cmd_opts = parser.parse_args() restricted_opts = [ "samples_filename_pattern", diff --git a/modules/ui.py b/modules/ui.py index 43dc88fc..a0cd052e 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1783,6 +1783,8 @@ for filename in sorted(os.listdir(jsdir)): with open(os.path.join(jsdir, filename), "r", encoding="utf8") as jsfile: javascript += f"\n" +if cmd_opts.dark_mode: + javascript += "\n\n" if 'gradio_routes_templates_response' not in globals(): def template_response(*args, **kwargs): -- cgit v1.2.3 From 665beebc0825a6fad410c8252f27f6f6f0bd900b Mon Sep 17 00:00:00 2001 From: Michoko Date: Mon, 17 Oct 2022 18:24:24 +0200 Subject: Use of a --theme argument for more flexibility Added possibility to set the theme (light or dark) --- modules/shared.py | 2 +- modules/ui.py | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index cbf158e4..fa084c69 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -69,7 +69,7 @@ parser.add_argument("--gradio-img2img-tool", type=str, help='gradio image upload parser.add_argument("--opt-channelslast", action='store_true', help="change memory type for stable diffusion to channels last") parser.add_argument("--styles-file", type=str, help="filename to use for styles", default=os.path.join(script_path, 'styles.csv')) parser.add_argument("--autolaunch", action='store_true', help="open the webui URL in the system's default browser upon launch", default=False) -parser.add_argument("--dark-mode", action='store_true', help="launches the UI in dark mode", default=False) +parser.add_argument("--theme", type=str, help="launches the UI with light or dark theme", default=None) parser.add_argument("--use-textbox-seed", action='store_true', help="use textbox for seeds in UI (no up/down, but possible to input long seeds)", default=False) parser.add_argument("--disable-console-progressbars", action='store_true', help="do not output progressbars to console", default=False) parser.add_argument("--enable-console-prompts", action='store_true', help="print prompts to console when generating with txt2img and img2img", default=False) diff --git a/modules/ui.py b/modules/ui.py index a0cd052e..d41715fa 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1783,8 +1783,8 @@ for filename in sorted(os.listdir(jsdir)): with open(os.path.join(jsdir, filename), "r", encoding="utf8") as jsfile: javascript += f"\n" -if cmd_opts.dark_mode: - javascript += "\n\n" +if cmd_opts.theme is not None: + javascript += f"\n\n" if 'gradio_routes_templates_response' not in globals(): def template_response(*args, **kwargs): -- cgit v1.2.3 From d62ef76614624cda99d842a2900242d5b7923eda Mon Sep 17 00:00:00 2001 From: guaneec Date: Tue, 18 Oct 2022 03:09:50 +0800 Subject: Don't eat colons in booru tags --- modules/deepbooru.py | 2 -- 1 file changed, 2 deletions(-) (limited to 'modules') diff --git a/modules/deepbooru.py b/modules/deepbooru.py index 4ad334a1..de16b13f 100644 --- a/modules/deepbooru.py +++ b/modules/deepbooru.py @@ -157,8 +157,6 @@ def get_deepbooru_tags_from_model(model, tags, pil_image, threshold, deepbooru_o # sort by reverse by likelihood and normal for alpha, and format tag text as requested unsorted_tags_in_theshold.sort(key=lambda y: y[sort_ndx], reverse=(not alpha_sort)) for weight, tag in unsorted_tags_in_theshold: - # note: tag_outformat will still have a colon if include_ranks is True - tag_outformat = tag.replace(':', ' ') if use_spaces: tag_outformat = tag_outformat.replace('_', ' ') if use_escape: -- cgit v1.2.3 From f80e914ac4aa69a9783b4040813253500b34d925 Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Mon, 17 Oct 2022 19:10:36 +0000 Subject: example API working with gradio --- modules/api/api.py | 9 ++++++-- modules/api/processing.py | 56 ++++++++++++++++++++++++++++++++--------------- modules/processing.py | 22 +++++++++++++------ 3 files changed, 60 insertions(+), 27 deletions(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index fd09d352..5e86c3bf 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -23,8 +23,13 @@ class Api: app.add_api_route("/v1/txt2img", self.text2imgapi, methods=["POST"]) def text2imgapi(self, txt2imgreq: StableDiffusionProcessingAPI ): - p = StableDiffusionProcessingTxt2Img(**vars(txt2imgreq)) - p.sd_model = shared.sd_model + populate = txt2imgreq.copy(update={ # Override __init__ params + "sd_model": shared.sd_model, + "sampler_index": 0, + } + ) + p = StableDiffusionProcessingTxt2Img(**vars(populate)) + # Override object param processed = process_images(p) b64images = [] diff --git a/modules/api/processing.py b/modules/api/processing.py index e4df93c5..b6798241 100644 --- a/modules/api/processing.py +++ b/modules/api/processing.py @@ -5,6 +5,24 @@ from modules.processing import StableDiffusionProcessing, Processed, StableDiffu import inspect +API_NOT_ALLOWED = [ + "self", + "kwargs", + "sd_model", + "outpath_samples", + "outpath_grids", + "sampler_index", + "do_not_save_samples", + "do_not_save_grid", + "extra_generation_params", + "overlay_images", + "do_not_reload_embeddings", + "seed_enable_extras", + "prompt_for_display", + "sampler_noise_scheduler_override", + "ddim_discretize" +] + class ModelDef(BaseModel): """Assistance Class for Pydantic Dynamic Model Generation""" @@ -14,7 +32,7 @@ class ModelDef(BaseModel): field_value: Any -class pydanticModelGenerator: +class PydanticModelGenerator: """ Takes in created classes and stubs them out in a way FastAPI/Pydantic is happy about: source_data is a snapshot of the default values produced by the class @@ -24,30 +42,33 @@ class pydanticModelGenerator: def __init__( self, model_name: str = None, - source_data: {} = {}, - params: Dict = {}, - overrides: Dict = {}, - optionals: Dict = {}, + class_instance = None ): - def field_type_generator(k, v, overrides, optionals): - field_type = str if not overrides.get(k) else overrides[k]["type"] - if v is None: - field_type = Any - else: - field_type = type(v) + def field_type_generator(k, v): + # field_type = str if not overrides.get(k) else overrides[k]["type"] + # print(k, v.annotation, v.default) + field_type = v.annotation return Optional[field_type] + def merge_class_params(class_): + all_classes = list(filter(lambda x: x is not object, inspect.getmro(class_))) + parameters = {} + for classes in all_classes: + parameters = {**parameters, **inspect.signature(classes.__init__).parameters} + return parameters + + self._model_name = model_name - self._json_data = source_data + self._class_data = merge_class_params(class_instance) self._model_def = [ ModelDef( field=underscore(k), field_alias=k, - field_type=field_type_generator(k, v, overrides, optionals), - field_value=v + field_type=field_type_generator(k, v), + field_value=v.default ) - for (k,v) in source_data.items() if k in params + for (k,v) in self._class_data.items() if k not in API_NOT_ALLOWED ] def generate_model(self): @@ -60,8 +81,7 @@ class pydanticModelGenerator: } DynamicModel = create_model(self._model_name, **fields) DynamicModel.__config__.allow_population_by_field_name = True + DynamicModel.__config__.allow_mutation = True return DynamicModel -StableDiffusionProcessingAPI = pydanticModelGenerator("StableDiffusionProcessing", - StableDiffusionProcessing().__dict__, - inspect.signature(StableDiffusionProcessing.__init__).parameters).generate_model() +StableDiffusionProcessingAPI = PydanticModelGenerator("StableDiffusionProcessingTxt2Img", StableDiffusionProcessingTxt2Img).generate_model() diff --git a/modules/processing.py b/modules/processing.py index 4a7c6ccc..024a4fc3 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -9,6 +9,7 @@ from PIL import Image, ImageFilter, ImageOps import random import cv2 from skimage import exposure +from typing import Any, Dict, List, Optional import modules.sd_hijack from modules import devices, prompt_parser, masking, sd_samplers, lowvram @@ -51,9 +52,15 @@ def get_correct_sampler(p): return sd_samplers.samplers elif isinstance(p, modules.processing.StableDiffusionProcessingImg2Img): return sd_samplers.samplers_for_img2img + elif isinstance(p, modules.api.processing.StableDiffusionProcessingAPI): + return sd_samplers.samplers -class StableDiffusionProcessing: - def __init__(self, sd_model=None, outpath_samples=None, outpath_grids=None, prompt="", styles=None, seed=-1, subseed=-1, subseed_strength=0, seed_resize_from_h=-1, seed_resize_from_w=-1, seed_enable_extras=True, sampler_index=0, batch_size=1, n_iter=1, steps=50, cfg_scale=7.0, width=512, height=512, restore_faces=False, tiling=False, do_not_save_samples=False, do_not_save_grid=False, extra_generation_params=None, overlay_images=None, negative_prompt=None, eta=None, do_not_reload_embeddings=False): +class StableDiffusionProcessing(): + """ + The first set of paramaters: sd_models -> do_not_reload_embeddings represent the minimum required to create a StableDiffusionProcessing + + """ + def __init__(self, sd_model=None, outpath_samples=None, outpath_grids=None, prompt: str="", styles: List[str]=None, seed: int=-1, subseed: int=-1, subseed_strength: float=0, seed_resize_from_h: int=-1, seed_resize_from_w: int=-1, seed_enable_extras: bool=True, sampler_index: int=0, batch_size: int=1, n_iter: int=1, steps:int =50, cfg_scale:float=7.0, width:int=512, height:int=512, restore_faces:bool=False, tiling:bool=False, do_not_save_samples:bool=False, do_not_save_grid:bool=False, extra_generation_params: Dict[Any,Any]=None, overlay_images: Any=None, negative_prompt: str=None, eta: float =None, do_not_reload_embeddings: bool=False, denoising_strength: float = 0, ddim_discretize: str = "uniform", s_churn: float = 0.0, s_tmax: float = None, s_tmin: float = 0.0, s_noise: float = 1.0): self.sd_model = sd_model self.outpath_samples: str = outpath_samples self.outpath_grids: str = outpath_grids @@ -86,10 +93,10 @@ class StableDiffusionProcessing: self.denoising_strength: float = 0 self.sampler_noise_scheduler_override = None self.ddim_discretize = opts.ddim_discretize - self.s_churn = opts.s_churn - self.s_tmin = opts.s_tmin - self.s_tmax = float('inf') # not representable as a standard ui option - self.s_noise = opts.s_noise + self.s_churn = s_churn or opts.s_churn + self.s_tmin = s_tmin or opts.s_tmin + self.s_tmax = s_tmax or float('inf') # not representable as a standard ui option + self.s_noise = s_noise or opts.s_noise if not seed_enable_extras: self.subseed = -1 @@ -97,6 +104,7 @@ class StableDiffusionProcessing: self.seed_resize_from_h = 0 self.seed_resize_from_w = 0 + def init(self, all_prompts, all_seeds, all_subseeds): pass @@ -497,7 +505,7 @@ def process_images(p: StableDiffusionProcessing) -> Processed: class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): sampler = None - def __init__(self, enable_hr=False, denoising_strength=0.75, firstphase_width=0, firstphase_height=0, **kwargs): + def __init__(self, enable_hr: bool=False, denoising_strength: float=0.75, firstphase_width: int=0, firstphase_height: int=0, **kwargs): super().__init__(**kwargs) self.enable_hr = enable_hr self.denoising_strength = denoising_strength -- cgit v1.2.3 From 2e28c841f438b2090caac2b9a54eb62ddbda837c Mon Sep 17 00:00:00 2001 From: guaneec Date: Tue, 18 Oct 2022 03:15:41 +0800 Subject: Oops --- modules/deepbooru.py | 1 + 1 file changed, 1 insertion(+) (limited to 'modules') diff --git a/modules/deepbooru.py b/modules/deepbooru.py index de16b13f..8914662d 100644 --- a/modules/deepbooru.py +++ b/modules/deepbooru.py @@ -157,6 +157,7 @@ def get_deepbooru_tags_from_model(model, tags, pil_image, threshold, deepbooru_o # sort by reverse by likelihood and normal for alpha, and format tag text as requested unsorted_tags_in_theshold.sort(key=lambda y: y[sort_ndx], reverse=(not alpha_sort)) for weight, tag in unsorted_tags_in_theshold: + tag_outformat = tag if use_spaces: tag_outformat = tag_outformat.replace('_', ' ') if use_escape: -- cgit v1.2.3 From f29b16bad19b6332a15b2ef439864d866277fffb Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Mon, 17 Oct 2022 20:36:14 +0000 Subject: prevent API from saving --- modules/api/api.py | 2 ++ 1 file changed, 2 insertions(+) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index 5e86c3bf..ce72c5ee 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -26,6 +26,8 @@ class Api: populate = txt2imgreq.copy(update={ # Override __init__ params "sd_model": shared.sd_model, "sampler_index": 0, + "do_not_save_samples": True, + "do_not_save_grid": True } ) p = StableDiffusionProcessingTxt2Img(**vars(populate)) -- cgit v1.2.3 From c3851a853d99ad35ccedcdd8dbeb6cfbe273439b Mon Sep 17 00:00:00 2001 From: Ryan Voots Date: Mon, 17 Oct 2022 12:49:33 -0400 Subject: Re-use webui fastapi application rather than requiring one or the other, not both. --- modules/api/api.py | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index ce72c5ee..8781cd86 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -2,15 +2,13 @@ from modules.api.processing import StableDiffusionProcessingAPI from modules.processing import StableDiffusionProcessingTxt2Img, process_images import modules.shared as shared import uvicorn -from fastapi import FastAPI, Body, APIRouter +from fastapi import Body, APIRouter from fastapi.responses import JSONResponse from pydantic import BaseModel, Field, Json import json import io import base64 -app = FastAPI() - class TextToImageResponse(BaseModel): images: list[str] = Field(default=None, title="Image", description="The generated image in base64 format.") parameters: Json @@ -18,7 +16,7 @@ class TextToImageResponse(BaseModel): class Api: - def __init__(self): + def __init__(self, app): self.router = APIRouter() app.add_api_route("/v1/txt2img", self.text2imgapi, methods=["POST"]) -- cgit v1.2.3 From 247aeb3aaaf2925c7d68a9cf47c975f3e6d3dd33 Mon Sep 17 00:00:00 2001 From: Ryan Voots Date: Mon, 17 Oct 2022 12:50:45 -0400 Subject: Put API under /sdapi/ so that routing is simpler in the future. This means that one could allow access to /sdapi/ but not the webui. --- modules/api/api.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index 8781cd86..14613d8c 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -18,7 +18,7 @@ class TextToImageResponse(BaseModel): class Api: def __init__(self, app): self.router = APIRouter() - app.add_api_route("/v1/txt2img", self.text2imgapi, methods=["POST"]) + app.add_api_route("/sdapi/v1/txt2img", self.text2imgapi, methods=["POST"]) def text2imgapi(self, txt2imgreq: StableDiffusionProcessingAPI ): populate = txt2imgreq.copy(update={ # Override __init__ params -- cgit v1.2.3 From 1df3ff25e6fe2e3f308e45f7a6dd37fb4f1988e6 Mon Sep 17 00:00:00 2001 From: Ryan Voots Date: Mon, 17 Oct 2022 12:58:34 -0400 Subject: Add --nowebui as a means of disabling the webui and run on the other port --- modules/shared.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index 6c6405fd..8b436970 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -74,7 +74,8 @@ parser.add_argument("--disable-console-progressbars", action='store_true', help= parser.add_argument("--enable-console-prompts", action='store_true', help="print prompts to console when generating with txt2img and img2img", default=False) parser.add_argument('--vae-path', type=str, help='Path to Variational Autoencoders model', default=None) parser.add_argument("--disable-safe-unpickle", action='store_true', help="disable checking pytorch models for malicious code", default=False) -parser.add_argument("--api", action='store_true', help="use api=True to launch the api instead of the webui") +parser.add_argument("--api", action='store_true', help="use api=True to launch the api with the webui") +parser.add_argument("--nowebui", action='store_true', help="use api=True to launch the api instead of the webui") cmd_opts = parser.parse_args() restricted_opts = [ -- cgit v1.2.3 From 8d5d863a9d11850464fdb6b64f34602803c15ccc Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Tue, 18 Oct 2022 06:51:53 +0000 Subject: gradio and FastAPI --- modules/api/api.py | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index 14613d8c..ce98cb8c 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -16,9 +16,11 @@ class TextToImageResponse(BaseModel): class Api: - def __init__(self, app): + def __init__(self, app, queue_lock): self.router = APIRouter() - app.add_api_route("/sdapi/v1/txt2img", self.text2imgapi, methods=["POST"]) + self.app = app + self.queue_lock = queue_lock + self.app.add_api_route("/sdapi/v1/txt2img", self.text2imgapi, methods=["POST"]) def text2imgapi(self, txt2imgreq: StableDiffusionProcessingAPI ): populate = txt2imgreq.copy(update={ # Override __init__ params @@ -30,7 +32,8 @@ class Api: ) p = StableDiffusionProcessingTxt2Img(**vars(populate)) # Override object param - processed = process_images(p) + with self.queue_lock: + processed = process_images(p) b64images = [] for i in processed.images: @@ -52,5 +55,5 @@ class Api: raise NotImplementedError def launch(self, server_name, port): - app.include_router(self.router) - uvicorn.run(app, host=server_name, port=port) + self.app.include_router(self.router) + uvicorn.run(self.app, host=server_name, port=port) -- cgit v1.2.3 From 786ed499226177d71e937e0342bcb9d3b1ff260f Mon Sep 17 00:00:00 2001 From: C43H66N12O12S2 <36072735+C43H66N12O12S2@users.noreply.github.com> Date: Mon, 17 Oct 2022 19:48:39 +0300 Subject: use legacy attnblock --- modules/sd_hijack.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 984b35c4..2407a461 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -27,7 +27,7 @@ def apply_optimizations(): if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and (6, 0) <= torch.cuda.get_device_capability(shared.device) <= (9, 0)): print("Applying xformers cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.xformers_attention_forward - ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.xformers_attnblock_forward + ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.cross_attention_attnblock_forward elif cmd_opts.opt_split_attention_v1: print("Applying v1 cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.split_cross_attention_forward_v1 -- cgit v1.2.3 From 2043c4a231eef838bb15044f502b864b55885037 Mon Sep 17 00:00:00 2001 From: C43H66N12O12S2 <36072735+C43H66N12O12S2@users.noreply.github.com> Date: Mon, 17 Oct 2022 19:49:11 +0300 Subject: delete xformers attnblock --- modules/sd_hijack_optimizations.py | 12 ------------ 1 file changed, 12 deletions(-) (limited to 'modules') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 79405525..60da7459 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -292,15 +292,3 @@ def cross_attention_attnblock_forward(self, x): return h3 -def xformers_attnblock_forward(self, x): - try: - h_ = x - h_ = self.norm(h_) - q1 = self.q(h_).contiguous() - k1 = self.k(h_).contiguous() - v = self.v(h_).contiguous() - out = xformers.ops.memory_efficient_attention(q1, k1, v) - out = self.proj_out(out) - return x + out - except NotImplementedError: - return cross_attention_attnblock_forward(self, x) -- cgit v1.2.3 From 84823275e896bcc1f7cb4ce098ae3c5d05e17b9a Mon Sep 17 00:00:00 2001 From: C43H66N12O12S2 <36072735+C43H66N12O12S2@users.noreply.github.com> Date: Mon, 17 Oct 2022 22:18:59 +0300 Subject: readd xformers attnblock --- modules/sd_hijack_optimizations.py | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'modules') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 60da7459..7ebef3f0 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -292,3 +292,18 @@ def cross_attention_attnblock_forward(self, x): return h3 +def xformers_attnblock_forward(self, x): + try: + h_ = x + h_ = self.norm(h_) + q = self.q(h_) + k = self.k(h_) + v = self.v(h_) + b, c, h, w = q.shape + q, k, v = map(lambda t: rearrange(t, 'b c h w -> b (h w) c'), (q, k, v)) + out = xformers.ops.memory_efficient_attention(q, k, v) + out = rearrange(out, 'b (h w) c -> b c h w', h=h) + out = self.proj_out(out) + return x + out + except NotImplementedError: + return cross_attention_attnblock_forward(self, x) -- cgit v1.2.3 From 73b5dbf72a93b64445551c74a4c0dc924986081d Mon Sep 17 00:00:00 2001 From: C43H66N12O12S2 <36072735+C43H66N12O12S2@users.noreply.github.com> Date: Mon, 17 Oct 2022 22:19:18 +0300 Subject: Update sd_hijack.py --- modules/sd_hijack.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 2407a461..984b35c4 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -27,7 +27,7 @@ def apply_optimizations(): if cmd_opts.force_enable_xformers or (cmd_opts.xformers and shared.xformers_available and torch.version.cuda and (6, 0) <= torch.cuda.get_device_capability(shared.device) <= (9, 0)): print("Applying xformers cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.xformers_attention_forward - ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.cross_attention_attnblock_forward + ldm.modules.diffusionmodules.model.AttnBlock.forward = sd_hijack_optimizations.xformers_attnblock_forward elif cmd_opts.opt_split_attention_v1: print("Applying v1 cross attention optimization.") ldm.modules.attention.CrossAttention.forward = sd_hijack_optimizations.split_cross_attention_forward_v1 -- cgit v1.2.3 From c71008c74156635558bb2e877d1628913f6f781e Mon Sep 17 00:00:00 2001 From: C43H66N12O12S2 <36072735+C43H66N12O12S2@users.noreply.github.com> Date: Tue, 18 Oct 2022 00:02:50 +0300 Subject: Update sd_hijack_optimizations.py --- modules/sd_hijack_optimizations.py | 3 +++ 1 file changed, 3 insertions(+) (limited to 'modules') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 7ebef3f0..a3345bb9 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -301,6 +301,9 @@ def xformers_attnblock_forward(self, x): v = self.v(h_) b, c, h, w = q.shape q, k, v = map(lambda t: rearrange(t, 'b c h w -> b (h w) c'), (q, k, v)) + q = q.contiguous() + k = k.contiguous() + v = v.contiguous() out = xformers.ops.memory_efficient_attention(q, k, v) out = rearrange(out, 'b (h w) c -> b c h w', h=h) out = self.proj_out(out) -- cgit v1.2.3 From 8b02662215917d39f76f86b703a322818d5a8ad4 Mon Sep 17 00:00:00 2001 From: trufty Date: Mon, 17 Oct 2022 10:58:21 -0400 Subject: Disable auto weights swap with config option --- modules/shared.py | 1 + modules/ui.py | 4 ++++ 2 files changed, 5 insertions(+) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index 9603d26e..8a1d1881 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -266,6 +266,7 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "enable_emphasis": OptionInfo(True, "Emphasis: use (text) to make model pay more attention to text and [text] to make it pay less attention"), "use_old_emphasis_implementation": OptionInfo(False, "Use old emphasis implementation. Can be useful to reproduce old seeds."), "enable_batch_seeds": OptionInfo(True, "Make K-diffusion samplers produce same images in a batch as when making a single image"), + "disable_weights_auto_swap": OptionInfo(False, "Disable auto swapping weights to match model hash in prompts"), "comma_padding_backtrack": OptionInfo(20, "Increase coherency by padding from the last comma within n tokens when using more than 75 tokens", gr.Slider, {"minimum": 0, "maximum": 74, "step": 1 }), "filter_nsfw": OptionInfo(False, "Filter NSFW content"), 'CLIP_stop_at_last_layers': OptionInfo(1, "Stop At last layers of CLIP model", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}), diff --git a/modules/ui.py b/modules/ui.py index 1dae4a65..75eb0b0c 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -542,6 +542,10 @@ def apply_setting(key, value): if value is None: return gr.update() + # dont allow model to be swapped when model hash exists in prompt + if key == "sd_model_checkpoint" and opts.disable_weights_auto_swap: + return gr.update() + if key == "sd_model_checkpoint": ckpt_info = sd_models.get_closet_checkpoint_match(value) -- cgit v1.2.3 From d2f459c5cf9f728256775dc1c3380c7e9a7e27fb Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 18 Oct 2022 14:22:52 +0300 Subject: clarify the comment for the new option from #2959 and move it to UI section. --- modules/shared.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index 8a1d1881..c0d87168 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -266,7 +266,6 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "enable_emphasis": OptionInfo(True, "Emphasis: use (text) to make model pay more attention to text and [text] to make it pay less attention"), "use_old_emphasis_implementation": OptionInfo(False, "Use old emphasis implementation. Can be useful to reproduce old seeds."), "enable_batch_seeds": OptionInfo(True, "Make K-diffusion samplers produce same images in a batch as when making a single image"), - "disable_weights_auto_swap": OptionInfo(False, "Disable auto swapping weights to match model hash in prompts"), "comma_padding_backtrack": OptionInfo(20, "Increase coherency by padding from the last comma within n tokens when using more than 75 tokens", gr.Slider, {"minimum": 0, "maximum": 74, "step": 1 }), "filter_nsfw": OptionInfo(False, "Filter NSFW content"), 'CLIP_stop_at_last_layers': OptionInfo(1, "Stop At last layers of CLIP model", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}), @@ -294,6 +293,7 @@ options_templates.update(options_section(('ui', "User interface"), { "do_not_show_images": OptionInfo(False, "Do not show any images in results for web"), "add_model_hash_to_info": OptionInfo(True, "Add model hash to generation information"), "add_model_name_to_info": OptionInfo(False, "Add model name to generation information"), + "disable_weights_auto_swap": OptionInfo(False, "When reading generation parameters from text into UI (from PNG info or pasted text), do not change the selected model/checkpoint."), "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"), -- cgit v1.2.3 From 97d3ba3941536215ea15431886c7f28300a9d915 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E3=81=B5=E3=81=81?= <34892635+fa0311@users.noreply.github.com> Date: Tue, 18 Oct 2022 17:29:42 +0900 Subject: Add scripts to ui-config,json --- modules/scripts.py | 15 +++++++++++++-- modules/ui.py | 5 +++++ 2 files changed, 18 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/scripts.py b/modules/scripts.py index ac66d448..3402066d 100644 --- a/modules/scripts.py +++ b/modules/scripts.py @@ -96,6 +96,7 @@ def wrap_call(func, filename, funcname, *args, default=None, **kwargs): class ScriptRunner: def __init__(self): self.scripts = [] + self.titles = [] def setup_ui(self, is_img2img): for script_class, path in scripts_data: @@ -107,9 +108,10 @@ class ScriptRunner: self.scripts.append(script) - titles = [wrap_call(script.title, script.filename, "title") or f"{script.filename} [error]" for script in self.scripts] + self.titles = [wrap_call(script.title, script.filename, "title") or f"{script.filename} [error]" for script in self.scripts] - dropdown = gr.Dropdown(label="Script", choices=["None"] + titles, value="None", type="index") + dropdown = gr.Dropdown(label="Script", choices=["None"] + self.titles, value="None", type="index") + dropdown.save_to_config = True inputs = [dropdown] for script in self.scripts: @@ -139,6 +141,15 @@ class ScriptRunner: return [ui.gr_show(True if i == 0 else args_from <= i < args_to) for i in range(len(inputs))] + def init_field(title): + if title == "None": + return + script_index = self.titles.index(title) + script = self.scripts[script_index] + for i in range(script.args_from, script.args_to): + inputs[i].visible = True + + dropdown.init_field = init_field dropdown.change( fn=select_script, inputs=[dropdown], diff --git a/modules/ui.py b/modules/ui.py index 75eb0b0c..39afbc4e 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1753,6 +1753,11 @@ Requested path was: {f} print(f'Warning: Bad ui setting value: {key}: {saved_value}; Default value "{getattr(obj, field)}" will be used instead.') else: setattr(obj, field, saved_value) + if getattr(x, 'init_field', False): + try: + x.init_field(saved_value) + except Exception: + print(f'Warning: Bad ui setting value: {key}: {saved_value}; Default value "{getattr(obj, field)}" will be used instead.') if type(x) in [gr.Slider, gr.Radio, gr.Checkbox, gr.Textbox, gr.Number] and x.visible: apply_field(x, 'visible') -- cgit v1.2.3 From de29ec0743fcfb141d8891a3ccbd537ea71bf5b4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E3=81=B5=E3=81=81?= <34892635+fa0311@users.noreply.github.com> Date: Tue, 18 Oct 2022 18:15:00 +0900 Subject: Remove exception handling --- modules/ui.py | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 39afbc4e..b38bfb3f 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1754,10 +1754,7 @@ Requested path was: {f} else: setattr(obj, field, saved_value) if getattr(x, 'init_field', False): - try: - x.init_field(saved_value) - except Exception: - print(f'Warning: Bad ui setting value: {key}: {saved_value}; Default value "{getattr(obj, field)}" will be used instead.') + x.init_field(saved_value) if type(x) in [gr.Slider, gr.Radio, gr.Checkbox, gr.Textbox, gr.Number] and x.visible: apply_field(x, 'visible') -- cgit v1.2.3 From 3003438088502774628656790d83fc8074d51ab4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E3=81=B5=E3=81=81?= <34892635+fa0311@users.noreply.github.com> Date: Tue, 18 Oct 2022 18:51:57 +0900 Subject: Add visible for dropdown --- modules/ui.py | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index b38bfb3f..fb6eb5a0 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1737,7 +1737,7 @@ Requested path was: {f} print(traceback.format_exc(), file=sys.stderr) def loadsave(path, x): - def apply_field(obj, field, condition=None): + def apply_field(obj, field, condition=None, init_field=None): key = path + "/" + field if getattr(obj,'custom_script_source',None) is not None: @@ -1753,8 +1753,8 @@ Requested path was: {f} print(f'Warning: Bad ui setting value: {key}: {saved_value}; Default value "{getattr(obj, field)}" will be used instead.') else: setattr(obj, field, saved_value) - if getattr(x, 'init_field', False): - x.init_field(saved_value) + if init_field is not None: + init_field(saved_value) if type(x) in [gr.Slider, gr.Radio, gr.Checkbox, gr.Textbox, gr.Number] and x.visible: apply_field(x, 'visible') @@ -1780,7 +1780,8 @@ Requested path was: {f} # Since there are many dropdowns that shouldn't be saved, # we only mark dropdowns that should be saved. if type(x) == gr.Dropdown and getattr(x, 'save_to_config', False): - apply_field(x, 'value', lambda val: val in x.choices) + apply_field(x, 'value', lambda val: val in x.choices, getattr(x, 'init_field', None)) + apply_field(x, 'visible') visit(txt2img_interface, loadsave, "txt2img") visit(img2img_interface, loadsave, "img2img") -- cgit v1.2.3 From 02622b19191f5f5112db7633c0630e5c7df1b2f7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E3=81=B5=E3=81=81?= <34892635+fa0311@users.noreply.github.com> Date: Tue, 18 Oct 2022 18:52:27 +0900 Subject: update scripts.py --- modules/scripts.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/scripts.py b/modules/scripts.py index 3402066d..1039fa9c 100644 --- a/modules/scripts.py +++ b/modules/scripts.py @@ -142,7 +142,7 @@ class ScriptRunner: return [ui.gr_show(True if i == 0 else args_from <= i < args_to) for i in range(len(inputs))] def init_field(title): - if title == "None": + if title == 'None': return script_index = self.titles.index(title) script = self.scripts[script_index] -- cgit v1.2.3 From 4c605c5174a9b211c3a88e9aff5f5be92b53fd92 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Sun, 16 Oct 2022 17:24:06 +0100 Subject: add shared option for update check --- modules/shared.py | 1 + 1 file changed, 1 insertion(+) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index c0d87168..50dc46ae 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -76,6 +76,7 @@ parser.add_argument("--disable-console-progressbars", action='store_true', help= parser.add_argument("--enable-console-prompts", action='store_true', help="print prompts to console when generating with txt2img and img2img", default=False) parser.add_argument('--vae-path', type=str, help='Path to Variational Autoencoders model', default=None) parser.add_argument("--disable-safe-unpickle", action='store_true', help="disable checking pytorch models for malicious code", default=False) +parser.add_argument("--update-check", action='store_true', help="enable http check to confirm that the currently running version is the most recent release.", default=False) cmd_opts = parser.parse_args() restricted_opts = [ -- cgit v1.2.3 From 433a7525c1f5eb5963340e0cc45d31038ede3f7e Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 18 Oct 2022 15:18:02 +0300 Subject: remove shared option for update check (because it is not an argument of webui) have launch.py examine both COMMANDLINE_ARGS as well as argv for its arguments --- modules/shared.py | 1 - 1 file changed, 1 deletion(-) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index 50dc46ae..c0d87168 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -76,7 +76,6 @@ parser.add_argument("--disable-console-progressbars", action='store_true', help= parser.add_argument("--enable-console-prompts", action='store_true', help="print prompts to console when generating with txt2img and img2img", default=False) parser.add_argument('--vae-path', type=str, help='Path to Variational Autoencoders model', default=None) parser.add_argument("--disable-safe-unpickle", action='store_true', help="disable checking pytorch models for malicious code", default=False) -parser.add_argument("--update-check", action='store_true', help="enable http check to confirm that the currently running version is the most recent release.", default=False) cmd_opts = parser.parse_args() restricted_opts = [ -- cgit v1.2.3 From 2f448d97a9427f9a7bad19cf608561b2878ab2da Mon Sep 17 00:00:00 2001 From: w-e-w <40751091+w-e-w@users.noreply.github.com> Date: Mon, 17 Oct 2022 23:18:21 +0900 Subject: styles.csv encoding utf8 to utf-8-sig utf-8-bom for better compatibility for some programs --- modules/styles.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/styles.py b/modules/styles.py index d44dfc1a..3bf5c5b6 100644 --- a/modules/styles.py +++ b/modules/styles.py @@ -45,7 +45,7 @@ class StyleDatabase: if not os.path.exists(path): return - with open(path, "r", encoding="utf8", newline='') as file: + with open(path, "r", encoding="utf-8-sig", newline='') as file: reader = csv.DictReader(file) for row in reader: # Support loading old CSV format with "name, text"-columns @@ -79,7 +79,7 @@ class StyleDatabase: def save_styles(self, path: str) -> None: # Write to temporary file first, so we don't nuke the file if something goes wrong fd, temp_path = tempfile.mkstemp(".csv") - with os.fdopen(fd, "w", encoding="utf8", newline='') as file: + with os.fdopen(fd, "w", encoding="utf-8-sig", newline='') as file: # _fields is actually part of the public API: typing.NamedTuple is a replacement for collections.NamedTuple, # and collections.NamedTuple has explicit documentation for accessing _fields. Same goes for _asdict() writer = csv.DictWriter(file, fieldnames=PromptStyle._fields) -- cgit v1.2.3 From e20b7e30fe17744acb74ad33c87c0963525ea921 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 18 Oct 2022 15:33:24 +0300 Subject: fix for add difference model merging --- modules/extras.py | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/extras.py b/modules/extras.py index c908b43e..03f6085e 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -216,8 +216,11 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam if theta_func1: for key in tqdm.tqdm(theta_1.keys()): if 'model' in key: - t2 = theta_2.get(key, torch.zeros_like(theta_1[key])) - theta_1[key] = theta_func1(theta_1[key], t2) + if key in theta_2: + t2 = theta_2.get(key, torch.zeros_like(theta_1[key])) + theta_1[key] = theta_func1(theta_1[key], t2) + else: + theta_1[key] = 0 del theta_2, teritary_model for key in tqdm.tqdm(theta_0.keys()): -- cgit v1.2.3 From ec1924ee5789b72c31c65932b549c59ccae0cdd6 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 18 Oct 2022 16:05:52 +0300 Subject: additional fix for difference model merging --- modules/extras.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/extras.py b/modules/extras.py index 03f6085e..b853fa5b 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -220,7 +220,7 @@ def run_modelmerger(primary_model_name, secondary_model_name, teritary_model_nam t2 = theta_2.get(key, torch.zeros_like(theta_1[key])) theta_1[key] = theta_func1(theta_1[key], t2) else: - theta_1[key] = 0 + theta_1[key] = torch.zeros_like(theta_1[key]) del theta_2, teritary_model for key in tqdm.tqdm(theta_0.keys()): -- cgit v1.2.3 From cbf15edbf90a68a08eeab40af5df577ba4ac90b6 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 18 Oct 2022 17:23:38 +0300 Subject: remove dependence on TQDM for sampler progress/interrupt functionality --- modules/processing.py | 6 --- modules/sd_samplers.py | 107 +++++++++++++++++++++++++++---------------------- 2 files changed, 58 insertions(+), 55 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index deb6125e..346eea88 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -402,12 +402,6 @@ def process_images(p: StableDiffusionProcessing) -> Processed: with devices.autocast(): samples_ddim = p.sample(conditioning=c, unconditional_conditioning=uc, seeds=seeds, subseeds=subseeds, subseed_strength=p.subseed_strength) - if state.interrupted or state.skipped: - - # if we are interrupted, sample returns just noise - # use the image collected previously in sampler loop - samples_ddim = shared.state.current_latent - samples_ddim = samples_ddim.to(devices.dtype_vae) x_samples_ddim = decode_first_stage(p.sd_model, samples_ddim) x_samples_ddim = torch.clamp((x_samples_ddim + 1.0) / 2.0, min=0.0, max=1.0) diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index 20309e06..b58e810b 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -98,25 +98,8 @@ def store_latent(decoded): shared.state.current_image = sample_to_image(decoded) - -def extended_tdqm(sequence, *args, desc=None, **kwargs): - state.sampling_steps = len(sequence) - state.sampling_step = 0 - - seq = sequence if cmd_opts.disable_console_progressbars else tqdm.tqdm(sequence, *args, desc=state.job, file=shared.progress_print_out, **kwargs) - - for x in seq: - if state.interrupted or state.skipped: - break - - yield x - - state.sampling_step += 1 - shared.total_tqdm.update() - - -ldm.models.diffusion.ddim.tqdm = lambda *args, desc=None, **kwargs: extended_tdqm(*args, desc=desc, **kwargs) -ldm.models.diffusion.plms.tqdm = lambda *args, desc=None, **kwargs: extended_tdqm(*args, desc=desc, **kwargs) +class InterruptedException(BaseException): + pass class VanillaStableDiffusionSampler: @@ -128,14 +111,32 @@ class VanillaStableDiffusionSampler: self.init_latent = None self.sampler_noises = None self.step = 0 + self.stop_at = None self.eta = None self.default_eta = 0.0 self.config = None + self.last_latent = None def number_of_needed_noises(self, p): return 0 + def launch_sampling(self, steps, func): + state.sampling_steps = steps + state.sampling_step = 0 + + try: + return func() + except InterruptedException: + return self.last_latent + def p_sample_ddim_hook(self, x_dec, cond, ts, unconditional_conditioning, *args, **kwargs): + if state.interrupted or state.skipped: + raise InterruptedException + + if self.stop_at is not None and self.step > self.stop_at: + raise InterruptedException + + conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step) unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step) @@ -159,11 +160,16 @@ class VanillaStableDiffusionSampler: res = self.orig_p_sample_ddim(x_dec, cond, ts, unconditional_conditioning=unconditional_conditioning, *args, **kwargs) if self.mask is not None: - store_latent(self.init_latent * self.mask + self.nmask * res[1]) + self.last_latent = self.init_latent * self.mask + self.nmask * res[1] else: - store_latent(res[1]) + self.last_latent = res[1] + + store_latent(self.last_latent) self.step += 1 + state.sampling_step = self.step + shared.total_tqdm.update() + return res def initialize(self, p): @@ -192,7 +198,7 @@ class VanillaStableDiffusionSampler: self.init_latent = x self.step = 0 - samples = self.sampler.decode(x1, conditioning, t_enc, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning) + samples = self.launch_sampling(steps, lambda: self.sampler.decode(x1, conditioning, t_enc, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning)) return samples @@ -206,9 +212,9 @@ class VanillaStableDiffusionSampler: # existing code fails with certain step counts, like 9 try: - samples_ddim, _ = self.sampler.sample(S=steps, conditioning=conditioning, batch_size=int(x.shape[0]), shape=x[0].shape, verbose=False, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning, x_T=x, eta=self.eta) + samples_ddim = self.launch_sampling(steps, lambda: self.sampler.sample(S=steps, conditioning=conditioning, batch_size=int(x.shape[0]), shape=x[0].shape, verbose=False, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning, x_T=x, eta=self.eta)[0]) except Exception: - samples_ddim, _ = self.sampler.sample(S=steps+1, conditioning=conditioning, batch_size=int(x.shape[0]), shape=x[0].shape, verbose=False, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning, x_T=x, eta=self.eta) + samples_ddim = self.launch_sampling(steps, lambda: self.sampler.sample(S=steps+1, conditioning=conditioning, batch_size=int(x.shape[0]), shape=x[0].shape, verbose=False, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning, x_T=x, eta=self.eta)[0]) return samples_ddim @@ -223,6 +229,9 @@ class CFGDenoiser(torch.nn.Module): self.step = 0 def forward(self, x, sigma, uncond, cond, cond_scale): + if state.interrupted or state.skipped: + raise InterruptedException + conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step) uncond = prompt_parser.reconstruct_cond_batch(uncond, self.step) @@ -268,25 +277,6 @@ class CFGDenoiser(torch.nn.Module): return denoised -def extended_trange(sampler, count, *args, **kwargs): - state.sampling_steps = count - state.sampling_step = 0 - - seq = range(count) if cmd_opts.disable_console_progressbars else tqdm.trange(count, *args, desc=state.job, file=shared.progress_print_out, **kwargs) - - for x in seq: - if state.interrupted or state.skipped: - break - - if sampler.stop_at is not None and x > sampler.stop_at: - break - - yield x - - state.sampling_step += 1 - shared.total_tqdm.update() - - class TorchHijack: def __init__(self, kdiff_sampler): self.kdiff_sampler = kdiff_sampler @@ -314,9 +304,28 @@ class KDiffusionSampler: self.eta = None self.default_eta = 1.0 self.config = None + self.last_latent = None def callback_state(self, d): - store_latent(d["denoised"]) + step = d['i'] + latent = d["denoised"] + store_latent(latent) + self.last_latent = latent + + if self.stop_at is not None and step > self.stop_at: + raise InterruptedException + + state.sampling_step = step + shared.total_tqdm.update() + + def launch_sampling(self, steps, func): + state.sampling_steps = steps + state.sampling_step = 0 + + try: + return func() + except InterruptedException: + return self.last_latent def number_of_needed_noises(self, p): return p.steps @@ -339,9 +348,6 @@ class KDiffusionSampler: self.sampler_noise_index = 0 self.eta = p.eta or opts.eta_ancestral - if hasattr(k_diffusion.sampling, 'trange'): - k_diffusion.sampling.trange = lambda *args, **kwargs: extended_trange(self, *args, **kwargs) - if self.sampler_noises is not None: k_diffusion.sampling.torch = TorchHijack(self) @@ -383,8 +389,9 @@ class KDiffusionSampler: self.model_wrap_cfg.init_latent = x - return self.func(self.model_wrap_cfg, xi, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs) + samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, xi, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs)) + return samples def sample(self, p, x, conditioning, unconditional_conditioning, steps=None): steps = steps or p.steps @@ -406,6 +413,8 @@ class KDiffusionSampler: extra_params_kwargs['n'] = steps else: extra_params_kwargs['sigmas'] = sigmas - samples = self.func(self.model_wrap_cfg, x, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs) + + samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs)) + return samples -- cgit v1.2.3 From 6021f7a75f7b5208a2be15cda5526028152f922d Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 00:51:36 +0900 Subject: add options to custom hypernetwork layer structure --- modules/hypernetworks/hypernetwork.py | 88 ++++++++++++++++++++++++++--------- modules/shared.py | 4 +- 2 files changed, 70 insertions(+), 22 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 4905710e..cadb9911 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -1,52 +1,98 @@ +import csv import datetime import glob import html import os import sys import traceback -import tqdm -import csv +import modules.textual_inversion.dataset import torch - -from ldm.util import default -from modules import devices, shared, processing, sd_models -import torch -from torch import einsum +import tqdm from einops import rearrange, repeat -import modules.textual_inversion.dataset +from ldm.util import default +from modules import devices, processing, sd_models, shared from modules.textual_inversion import textual_inversion from modules.textual_inversion.learn_schedule import LearnRateScheduler +from torch import einsum + + +def parse_layer_structure(dim, state_dict): + i = 0 + res = [1] + while (key := "linear.{}.weight".format(i)) in state_dict: + weight = state_dict[key] + res.append(len(weight) // dim) + i += 1 + return res class HypernetworkModule(torch.nn.Module): multiplier = 1.0 + layer_structure = None + add_layer_norm = False def __init__(self, dim, state_dict=None): super().__init__() + if (state_dict is None or 'linear.0.weight' not in state_dict) and self.layer_structure is None: + layer_structure = (1, 2, 1) + else: + if self.layer_structure is not None: + assert self.layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" + assert self.layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" + layer_structure = self.layer_structure + else: + layer_structure = parse_layer_structure(dim, state_dict) + + linears = [] + for i in range(len(layer_structure) - 1): + linears.append(torch.nn.Linear(int(dim * layer_structure[i]), int(dim * layer_structure[i+1]))) + if self.add_layer_norm: + linears.append(torch.nn.LayerNorm(int(dim * layer_structure[i+1]))) - self.linear1 = torch.nn.Linear(dim, dim * 2) - self.linear2 = torch.nn.Linear(dim * 2, dim) + self.linear = torch.nn.Sequential(*linears) if state_dict is not None: - self.load_state_dict(state_dict, strict=True) + try: + self.load_state_dict(state_dict) + except RuntimeError: + self.try_load_previous(state_dict) else: - - self.linear1.weight.data.normal_(mean=0.0, std=0.01) - self.linear1.bias.data.zero_() - self.linear2.weight.data.normal_(mean=0.0, std=0.01) - self.linear2.bias.data.zero_() + for layer in self.linear: + layer.weight.data.normal_(mean = 0.0, std = 0.01) + layer.bias.data.zero_() self.to(devices.device) + def try_load_previous(self, state_dict): + states = self.state_dict() + states['linear.0.bias'].copy_(state_dict['linear1.bias']) + states['linear.0.weight'].copy_(state_dict['linear1.weight']) + states['linear.1.bias'].copy_(state_dict['linear2.bias']) + states['linear.1.weight'].copy_(state_dict['linear2.weight']) + def forward(self, x): - return x + (self.linear2(self.linear1(x))) * self.multiplier + return x + self.linear(x) * self.multiplier + + def trainables(self): + res = [] + for layer in self.linear: + res += [layer.weight, layer.bias] + return res def apply_strength(value=None): HypernetworkModule.multiplier = value if value is not None else shared.opts.sd_hypernetwork_strength +def apply_layer_structure(value=None): + HypernetworkModule.layer_structure = value if value is not None else shared.opts.sd_hypernetwork_layer_structure + + +def apply_layer_norm(value=None): + HypernetworkModule.add_layer_norm = value if value is not None else shared.opts.sd_hypernetwork_add_layer_norm + + class Hypernetwork: filename = None name = None @@ -68,7 +114,7 @@ class Hypernetwork: for k, layers in self.layers.items(): for layer in layers: layer.train() - res += [layer.linear1.weight, layer.linear1.bias, layer.linear2.weight, layer.linear2.bias] + res += layer.trainables() return res @@ -226,7 +272,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..." with torch.autocast("cuda"): ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size) - + assert ds.length > 1, "Dataset should contain more than 1 images" if unload: shared.sd_model.cond_stage_model.to(devices.cpu) shared.sd_model.first_stage_model.to(devices.cpu) @@ -261,7 +307,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log with torch.autocast("cuda"): c = stack_conds([entry.cond for entry in entries]).to(devices.device) -# c = torch.vstack([entry.cond for entry in entries]).to(devices.device) + c = torch.vstack([entry.cond for entry in entries]).to(devices.device) x = torch.stack([entry.latent for entry in entries]).to(devices.device) loss = shared.sd_model(x, c)[0] del x @@ -283,7 +329,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log textual_inversion.write_loss(log_directory, "hypernetwork_loss.csv", hypernetwork.step, len(ds), { "loss": f"{mean_loss:.7f}", - "learn_rate": scheduler.learn_rate + "learn_rate": f"{scheduler.learn_rate:.7f}" }) if hypernetwork.step > 0 and images_dir is not None and hypernetwork.step % create_image_every == 0: diff --git a/modules/shared.py b/modules/shared.py index c0d87168..c87ce70e 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -13,7 +13,7 @@ import modules.memmon import modules.sd_models import modules.styles import modules.devices as devices -from modules import sd_samplers, sd_models, localization +from modules import sd_models, sd_samplers, localization from modules.hypernetworks import hypernetwork from modules.paths import models_path, script_path, sd_path @@ -258,6 +258,8 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "sd_model_checkpoint": OptionInfo(None, "Stable Diffusion checkpoint", gr.Dropdown, lambda: {"choices": modules.sd_models.checkpoint_tiles()}, refresh=sd_models.list_models), "sd_checkpoint_cache": OptionInfo(0, "Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}), "sd_hypernetwork": OptionInfo("None", "Hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}, refresh=reload_hypernetworks), + "sd_hypernetwork_layer_structure": OptionInfo(None, "Hypernetwork layer structure Default: (1,2,1).", gr.Dropdown, lambda: {"choices": [(1, 2, 1), (1, 2, 2, 1), (1, 2, 4, 2, 1)]}), + "sd_hypernetwork_add_layer_norm": OptionInfo(False, "Add layer normalization to hypernetwork architecture."), "sd_hypernetwork_strength": OptionInfo(1.0, "Hypernetwork strength", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.001}), "img2img_color_correction": OptionInfo(False, "Apply color correction to img2img results to match original colors."), "save_images_before_color_correction": OptionInfo(False, "Save a copy of image before applying color correction to img2img results"), -- cgit v1.2.3 From a5611ea5026bd8e12d8e84023384c369d0511dda Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 01:00:01 +0900 Subject: update --- modules/hypernetworks/hypernetwork.py | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index cadb9911..c5835bce 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -1,20 +1,22 @@ -import csv import datetime import glob import html import os import sys import traceback +import tqdm +import csv -import modules.textual_inversion.dataset import torch -import tqdm -from einops import rearrange, repeat + from ldm.util import default -from modules import devices, processing, sd_models, shared +from modules import devices, shared, processing, sd_models +import torch +from torch import einsum +from einops import rearrange, repeat +import modules.textual_inversion.dataset from modules.textual_inversion import textual_inversion from modules.textual_inversion.learn_schedule import LearnRateScheduler -from torch import einsum def parse_layer_structure(dim, state_dict): -- cgit v1.2.3 From 7f2095c6c8db82a5c9cd7c7177f6ba856a2cc676 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 01:01:22 +0900 Subject: update --- modules/shared.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/shared.py b/modules/shared.py index c87ce70e..6b6d5c41 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -13,7 +13,7 @@ import modules.memmon import modules.sd_models import modules.styles import modules.devices as devices -from modules import sd_models, sd_samplers, localization +from modules import sd_samplers, sd_models, localization from modules.hypernetworks import hypernetwork from modules.paths import models_path, script_path, sd_path @@ -135,7 +135,7 @@ class State: self.job_no += 1 self.sampling_step = 0 self.current_image_sampling_step = 0 - + def get_job_timestamp(self): return datetime.datetime.now().strftime("%Y%m%d%H%M%S") # shouldn't this return job_timestamp? -- cgit v1.2.3 From e40ba281f1b419cf99552962ea01d87d699840a5 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 01:03:58 +0900 Subject: update --- modules/hypernetworks/hypernetwork.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index c5835bce..082165f4 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -309,7 +309,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log with torch.autocast("cuda"): c = stack_conds([entry.cond for entry in entries]).to(devices.device) - c = torch.vstack([entry.cond for entry in entries]).to(devices.device) + # c = torch.vstack([entry.cond for entry in entries]).to(devices.device) x = torch.stack([entry.latent for entry in entries]).to(devices.device) loss = shared.sd_model(x, c)[0] del x @@ -331,7 +331,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log textual_inversion.write_loss(log_directory, "hypernetwork_loss.csv", hypernetwork.step, len(ds), { "loss": f"{mean_loss:.7f}", - "learn_rate": f"{scheduler.learn_rate:.7f}" + "learn_rate": scheduler.learn_rate }) if hypernetwork.step > 0 and images_dir is not None and hypernetwork.step % create_image_every == 0: -- cgit v1.2.3 From e7f4808505f7a6339927c32b9a0c01bc9134bdeb Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Tue, 18 Oct 2022 19:04:56 +0000 Subject: provide sampler by name --- modules/api/api.py | 12 ++++++++++-- modules/api/processing.py | 16 ++++++++++++++-- 2 files changed, 24 insertions(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index ce98cb8c..ff9df0d1 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -1,14 +1,17 @@ from modules.api.processing import StableDiffusionProcessingAPI from modules.processing import StableDiffusionProcessingTxt2Img, process_images +from modules.sd_samplers import samplers_k_diffusion import modules.shared as shared import uvicorn -from fastapi import Body, APIRouter +from fastapi import Body, APIRouter, HTTPException from fastapi.responses import JSONResponse from pydantic import BaseModel, Field, Json import json import io import base64 +sampler_to_index = lambda name: next(filter(lambda row: name in row[1][2], enumerate(samplers_k_diffusion)), None) + class TextToImageResponse(BaseModel): images: list[str] = Field(default=None, title="Image", description="The generated image in base64 format.") parameters: Json @@ -23,9 +26,14 @@ class Api: self.app.add_api_route("/sdapi/v1/txt2img", self.text2imgapi, methods=["POST"]) def text2imgapi(self, txt2imgreq: StableDiffusionProcessingAPI ): + sampler_index = sampler_to_index(txt2imgreq.sampler_index) + + if sampler_index is None: + raise HTTPException(status_code=404, detail="Sampler not found") + populate = txt2imgreq.copy(update={ # Override __init__ params "sd_model": shared.sd_model, - "sampler_index": 0, + "sampler_index": sampler_index[0], "do_not_save_samples": True, "do_not_save_grid": True } diff --git a/modules/api/processing.py b/modules/api/processing.py index b6798241..2e6483ee 100644 --- a/modules/api/processing.py +++ b/modules/api/processing.py @@ -42,7 +42,8 @@ class PydanticModelGenerator: def __init__( self, model_name: str = None, - class_instance = None + class_instance = None, + additional_fields = None, ): def field_type_generator(k, v): # field_type = str if not overrides.get(k) else overrides[k]["type"] @@ -70,6 +71,13 @@ class PydanticModelGenerator: ) for (k,v) in self._class_data.items() if k not in API_NOT_ALLOWED ] + + for fields in additional_fields: + self._model_def.append(ModelDef( + field=underscore(fields["key"]), + field_alias=fields["key"], + field_type=fields["type"], + field_value=fields["default"])) def generate_model(self): """ @@ -84,4 +92,8 @@ class PydanticModelGenerator: DynamicModel.__config__.allow_mutation = True return DynamicModel -StableDiffusionProcessingAPI = PydanticModelGenerator("StableDiffusionProcessingTxt2Img", StableDiffusionProcessingTxt2Img).generate_model() +StableDiffusionProcessingAPI = PydanticModelGenerator( + "StableDiffusionProcessingTxt2Img", + StableDiffusionProcessingTxt2Img, + [{"key": "sampler_index", "type": str, "default": "k_euler_a"}] +).generate_model() -- cgit v1.2.3 From 0f0d6ab8e06898ce066251fc769fe14e77e98ced Mon Sep 17 00:00:00 2001 From: arcticfaded Date: Wed, 19 Oct 2022 05:19:01 +0000 Subject: call sampler by name --- modules/api/api.py | 11 ++++++----- modules/api/processing.py | 6 +++--- 2 files changed, 9 insertions(+), 8 deletions(-) (limited to 'modules') diff --git a/modules/api/api.py b/modules/api/api.py index ff9df0d1..5b0c934e 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -1,6 +1,7 @@ from modules.api.processing import StableDiffusionProcessingAPI from modules.processing import StableDiffusionProcessingTxt2Img, process_images -from modules.sd_samplers import samplers_k_diffusion +from modules.sd_samplers import all_samplers +from modules.extras import run_pnginfo import modules.shared as shared import uvicorn from fastapi import Body, APIRouter, HTTPException @@ -10,7 +11,7 @@ import json import io import base64 -sampler_to_index = lambda name: next(filter(lambda row: name in row[1][2], enumerate(samplers_k_diffusion)), None) +sampler_to_index = lambda name: next(filter(lambda row: name.lower() == row[1].name.lower(), enumerate(all_samplers)), None) class TextToImageResponse(BaseModel): images: list[str] = Field(default=None, title="Image", description="The generated image in base64 format.") @@ -53,13 +54,13 @@ class Api: - def img2imgendoint(self): + def img2imgapi(self): raise NotImplementedError - def extrasendoint(self): + def extrasapi(self): raise NotImplementedError - def pnginfoendoint(self): + def pnginfoapi(self): raise NotImplementedError def launch(self, server_name, port): diff --git a/modules/api/processing.py b/modules/api/processing.py index 2e6483ee..4c541241 100644 --- a/modules/api/processing.py +++ b/modules/api/processing.py @@ -1,7 +1,7 @@ from inflection import underscore from typing import Any, Dict, Optional from pydantic import BaseModel, Field, create_model -from modules.processing import StableDiffusionProcessing, Processed, StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img, process_images +from modules.processing import StableDiffusionProcessingTxt2Img import inspect @@ -95,5 +95,5 @@ class PydanticModelGenerator: StableDiffusionProcessingAPI = PydanticModelGenerator( "StableDiffusionProcessingTxt2Img", StableDiffusionProcessingTxt2Img, - [{"key": "sampler_index", "type": str, "default": "k_euler_a"}] -).generate_model() + [{"key": "sampler_index", "type": str, "default": "Euler"}] +).generate_model() \ No newline at end of file -- cgit v1.2.3 From 10aca1ca3e81e69e08f556a500c3dc603451429b Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Wed, 19 Oct 2022 08:42:22 +0300 Subject: more careful loading of model weights (eliminates some issues with checkpoints that have weird cond_stage_model layer names) --- modules/sd_models.py | 28 +++++++++++++++++++++++++--- 1 file changed, 25 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/sd_models.py b/modules/sd_models.py index 3aa21ec1..7ad6d474 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -122,11 +122,33 @@ def select_checkpoint(): return checkpoint_info +chckpoint_dict_replacements = { + 'cond_stage_model.transformer.embeddings.': 'cond_stage_model.transformer.text_model.embeddings.', + 'cond_stage_model.transformer.encoder.': 'cond_stage_model.transformer.text_model.encoder.', + 'cond_stage_model.transformer.final_layer_norm.': 'cond_stage_model.transformer.text_model.final_layer_norm.', +} + + +def transform_checkpoint_dict_key(k): + for text, replacement in chckpoint_dict_replacements.items(): + if k.startswith(text): + k = replacement + k[len(text):] + + return k + + def get_state_dict_from_checkpoint(pl_sd): if "state_dict" in pl_sd: - return pl_sd["state_dict"] + pl_sd = pl_sd["state_dict"] + + sd = {} + for k, v in pl_sd.items(): + new_key = transform_checkpoint_dict_key(k) + + if new_key is not None: + sd[new_key] = v - return pl_sd + return sd def load_model_weights(model, checkpoint_info): @@ -141,7 +163,7 @@ def load_model_weights(model, checkpoint_info): print(f"Global Step: {pl_sd['global_step']}") sd = get_state_dict_from_checkpoint(pl_sd) - model.load_state_dict(sd, strict=False) + missing, extra = model.load_state_dict(sd, strict=False) if shared.cmd_opts.opt_channelslast: model.to(memory_format=torch.channels_last) -- cgit v1.2.3 From da72becb13e4b750fbcb3d158c3f843311ef9938 Mon Sep 17 00:00:00 2001 From: Silent <16026653+s-ilent@users.noreply.github.com> Date: Wed, 19 Oct 2022 16:14:33 +1030 Subject: Use training width/height when training hypernetworks. --- modules/hypernetworks/hypernetwork.py | 4 ++-- modules/ui.py | 2 ++ 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 4905710e..b8695fc1 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -196,7 +196,7 @@ def stack_conds(conds): return torch.stack(conds) -def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height): +def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height): assert hypernetwork_name, 'hypernetwork not selected' path = shared.hypernetworks.get(hypernetwork_name, None) @@ -225,7 +225,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log shared.state.textinfo = f"Preparing dataset from {html.escape(data_root)}..." with torch.autocast("cuda"): - ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size) + ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=training_width, height=training_height, repeats=shared.opts.training_image_repeats_per_epoch, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True, batch_size=batch_size) if unload: shared.sd_model.cond_stage_model.to(devices.cpu) diff --git a/modules/ui.py b/modules/ui.py index fb6eb5a0..ca46343f 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1341,6 +1341,8 @@ def create_ui(wrap_gradio_gpu_call): batch_size, dataset_directory, log_directory, + training_width, + training_height, steps, create_image_every, save_embedding_every, -- cgit v1.2.3 From 2fd7935ef4ed296db5dfd8c7fea99244816f8cf0 Mon Sep 17 00:00:00 2001 From: Cheka Date: Tue, 18 Oct 2022 20:28:28 -0300 Subject: Remove wrong self reference in CUDA support for invokeai --- modules/sd_hijack_optimizations.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index a3345bb9..98123fbf 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -181,7 +181,7 @@ def einsum_op_cuda(q, k, v): mem_free_torch = mem_reserved - mem_active mem_free_total = mem_free_cuda + mem_free_torch # Divide factor of safety as there's copying and fragmentation - return self.einsum_op_tensor_mem(q, k, v, mem_free_total / 3.3 / (1 << 20)) + 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': -- cgit v1.2.3 From bcfbb33e50a48b237d8d961cc2be038db53774d5 Mon Sep 17 00:00:00 2001 From: Anastasius Date: Mon, 17 Oct 2022 13:35:20 -0700 Subject: Added time left estimation --- modules/ui.py | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index ca46343f..9a54aa16 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -261,6 +261,15 @@ def wrap_gradio_call(func, extra_outputs=None): return f +def calc_time_left(progress): + if progress == 0: + return "N/A" + else: + time_since_start = time.time() - shared.state.time_start + eta = (time_since_start/progress) + return time.strftime('%H:%M:%S', time.gmtime(eta-time_since_start)) + + def check_progress_call(id_part): if shared.state.job_count == 0: return "", gr_show(False), gr_show(False), gr_show(False) @@ -272,11 +281,13 @@ def check_progress_call(id_part): if shared.state.sampling_steps > 0: progress += 1 / shared.state.job_count * shared.state.sampling_step / shared.state.sampling_steps + time_left = calc_time_left( progress ) + progress = min(progress, 1) progressbar = "" if opts.show_progressbar: - progressbar = f"""
{str(int(progress*100))+"%" if progress > 0.01 else ""}
""" + progressbar = f"""
{str(int(progress*100))+"% ETA:"+time_left if progress > 0.01 else ""}
""" image = gr_show(False) preview_visibility = gr_show(False) @@ -308,6 +319,7 @@ def check_progress_call_initial(id_part): shared.state.current_latent = None shared.state.current_image = None shared.state.textinfo = None + shared.state.time_start = time.time() return check_progress_call(id_part) -- cgit v1.2.3 From 442dbedc159bb7e9cf94f0c3626f8a409e0a50eb Mon Sep 17 00:00:00 2001 From: Anastasius Date: Tue, 18 Oct 2022 10:38:07 -0700 Subject: Estimated time displayed if jobs take more 60 sec --- modules/ui.py | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 9a54aa16..fa54110b 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -261,13 +261,17 @@ def wrap_gradio_call(func, extra_outputs=None): return f -def calc_time_left(progress): +def calc_time_left(progress, threshold, label, force_display): if progress == 0: - return "N/A" + return "" else: time_since_start = time.time() - shared.state.time_start eta = (time_since_start/progress) - return time.strftime('%H:%M:%S', time.gmtime(eta-time_since_start)) + eta_relative = eta-time_since_start + if eta_relative > threshold or force_display: + return label + time.strftime('%H:%M:%S', time.gmtime(eta_relative)) + else: + return "" def check_progress_call(id_part): @@ -281,13 +285,15 @@ def check_progress_call(id_part): if shared.state.sampling_steps > 0: progress += 1 / shared.state.job_count * shared.state.sampling_step / shared.state.sampling_steps - time_left = calc_time_left( progress ) + time_left = calc_time_left( progress, 60, " ETA:", shared.state.time_left_force_display ) + if time_left != "": + shared.state.time_left_force_display = True progress = min(progress, 1) progressbar = "" if opts.show_progressbar: - progressbar = f"""
{str(int(progress*100))+"% ETA:"+time_left if progress > 0.01 else ""}
""" + progressbar = f"""
{str(int(progress*100))+"%"+time_left if progress > 0.01 else ""}
""" image = gr_show(False) preview_visibility = gr_show(False) @@ -320,6 +326,7 @@ def check_progress_call_initial(id_part): shared.state.current_image = None shared.state.textinfo = None shared.state.time_start = time.time() + shared.state.time_left_force_display = False return check_progress_call(id_part) -- cgit v1.2.3 From 1d4aa376e6111e90888a30ae24d2bcd7f978ec51 Mon Sep 17 00:00:00 2001 From: Anastasius Date: Tue, 18 Oct 2022 12:42:39 -0700 Subject: Predictable long operation check for time estimation --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index fa54110b..38ba1138 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -268,7 +268,7 @@ def calc_time_left(progress, threshold, label, force_display): time_since_start = time.time() - shared.state.time_start eta = (time_since_start/progress) eta_relative = eta-time_since_start - if eta_relative > threshold or force_display: + if (eta_relative > threshold and progress > 0.02) or force_display: return label + time.strftime('%H:%M:%S', time.gmtime(eta_relative)) else: return "" -- cgit v1.2.3 From bb0e7232b301d1706bbd0e09367dece3bb7ac07c Mon Sep 17 00:00:00 2001 From: Ikko Ashimine Date: Wed, 19 Oct 2022 02:18:56 +0900 Subject: Fix typo in prompt_parser.py assoicated -> associated --- modules/prompt_parser.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/prompt_parser.py b/modules/prompt_parser.py index 919d5d31..f70872c4 100644 --- a/modules/prompt_parser.py +++ b/modules/prompt_parser.py @@ -275,7 +275,7 @@ re_attention = re.compile(r""" def parse_prompt_attention(text): """ - Parses a string with attention tokens and returns a list of pairs: text and its assoicated weight. + Parses a string with attention tokens and returns a list of pairs: text and its associated weight. Accepted tokens are: (abc) - increases attention to abc by a multiplier of 1.1 (abc:3.12) - increases attention to abc by a multiplier of 3.12 -- cgit v1.2.3 From f894dd552f68bea27476f1f360ab8e79f3a65b4f Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Wed, 19 Oct 2022 12:45:30 +0300 Subject: fix for broken checkpoint merger --- modules/sd_models.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/sd_models.py b/modules/sd_models.py index 7ad6d474..eae22e87 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -148,7 +148,10 @@ def get_state_dict_from_checkpoint(pl_sd): if new_key is not None: sd[new_key] = v - return sd + pl_sd.clear() + pl_sd.update(sd) + + return pl_sd def load_model_weights(model, checkpoint_info): -- cgit v1.2.3 From 42fbda83bb9830af18187fddb50c1bedd01da502 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 14:30:33 +0000 Subject: layer options moves into create hnet ui --- modules/hypernetworks/hypernetwork.py | 64 +++++++++++++++++------------------ modules/hypernetworks/ui.py | 9 +++-- modules/shared.py | 2 -- modules/ui.py | 8 +++-- 4 files changed, 45 insertions(+), 38 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 583ada31..7d519cd9 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -19,37 +19,21 @@ from modules.textual_inversion import textual_inversion from modules.textual_inversion.learn_schedule import LearnRateScheduler -def parse_layer_structure(dim, state_dict): - i = 0 - res = [1] - while (key := "linear.{}.weight".format(i)) in state_dict: - weight = state_dict[key] - res.append(len(weight) // dim) - i += 1 - return res - - class HypernetworkModule(torch.nn.Module): multiplier = 1.0 - layer_structure = None - add_layer_norm = False - def __init__(self, dim, state_dict=None): + def __init__(self, dim, state_dict=None, layer_structure=None, add_layer_norm=False): super().__init__() - if (state_dict is None or 'linear.0.weight' not in state_dict) and self.layer_structure is None: - layer_structure = (1, 2, 1) + if layer_structure is not None: + assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" + assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" else: - if self.layer_structure is not None: - assert self.layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" - assert self.layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" - layer_structure = self.layer_structure - else: - layer_structure = parse_layer_structure(dim, state_dict) + layer_structure = parse_layer_structure(dim, state_dict) linears = [] for i in range(len(layer_structure) - 1): linears.append(torch.nn.Linear(int(dim * layer_structure[i]), int(dim * layer_structure[i+1]))) - if self.add_layer_norm: + if add_layer_norm: linears.append(torch.nn.LayerNorm(int(dim * layer_structure[i+1]))) self.linear = torch.nn.Sequential(*linears) @@ -77,38 +61,47 @@ class HypernetworkModule(torch.nn.Module): return x + self.linear(x) * self.multiplier def trainables(self): - res = [] + layer_structure = [] for layer in self.linear: - res += [layer.weight, layer.bias] - return res + layer_structure += [layer.weight, layer.bias] + return layer_structure def apply_strength(value=None): HypernetworkModule.multiplier = value if value is not None else shared.opts.sd_hypernetwork_strength -def apply_layer_structure(value=None): - HypernetworkModule.layer_structure = value if value is not None else shared.opts.sd_hypernetwork_layer_structure +def parse_layer_structure(dim, state_dict): + i = 0 + layer_structure = [1] + while (key := "linear.{}.weight".format(i)) in state_dict: + weight = state_dict[key] + layer_structure.append(len(weight) // dim) + i += 1 -def apply_layer_norm(value=None): - HypernetworkModule.add_layer_norm = value if value is not None else shared.opts.sd_hypernetwork_add_layer_norm + return layer_structure class Hypernetwork: filename = None name = None - def __init__(self, name=None, enable_sizes=None): + def __init__(self, name=None, enable_sizes=None, layer_structure=None, add_layer_norm=False): self.filename = None self.name = name self.layers = {} self.step = 0 self.sd_checkpoint = None self.sd_checkpoint_name = None + self.layer_structure = layer_structure + self.add_layer_norm = add_layer_norm for size in enable_sizes or []: - self.layers[size] = (HypernetworkModule(size), HypernetworkModule(size)) + self.layers[size] = ( + HypernetworkModule(size, None, self.layer_structure, self.add_layer_norm), + HypernetworkModule(size, None, self.layer_structure, self.add_layer_norm), + ) def weights(self): res = [] @@ -128,6 +121,8 @@ class Hypernetwork: state_dict['step'] = self.step state_dict['name'] = self.name + state_dict['layer_structure'] = self.layer_structure + state_dict['is_layer_norm'] = self.add_layer_norm state_dict['sd_checkpoint'] = self.sd_checkpoint state_dict['sd_checkpoint_name'] = self.sd_checkpoint_name @@ -142,10 +137,15 @@ class Hypernetwork: for size, sd in state_dict.items(): if type(size) == int: - self.layers[size] = (HypernetworkModule(size, sd[0]), HypernetworkModule(size, sd[1])) + self.layers[size] = ( + HypernetworkModule(size, sd[0], state_dict["layer_structure"], state_dict["is_layer_norm"]), + HypernetworkModule(size, sd[1], state_dict["layer_structure"], state_dict["is_layer_norm"]), + ) self.name = state_dict.get('name', self.name) self.step = state_dict.get('step', 0) + self.layer_structure = state_dict.get('layer_structure', None) + self.add_layer_norm = state_dict.get('is_layer_norm', False) self.sd_checkpoint = state_dict.get('sd_checkpoint', None) self.sd_checkpoint_name = state_dict.get('sd_checkpoint_name', None) diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index dfa599af..7e8ea95e 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -9,11 +9,16 @@ from modules import sd_hijack, shared, devices from modules.hypernetworks import hypernetwork -def create_hypernetwork(name, enable_sizes): +def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm=False): fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") assert not os.path.exists(fn), f"file {fn} already exists" - hypernet = modules.hypernetworks.hypernetwork.Hypernetwork(name=name, enable_sizes=[int(x) for x in enable_sizes]) + hypernet = modules.hypernetworks.hypernetwork.Hypernetwork( + name=name, + enable_sizes=[int(x) for x in enable_sizes], + layer_structure=layer_structure, + add_layer_norm=add_layer_norm, + ) hypernet.save(fn) shared.reload_hypernetworks() diff --git a/modules/shared.py b/modules/shared.py index 0540cae9..faede821 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -260,8 +260,6 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "sd_model_checkpoint": OptionInfo(None, "Stable Diffusion checkpoint", gr.Dropdown, lambda: {"choices": modules.sd_models.checkpoint_tiles()}, refresh=sd_models.list_models), "sd_checkpoint_cache": OptionInfo(0, "Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}), "sd_hypernetwork": OptionInfo("None", "Hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}, refresh=reload_hypernetworks), - "sd_hypernetwork_layer_structure": OptionInfo(None, "Hypernetwork layer structure Default: (1,2,1).", gr.Dropdown, lambda: {"choices": [(1, 2, 1), (1, 2, 2, 1), (1, 2, 4, 2, 1)]}), - "sd_hypernetwork_add_layer_norm": OptionInfo(False, "Add layer normalization to hypernetwork architecture."), "sd_hypernetwork_strength": OptionInfo(1.0, "Hypernetwork strength", gr.Slider, {"minimum": 0.0, "maximum": 1.0, "step": 0.001}), "img2img_color_correction": OptionInfo(False, "Apply color correction to img2img results to match original colors."), "save_images_before_color_correction": OptionInfo(False, "Save a copy of image before applying color correction to img2img results"), diff --git a/modules/ui.py b/modules/ui.py index ca46343f..d9ee462f 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -458,14 +458,14 @@ def create_toprow(is_img2img): with gr.Row(): with gr.Column(scale=80): with gr.Row(): - prompt = gr.Textbox(label="Prompt", elem_id=f"{id_part}_prompt", show_label=False, lines=2, + prompt = gr.Textbox(label="Prompt", elem_id=f"{id_part}_prompt", show_label=False, lines=2, placeholder="Prompt (press Ctrl+Enter or Alt+Enter to generate)" ) with gr.Row(): with gr.Column(scale=80): with gr.Row(): - negative_prompt = gr.Textbox(label="Negative prompt", elem_id=f"{id_part}_neg_prompt", show_label=False, lines=2, + negative_prompt = gr.Textbox(label="Negative prompt", elem_id=f"{id_part}_neg_prompt", show_label=False, lines=2, placeholder="Negative prompt (press Ctrl+Enter or Alt+Enter to generate)" ) @@ -1198,6 +1198,8 @@ def create_ui(wrap_gradio_gpu_call): with gr.Tab(label="Create hypernetwork"): new_hypernetwork_name = gr.Textbox(label="Name") new_hypernetwork_sizes = gr.CheckboxGroup(label="Modules", value=["768", "320", "640", "1280"], choices=["768", "320", "640", "1280"]) + new_hypernetwork_layer_structure = gr.Dropdown(label="Hypernetwork layer structure", choices=[(1, 2, 1), (1, 2, 2, 1), (1, 2, 4, 2, 1)]) + new_hypernetwork_add_layer_norm = gr.Checkbox(label="Add layer normalization") with gr.Row(): with gr.Column(scale=3): @@ -1280,6 +1282,8 @@ def create_ui(wrap_gradio_gpu_call): inputs=[ new_hypernetwork_name, new_hypernetwork_sizes, + new_hypernetwork_layer_structure, + new_hypernetwork_add_layer_norm, ], outputs=[ train_hypernetwork_name, -- cgit v1.2.3 From 3770b8d2fa62066d472a04739c7b84bce8538832 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 15:28:42 +0000 Subject: enable to write layer structure of hn himself --- modules/hypernetworks/ui.py | 4 ++++ modules/ui.py | 2 +- 2 files changed, 5 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 7e8ea95e..08f75f15 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -1,5 +1,6 @@ import html import os +import re import gradio as gr @@ -13,6 +14,9 @@ def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") assert not os.path.exists(fn), f"file {fn} already exists" + if type(layer_structure) == str: + layer_structure = tuple(map(int, re.sub(r'\D', '', layer_structure))) + hypernet = modules.hypernetworks.hypernetwork.Hypernetwork( name=name, enable_sizes=[int(x) for x in enable_sizes], diff --git a/modules/ui.py b/modules/ui.py index d9ee462f..18a2add0 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1198,7 +1198,7 @@ def create_ui(wrap_gradio_gpu_call): with gr.Tab(label="Create hypernetwork"): new_hypernetwork_name = gr.Textbox(label="Name") new_hypernetwork_sizes = gr.CheckboxGroup(label="Modules", value=["768", "320", "640", "1280"], choices=["768", "320", "640", "1280"]) - new_hypernetwork_layer_structure = gr.Dropdown(label="Hypernetwork layer structure", choices=[(1, 2, 1), (1, 2, 2, 1), (1, 2, 4, 2, 1)]) + new_hypernetwork_layer_structure = gr.Textbox("1, 2, 1", label="Enter hypernetwork layer structure", placeholder="1st and last digit must be 1. ex:'1, 2, 1'") new_hypernetwork_add_layer_norm = gr.Checkbox(label="Add layer normalization") with gr.Row(): -- cgit v1.2.3 From 019a3a88f07766f2d32c32fbe8e41625f28ecb5e Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Wed, 19 Oct 2022 17:15:47 +0100 Subject: Update ui.py --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index d2e24880..1573ef82 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1247,7 +1247,7 @@ def create_ui(wrap_gradio_gpu_call): run_preprocess = gr.Button(value="Preprocess", variant='primary') with gr.Tab(label="Train"): - gr.HTML(value="

Train an embedding; must specify a directory with a set of 1:1 ratio images

") + gr.HTML(value="

Train an embedding or Hypernetwork; you must specify a directory with a set of 1:1 ratio images
Initial learning rates: 0.005 for an Embedding, 0.00001 for Hypernetwork wiki

") with gr.Row(): train_embedding_name = gr.Dropdown(label='Embedding', elem_id="train_embedding", choices=sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys())) create_refresh_button(train_embedding_name, sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings, lambda: {"choices": sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys())}, "refresh_train_embedding_name") -- cgit v1.2.3 From c6e9fed5003631c87d548e74d6e359678959a453 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Wed, 19 Oct 2022 19:21:16 +0300 Subject: fix for #3086 failing to load any previous hypernet --- modules/hypernetworks/hypernetwork.py | 60 ++++++++++++++++------------------- 1 file changed, 28 insertions(+), 32 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 7d519cd9..74300122 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -24,11 +24,10 @@ class HypernetworkModule(torch.nn.Module): def __init__(self, dim, state_dict=None, layer_structure=None, add_layer_norm=False): super().__init__() - if layer_structure is not None: - assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" - assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" - else: - layer_structure = parse_layer_structure(dim, state_dict) + + assert layer_structure is not None, "layer_structure mut not be None" + assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" + assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" linears = [] for i in range(len(layer_structure) - 1): @@ -39,23 +38,30 @@ class HypernetworkModule(torch.nn.Module): self.linear = torch.nn.Sequential(*linears) if state_dict is not None: - try: - self.load_state_dict(state_dict) - except RuntimeError: - self.try_load_previous(state_dict) + self.fix_old_state_dict(state_dict) + self.load_state_dict(state_dict) else: for layer in self.linear: - layer.weight.data.normal_(mean = 0.0, std = 0.01) + layer.weight.data.normal_(mean=0.0, std=0.01) layer.bias.data.zero_() self.to(devices.device) - def try_load_previous(self, state_dict): - states = self.state_dict() - states['linear.0.bias'].copy_(state_dict['linear1.bias']) - states['linear.0.weight'].copy_(state_dict['linear1.weight']) - states['linear.1.bias'].copy_(state_dict['linear2.bias']) - states['linear.1.weight'].copy_(state_dict['linear2.weight']) + def fix_old_state_dict(self, state_dict): + changes = { + 'linear1.bias': 'linear.0.bias', + 'linear1.weight': 'linear.0.weight', + 'linear2.bias': 'linear.1.bias', + 'linear2.weight': 'linear.1.weight', + } + + for fr, to in changes.items(): + x = state_dict.get(fr, None) + if x is None: + continue + + del state_dict[fr] + state_dict[to] = x def forward(self, x): return x + self.linear(x) * self.multiplier @@ -71,18 +77,6 @@ def apply_strength(value=None): HypernetworkModule.multiplier = value if value is not None else shared.opts.sd_hypernetwork_strength -def parse_layer_structure(dim, state_dict): - i = 0 - layer_structure = [1] - - while (key := "linear.{}.weight".format(i)) in state_dict: - weight = state_dict[key] - layer_structure.append(len(weight) // dim) - i += 1 - - return layer_structure - - class Hypernetwork: filename = None name = None @@ -135,17 +129,18 @@ class Hypernetwork: state_dict = torch.load(filename, map_location='cpu') + self.layer_structure = state_dict.get('layer_structure', [1, 2, 1]) + self.add_layer_norm = state_dict.get('is_layer_norm', False) + for size, sd in state_dict.items(): if type(size) == int: self.layers[size] = ( - HypernetworkModule(size, sd[0], state_dict["layer_structure"], state_dict["is_layer_norm"]), - HypernetworkModule(size, sd[1], state_dict["layer_structure"], state_dict["is_layer_norm"]), + HypernetworkModule(size, sd[0], self.layer_structure, self.add_layer_norm), + HypernetworkModule(size, sd[1], self.layer_structure, self.add_layer_norm), ) self.name = state_dict.get('name', self.name) self.step = state_dict.get('step', 0) - self.layer_structure = state_dict.get('layer_structure', None) - self.add_layer_norm = state_dict.get('is_layer_norm', False) self.sd_checkpoint = state_dict.get('sd_checkpoint', None) self.sd_checkpoint_name = state_dict.get('sd_checkpoint_name', None) @@ -244,6 +239,7 @@ def stack_conds(conds): return torch.stack(conds) + def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height): assert hypernetwork_name, 'hypernetwork not selected' -- cgit v1.2.3 From 2ce52d32e41fb523d1494f45073fd18496e52d35 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Wed, 19 Oct 2022 16:31:12 +0000 Subject: fix for #3086 failing to load any previous hypernet --- modules/hypernetworks/hypernetwork.py | 60 ++++++++++++++++------------------- 1 file changed, 28 insertions(+), 32 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 7d519cd9..74300122 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -24,11 +24,10 @@ class HypernetworkModule(torch.nn.Module): def __init__(self, dim, state_dict=None, layer_structure=None, add_layer_norm=False): super().__init__() - if layer_structure is not None: - assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" - assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" - else: - layer_structure = parse_layer_structure(dim, state_dict) + + assert layer_structure is not None, "layer_structure mut not be None" + assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" + assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" linears = [] for i in range(len(layer_structure) - 1): @@ -39,23 +38,30 @@ class HypernetworkModule(torch.nn.Module): self.linear = torch.nn.Sequential(*linears) if state_dict is not None: - try: - self.load_state_dict(state_dict) - except RuntimeError: - self.try_load_previous(state_dict) + self.fix_old_state_dict(state_dict) + self.load_state_dict(state_dict) else: for layer in self.linear: - layer.weight.data.normal_(mean = 0.0, std = 0.01) + layer.weight.data.normal_(mean=0.0, std=0.01) layer.bias.data.zero_() self.to(devices.device) - def try_load_previous(self, state_dict): - states = self.state_dict() - states['linear.0.bias'].copy_(state_dict['linear1.bias']) - states['linear.0.weight'].copy_(state_dict['linear1.weight']) - states['linear.1.bias'].copy_(state_dict['linear2.bias']) - states['linear.1.weight'].copy_(state_dict['linear2.weight']) + def fix_old_state_dict(self, state_dict): + changes = { + 'linear1.bias': 'linear.0.bias', + 'linear1.weight': 'linear.0.weight', + 'linear2.bias': 'linear.1.bias', + 'linear2.weight': 'linear.1.weight', + } + + for fr, to in changes.items(): + x = state_dict.get(fr, None) + if x is None: + continue + + del state_dict[fr] + state_dict[to] = x def forward(self, x): return x + self.linear(x) * self.multiplier @@ -71,18 +77,6 @@ def apply_strength(value=None): HypernetworkModule.multiplier = value if value is not None else shared.opts.sd_hypernetwork_strength -def parse_layer_structure(dim, state_dict): - i = 0 - layer_structure = [1] - - while (key := "linear.{}.weight".format(i)) in state_dict: - weight = state_dict[key] - layer_structure.append(len(weight) // dim) - i += 1 - - return layer_structure - - class Hypernetwork: filename = None name = None @@ -135,17 +129,18 @@ class Hypernetwork: state_dict = torch.load(filename, map_location='cpu') + self.layer_structure = state_dict.get('layer_structure', [1, 2, 1]) + self.add_layer_norm = state_dict.get('is_layer_norm', False) + for size, sd in state_dict.items(): if type(size) == int: self.layers[size] = ( - HypernetworkModule(size, sd[0], state_dict["layer_structure"], state_dict["is_layer_norm"]), - HypernetworkModule(size, sd[1], state_dict["layer_structure"], state_dict["is_layer_norm"]), + HypernetworkModule(size, sd[0], self.layer_structure, self.add_layer_norm), + HypernetworkModule(size, sd[1], self.layer_structure, self.add_layer_norm), ) self.name = state_dict.get('name', self.name) self.step = state_dict.get('step', 0) - self.layer_structure = state_dict.get('layer_structure', None) - self.add_layer_norm = state_dict.get('is_layer_norm', False) self.sd_checkpoint = state_dict.get('sd_checkpoint', None) self.sd_checkpoint_name = state_dict.get('sd_checkpoint_name', None) @@ -244,6 +239,7 @@ def stack_conds(conds): return torch.stack(conds) + def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height): assert hypernetwork_name, 'hypernetwork not selected' -- cgit v1.2.3 From 57eb1a64c85d995cacb4fa3832e87405bf6820b9 Mon Sep 17 00:00:00 2001 From: Alexandre Simard Date: Wed, 19 Oct 2022 12:28:27 -0400 Subject: Update ui.py --- modules/ui.py | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index d2e24880..c9a923ab 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -268,8 +268,13 @@ def calc_time_left(progress, threshold, label, force_display): time_since_start = time.time() - shared.state.time_start eta = (time_since_start/progress) eta_relative = eta-time_since_start - if (eta_relative > threshold and progress > 0.02) or force_display: - return label + time.strftime('%H:%M:%S', time.gmtime(eta_relative)) + if (eta_relative > threshold and progress > 0.02) or force_display: + if eta_relative > 3600: + return label + time.strftime('%H:%M:%S', time.gmtime(eta_relative)) + elif eta_relative > 60: + return label + time.strftime('%M:%S', time.gmtime(eta_relative)) + else: + return label + time.strftime('%Ss', time.gmtime(eta_relative)) else: return "" @@ -285,7 +290,7 @@ def check_progress_call(id_part): if shared.state.sampling_steps > 0: progress += 1 / shared.state.job_count * shared.state.sampling_step / shared.state.sampling_steps - time_left = calc_time_left( progress, 60, " ETA:", shared.state.time_left_force_display ) + time_left = calc_time_left( progress, 1, " ETA: ", shared.state.time_left_force_display ) if time_left != "": shared.state.time_left_force_display = True @@ -293,7 +298,7 @@ def check_progress_call(id_part): progressbar = "" if opts.show_progressbar: - progressbar = f"""
{str(int(progress*100))+"%"+time_left if progress > 0.01 else ""}
""" + progressbar = f"""
{str(int(progress*100))+"%"+time_left if progress > 0.01 else ""}
""" image = gr_show(False) preview_visibility = gr_show(False) -- cgit v1.2.3 From 1e4809b251d478a102fd980dcfc26e21d6d3730b Mon Sep 17 00:00:00 2001 From: Alexandre Simard Date: Wed, 19 Oct 2022 12:53:23 -0400 Subject: Added a bit of padding to the left --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index c9a923ab..a2dbd41e 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -298,7 +298,7 @@ def check_progress_call(id_part): progressbar = "" if opts.show_progressbar: - progressbar = f"""
{str(int(progress*100))+"%"+time_left if progress > 0.01 else ""}
""" + progressbar = f"""
{" " * 2 + str(int(progress*100))+"%" + time_left if progress > 0.01 else ""}
""" image = gr_show(False) preview_visibility = gr_show(False) -- cgit v1.2.3 From 5e012e4dfa5dcfeade0394678cf14b70682dba6c Mon Sep 17 00:00:00 2001 From: timntorres Date: Wed, 19 Oct 2022 06:17:47 -0700 Subject: Infotext saves more specific hypernet name. --- modules/processing.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index ea926fc3..bcb0c32c 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -304,7 +304,7 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration "Size": f"{p.width}x{p.height}", "Model hash": getattr(p, 'sd_model_hash', None if not opts.add_model_hash_to_info or not shared.sd_model.sd_model_hash else shared.sd_model.sd_model_hash), "Model": (None if not opts.add_model_name_to_info or not shared.sd_model.sd_checkpoint_info.model_name else shared.sd_model.sd_checkpoint_info.model_name.replace(',', '').replace(':', '')), - "Hypernet": (None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.name.replace(',', '').replace(':', '')), + "Hypernet": (None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.filename.split('\\')[-1].split('.')[0]), "Batch size": (None if p.batch_size < 2 else p.batch_size), "Batch pos": (None if p.batch_size < 2 else position_in_batch), "Variation seed": (None if p.subseed_strength == 0 else all_subseeds[index]), -- cgit v1.2.3 From 46122c4ff6aadc0f96e657f88dbac7bbd9f9bf99 Mon Sep 17 00:00:00 2001 From: Vladimir Repin <32306715+mezotaken@users.noreply.github.com> Date: Wed, 19 Oct 2022 19:18:52 +0300 Subject: Send empty prompts as valid generation parameter --- modules/generation_parameters_copypaste.py | 3 --- 1 file changed, 3 deletions(-) (limited to 'modules') diff --git a/modules/generation_parameters_copypaste.py b/modules/generation_parameters_copypaste.py index c27826b6..98d24406 100644 --- a/modules/generation_parameters_copypaste.py +++ b/modules/generation_parameters_copypaste.py @@ -45,10 +45,7 @@ Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 965400086, Size: 512x512, Model else: prompt += ("" if prompt == "" else "\n") + line - if len(prompt) > 0: res["Prompt"] = prompt - - if len(negative_prompt) > 0: res["Negative prompt"] = negative_prompt for k, v in re_param.findall(lastline): -- cgit v1.2.3 From b748b583c0b9f771c1be509175a6913e3f2ad97c Mon Sep 17 00:00:00 2001 From: Mackerel Date: Wed, 19 Oct 2022 14:22:03 -0400 Subject: generation_parameters_copypaste.py: fix indent --- modules/generation_parameters_copypaste.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/generation_parameters_copypaste.py b/modules/generation_parameters_copypaste.py index 98d24406..0f041449 100644 --- a/modules/generation_parameters_copypaste.py +++ b/modules/generation_parameters_copypaste.py @@ -45,8 +45,8 @@ Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 965400086, Size: 512x512, Model else: prompt += ("" if prompt == "" else "\n") + line - res["Prompt"] = prompt - res["Negative prompt"] = negative_prompt + res["Prompt"] = prompt + res["Negative prompt"] = negative_prompt for k, v in re_param.findall(lastline): m = re_imagesize.match(v) -- cgit v1.2.3 From eb7ba4b713ac2fb960ecf6365b1de0c89451e583 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Wed, 19 Oct 2022 19:50:46 +0100 Subject: update training header text --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 1573ef82..93c0767c 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1247,7 +1247,7 @@ def create_ui(wrap_gradio_gpu_call): run_preprocess = gr.Button(value="Preprocess", variant='primary') with gr.Tab(label="Train"): - gr.HTML(value="

Train an embedding or Hypernetwork; you must specify a directory with a set of 1:1 ratio images
Initial learning rates: 0.005 for an Embedding, 0.00001 for Hypernetwork wiki

") + gr.HTML(value="

Train an embedding or Hypernetwork; you must specify a directory with a set of 1:1 ratio images
Initial learning rates: 0.005 for an Embedding, 0.00001 for Hypernetwork [wiki]

") with gr.Row(): train_embedding_name = gr.Dropdown(label='Embedding', elem_id="train_embedding", choices=sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys())) create_refresh_button(train_embedding_name, sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings, lambda: {"choices": sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys())}, "refresh_train_embedding_name") -- cgit v1.2.3 From 4d663055ded968831ec97f047dfa8e94036cf1c1 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Wed, 19 Oct 2022 20:33:18 +0100 Subject: update ui with extra training options --- modules/ui.py | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 93c0767c..cdb9d335 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1206,6 +1206,7 @@ def create_ui(wrap_gradio_gpu_call): new_embedding_name = gr.Textbox(label="Name") initialization_text = gr.Textbox(label="Initialization text", value="*") nvpt = gr.Slider(label="Number of vectors per token", minimum=1, maximum=75, step=1, value=1) + overwrite_old_embedding = gr.Checkbox(value=False, label="Overwrite Old Embedding") with gr.Row(): with gr.Column(scale=3): @@ -1219,6 +1220,7 @@ def create_ui(wrap_gradio_gpu_call): new_hypernetwork_sizes = gr.CheckboxGroup(label="Modules", value=["768", "320", "640", "1280"], choices=["768", "320", "640", "1280"]) new_hypernetwork_layer_structure = gr.Textbox("1, 2, 1", label="Enter hypernetwork layer structure", placeholder="1st and last digit must be 1. ex:'1, 2, 1'") new_hypernetwork_add_layer_norm = gr.Checkbox(label="Add layer normalization") + overwrite_old_hypernetwork = gr.Checkbox(value=False, label="Overwrite Old Hypernetwork") with gr.Row(): with gr.Column(scale=3): @@ -1247,14 +1249,17 @@ def create_ui(wrap_gradio_gpu_call): run_preprocess = gr.Button(value="Preprocess", variant='primary') with gr.Tab(label="Train"): - gr.HTML(value="

Train an embedding or Hypernetwork; you must specify a directory with a set of 1:1 ratio images
Initial learning rates: 0.005 for an Embedding, 0.00001 for Hypernetwork [wiki]

") + gr.HTML(value="

Train an embedding or Hypernetwork; you must specify a directory with a set of 1:1 ratio images [wiki]

") with gr.Row(): train_embedding_name = gr.Dropdown(label='Embedding', elem_id="train_embedding", choices=sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys())) create_refresh_button(train_embedding_name, sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings, lambda: {"choices": sorted(sd_hijack.model_hijack.embedding_db.word_embeddings.keys())}, "refresh_train_embedding_name") with gr.Row(): train_hypernetwork_name = gr.Dropdown(label='Hypernetwork', elem_id="train_hypernetwork", choices=[x for x in shared.hypernetworks.keys()]) create_refresh_button(train_hypernetwork_name, shared.reload_hypernetworks, lambda: {"choices": sorted([x for x in shared.hypernetworks.keys()])}, "refresh_train_hypernetwork_name") - learn_rate = gr.Textbox(label='Learning rate', placeholder="Learning rate", value="0.005") + with gr.Row(): + embedding_learn_rate = gr.Textbox(label='Embedding Learning rate', placeholder="Embedding Learning rate", value="0.005") + hypernetwork_learn_rate = gr.Textbox(label='Hypernetwork Learning rate', placeholder="Hypernetwork Learning rate", value="0.00001") + batch_size = gr.Number(label='Batch size', value=1, precision=0) dataset_directory = gr.Textbox(label='Dataset directory', placeholder="Path to directory with input images") log_directory = gr.Textbox(label='Log directory', placeholder="Path to directory where to write outputs", value="textual_inversion") @@ -1288,6 +1293,7 @@ def create_ui(wrap_gradio_gpu_call): new_embedding_name, initialization_text, nvpt, + overwrite_old_embedding, ], outputs=[ train_embedding_name, @@ -1303,6 +1309,7 @@ def create_ui(wrap_gradio_gpu_call): new_hypernetwork_sizes, new_hypernetwork_layer_structure, new_hypernetwork_add_layer_norm, + overwrite_old_hypernetwork, ], outputs=[ train_hypernetwork_name, -- cgit v1.2.3 From 8e7097d06a6a261580d34375c9d2a9e4ffc63ffa Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Wed, 19 Oct 2022 13:47:45 -0700 Subject: Added support for RunwayML inpainting model --- modules/processing.py | 34 ++++++- modules/sd_hijack_inpainting.py | 208 ++++++++++++++++++++++++++++++++++++++++ modules/sd_models.py | 16 +++- modules/sd_samplers.py | 50 +++++++--- 4 files changed, 293 insertions(+), 15 deletions(-) create mode 100644 modules/sd_hijack_inpainting.py (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index bcb0c32c..a6c308f9 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -546,7 +546,16 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): if not self.enable_hr: x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) - samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning) + + # The "masked-image" in this case will just be all zeros since the entire image is masked. + image_conditioning = torch.zeros(x.shape[0], 3, self.height, self.width, device=x.device) + image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning)) + + # Add the fake full 1s mask to the first dimension. + image_conditioning = torch.nn.functional.pad(image_conditioning, (0, 0, 0, 0, 1, 0), value=1.0) + image_conditioning = image_conditioning.to(x.dtype) + + samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=image_conditioning) return samples x = create_random_tensors([opt_C, self.firstphase_height // opt_f, self.firstphase_width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) @@ -714,10 +723,31 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): elif self.inpainting_fill == 3: self.init_latent = self.init_latent * self.mask + if self.image_mask is not None: + conditioning_mask = np.array(self.image_mask.convert("L")) + conditioning_mask = conditioning_mask.astype(np.float32) / 255.0 + conditioning_mask = torch.from_numpy(conditioning_mask[None, None]) + + # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0 + conditioning_mask = torch.round(conditioning_mask) + else: + conditioning_mask = torch.ones(1, 1, *image.shape[-2:]) + + # Create another latent image, this time with a masked version of the original input. + conditioning_mask = conditioning_mask.to(image.device) + conditioning_image = image * (1.0 - conditioning_mask) + conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image)) + + # Create the concatenated conditioning tensor to be fed to `c_concat` + conditioning_mask = torch.nn.functional.interpolate(conditioning_mask, size=self.init_latent.shape[-2:]) + conditioning_mask = conditioning_mask.expand(conditioning_image.shape[0], -1, -1, -1) + self.image_conditioning = torch.cat([conditioning_mask, conditioning_image], dim=1) + self.image_conditioning = self.image_conditioning.to(shared.device).type(self.sd_model.dtype) + def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) - samples = self.sampler.sample_img2img(self, self.init_latent, x, conditioning, unconditional_conditioning) + samples = self.sampler.sample_img2img(self, self.init_latent, x, conditioning, unconditional_conditioning, image_conditioning=self.image_conditioning) if self.mask is not None: samples = samples * self.nmask + self.init_latent * self.mask diff --git a/modules/sd_hijack_inpainting.py b/modules/sd_hijack_inpainting.py new file mode 100644 index 00000000..7e5670d6 --- /dev/null +++ b/modules/sd_hijack_inpainting.py @@ -0,0 +1,208 @@ +import torch +import numpy as np + +from tqdm import tqdm +from einops import rearrange, repeat +from omegaconf import ListConfig + +from types import MethodType + +import ldm.models.diffusion.ddpm +import ldm.models.diffusion.ddim + +from ldm.models.diffusion.ddpm import LatentDiffusion +from ldm.models.diffusion.ddim import DDIMSampler, noise_like + +# ================================================================================================= +# Monkey patch DDIMSampler methods from RunwayML repo directly. +# Adapted from: +# https://github.com/runwayml/stable-diffusion/blob/main/ldm/models/diffusion/ddim.py +# ================================================================================================= +@torch.no_grad() +def sample( + self, + S, + batch_size, + shape, + conditioning=None, + callback=None, + normals_sequence=None, + img_callback=None, + quantize_x0=False, + eta=0., + mask=None, + x0=None, + temperature=1., + noise_dropout=0., + score_corrector=None, + corrector_kwargs=None, + verbose=True, + x_T=None, + log_every_t=100, + unconditional_guidance_scale=1., + unconditional_conditioning=None, + # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ... + **kwargs + ): + if conditioning is not None: + if isinstance(conditioning, dict): + ctmp = conditioning[list(conditioning.keys())[0]] + while isinstance(ctmp, list): + ctmp = elf.inpainting_fill == 2: + self.init_latent = self.init_latent * self.mask + create_random_tensors(self.init_latent.shape[1:], all_seeds[0:self.init_latent.shape[0]]) * self.nmask + elif self.inpainting_fill == 3: + self.init_latent = self.init_latent * self.mask + + if self.image_mask is not None: + conditioning_mask = np.array(self.image_mask.convert("L")) + conditioning_mask = conditioning_mask.astype(np.float32) / 255.0 + conditioning_mask = torch.from_numpy(conditioning_mask[None, None]) + + # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0 + conditioning_mask = torch.round(conditioning_mask) + else: + conditioning_mask = torch.ones(1, 1, *image.shape[-2:]) + + # Create another latent image, this time with a masked version of the original input. + conditioning_mask = conditioning_mask.to(image.device) + conditioning_image = image * (1.0 - conditioning_mask) + conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image)) + + # Create the concatenated conditioning tensor to be fed to `c_concat` + conditioning_mask = torch.nn.functional.interpolate(conditioning_mask, size=self.init_latent.shape[-2:]) + conditioning_mask = conditioning_mask.expand(conditioning_image.shape[0], -1, -1, -1) + self.image_conditioning = torch.cat([conditioning_mask, conditioning_image], dim=1) + self.image_conditioning = self.image_conditioning.to(shared.device).type(self.sd_model.dtype) + + def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): + x = create_random_tensors([opctmp[0] + cbs = ctmp.shape[0] + if cbs != batch_size: + print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}") + else: + if conditioning.shape[0] != batch_size: + print(f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}") + + self.make_schedule(ddim_num_steps=S, ddim_eta=eta, verbose=verbose) + # sampling + C, H, W = shape + size = (batch_size, C, H, W) + print(f'Data shape for DDIM sampling is {size}, eta {eta}') + + samples, intermediates = self.ddim_sampling(conditioning, size, + callback=callback, + img_callback=img_callback, + quantize_denoised=quantize_x0, + mask=mask, x0=x0, + ddim_use_original_steps=False, + noise_dropout=noise_dropout, + temperature=temperature, + score_corrector=score_corrector, + corrector_kwargs=corrector_kwargs, + x_T=x_T, + log_every_t=log_every_t, + unconditional_guidance_scale=unconditional_guidance_scale, + unconditional_conditioning=unconditional_conditioning, + ) + return samples, intermediates + + +@torch.no_grad() +def p_sample_ddim(self, x, c, t, index, repeat_noise=False, use_original_steps=False, quantize_denoised=False, + temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None, + unconditional_guidance_scale=1., unconditional_conditioning=None): + b, *_, device = *x.shape, x.device + + if unconditional_conditioning is None or unconditional_guidance_scale == 1.: + e_t = self.model.apply_model(x, t, c) + else: + x_in = torch.cat([x] * 2) + t_in = torch.cat([t] * 2) + if isinstance(c, dict): + assert isinstance(unconditional_conditioning, dict) + c_in = dict() + for k in c: + if isinstance(c[k], list): + c_in[k] = [ + torch.cat([unconditional_conditioning[k][i], c[k][i]]) + for i in range(len(c[k])) + ] + else: + c_in[k] = torch.cat([unconditional_conditioning[k], c[k]]) + else: + c_in = torch.cat([unconditional_conditioning, c]) + e_t_uncond, e_t = self.model.apply_model(x_in, t_in, c_in).chunk(2) + e_t = e_t_uncond + unconditional_guidance_scale * (e_t - e_t_uncond) + + if score_corrector is not None: + assert self.model.parameterization == "eps" + e_t = score_corrector.modify_score(self.model, e_t, x, t, c, **corrector_kwargs) + + alphas = self.model.alphas_cumprod if use_original_steps else self.ddim_alphas + alphas_prev = self.model.alphas_cumprod_prev if use_original_steps else self.ddim_alphas_prev + sqrt_one_minus_alphas = self.model.sqrt_one_minus_alphas_cumprod if use_original_steps else self.ddim_sqrt_one_minus_alphas + sigmas = self.model.ddim_sigmas_for_original_num_steps if use_original_steps else self.ddim_sigmas + # select parameters corresponding to the currently considered timestep + a_t = torch.full((b, 1, 1, 1), alphas[index], device=device) + a_prev = torch.full((b, 1, 1, 1), alphas_prev[index], device=device) + sigma_t = torch.full((b, 1, 1, 1), sigmas[index], device=device) + sqrt_one_minus_at = torch.full((b, 1, 1, 1), sqrt_one_minus_alphas[index],device=device) + + # current prediction for x_0 + pred_x0 = (x - sqrt_one_minus_at * e_t) / a_t.sqrt() + if quantize_denoised: + pred_x0, _, *_ = self.model.first_stage_model.quantize(pred_x0) + # direction pointing to x_t + dir_xt = (1. - a_prev - sigma_t**2).sqrt() * e_t + noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature + if noise_dropout > 0.: + noise = torch.nn.functional.dropout(noise, p=noise_dropout) + x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise + return x_prev, pred_x0 + + +# ================================================================================================= +# Monkey patch LatentInpaintDiffusion to load the checkpoint with a proper config. +# Adapted from: +# https://github.com/runwayml/stable-diffusion/blob/main/ldm/models/diffusion/ddpm.py +# ================================================================================================= + +@torch.no_grad() +def get_unconditional_conditioning(self, batch_size, null_label=None): + if null_label is not None: + xc = null_label + if isinstance(xc, ListConfig): + xc = list(xc) + if isinstance(xc, dict) or isinstance(xc, list): + c = self.get_learned_conditioning(xc) + else: + if hasattr(xc, "to"): + xc = xc.to(self.device) + c = self.get_learned_conditioning(xc) + else: + # todo: get null label from cond_stage_model + raise NotImplementedError() + c = repeat(c, "1 ... -> b ...", b=batch_size).to(self.device) + return c + +class LatentInpaintDiffusion(LatentDiffusion): + def __init__( + self, + concat_keys=("mask", "masked_image"), + masked_image_key="masked_image", + *args, + **kwargs, + ): + super().__init__(*args, **kwargs) + self.masked_image_key = masked_image_key + assert self.masked_image_key in concat_keys + self.concat_keys = concat_keys + +def should_hijack_inpainting(checkpoint_info): + return str(checkpoint_info.filename).endswith("inpainting.ckpt") and not checkpoint_info.config.endswith("inpainting.yaml") + +def do_inpainting_hijack(): + ldm.models.diffusion.ddpm.get_unconditional_conditioning = get_unconditional_conditioning + ldm.models.diffusion.ddpm.LatentInpaintDiffusion = LatentInpaintDiffusion + ldm.models.diffusion.ddim.DDIMSampler.p_sample_ddim = p_sample_ddim + ldm.models.diffusion.ddim.DDIMSampler.sample = sample \ No newline at end of file diff --git a/modules/sd_models.py b/modules/sd_models.py index eae22e87..47836d25 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -9,6 +9,7 @@ from ldm.util import instantiate_from_config from modules import shared, modelloader, devices from modules.paths import models_path +from modules.sd_hijack_inpainting import do_inpainting_hijack, should_hijack_inpainting model_dir = "Stable-diffusion" model_path = os.path.abspath(os.path.join(models_path, model_dir)) @@ -211,6 +212,19 @@ def load_model(): print(f"Loading config from: {checkpoint_info.config}") sd_config = OmegaConf.load(checkpoint_info.config) + + if should_hijack_inpainting(checkpoint_info): + do_inpainting_hijack() + + # Hardcoded config for now... + sd_config.model.target = "ldm.models.diffusion.ddpm.LatentInpaintDiffusion" + sd_config.model.params.use_ema = False + sd_config.model.params.conditioning_key = "hybrid" + sd_config.model.params.unet_config.params.in_channels = 9 + + # Create a "fake" config with a different name so that we know to unload it when switching models. + checkpoint_info = checkpoint_info._replace(config=checkpoint_info.config.replace(".yaml", "-inpainting.yaml")) + sd_model = instantiate_from_config(sd_config.model) load_model_weights(sd_model, checkpoint_info) @@ -234,7 +248,7 @@ def reload_model_weights(sd_model, info=None): if sd_model.sd_model_checkpoint == checkpoint_info.filename: return - if sd_model.sd_checkpoint_info.config != checkpoint_info.config: + if sd_model.sd_checkpoint_info.config != checkpoint_info.config or should_hijack_inpainting(checkpoint_info) != should_hijack_inpainting(sd_model.sd_checkpoint_info): checkpoints_loaded.clear() shared.sd_model = load_model() return shared.sd_model diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index b58e810b..9d3cf289 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -136,9 +136,15 @@ class VanillaStableDiffusionSampler: if self.stop_at is not None and self.step > self.stop_at: raise InterruptedException + # Have to unwrap the inpainting conditioning here to perform pre-preocessing + image_conditioning = None + if isinstance(cond, dict): + image_conditioning = cond["c_concat"][0] + cond = cond["c_crossattn"][0] + unconditional_conditioning = unconditional_conditioning["c_crossattn"][0] conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step) - unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step) + unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step) assert all([len(conds) == 1 for conds in conds_list]), 'composition via AND is not supported for DDIM/PLMS samplers' cond = tensor @@ -157,6 +163,10 @@ class VanillaStableDiffusionSampler: img_orig = self.sampler.model.q_sample(self.init_latent, ts) x_dec = img_orig * self.mask + self.nmask * x_dec + if image_conditioning is not None: + cond = {"c_concat": [image_conditioning], "c_crossattn": [cond]} + unconditional_conditioning = {"c_concat": [image_conditioning], "c_crossattn": [unconditional_conditioning]} + res = self.orig_p_sample_ddim(x_dec, cond, ts, unconditional_conditioning=unconditional_conditioning, *args, **kwargs) if self.mask is not None: @@ -182,7 +192,7 @@ class VanillaStableDiffusionSampler: self.mask = p.mask if hasattr(p, 'mask') else None self.nmask = p.nmask if hasattr(p, 'nmask') else None - def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None): + def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None, image_conditioning=None): steps, t_enc = setup_img2img_steps(p, steps) self.initialize(p) @@ -202,7 +212,7 @@ class VanillaStableDiffusionSampler: return samples - def sample(self, p, x, conditioning, unconditional_conditioning, steps=None): + def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, image_conditioning=None): self.initialize(p) self.init_latent = None @@ -210,6 +220,11 @@ class VanillaStableDiffusionSampler: steps = steps or p.steps + # Wrap the conditioning models with additional image conditioning for inpainting model + if image_conditioning is not None: + conditioning = {"c_concat": [image_conditioning], "c_crossattn": [conditioning]} + unconditional_conditioning = {"c_concat": [image_conditioning], "c_crossattn": [unconditional_conditioning]} + # existing code fails with certain step counts, like 9 try: samples_ddim = self.launch_sampling(steps, lambda: self.sampler.sample(S=steps, conditioning=conditioning, batch_size=int(x.shape[0]), shape=x[0].shape, verbose=False, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning, x_T=x, eta=self.eta)[0]) @@ -228,7 +243,7 @@ class CFGDenoiser(torch.nn.Module): self.init_latent = None self.step = 0 - def forward(self, x, sigma, uncond, cond, cond_scale): + def forward(self, x, sigma, uncond, cond, cond_scale, image_cond): if state.interrupted or state.skipped: raise InterruptedException @@ -239,28 +254,29 @@ class CFGDenoiser(torch.nn.Module): repeats = [len(conds_list[i]) for i in range(batch_size)] x_in = torch.cat([torch.stack([x[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [x]) + image_cond_in = torch.cat([torch.stack([image_cond[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [image_cond]) sigma_in = torch.cat([torch.stack([sigma[i] for _ in range(n)]) for i, n in enumerate(repeats)] + [sigma]) if tensor.shape[1] == uncond.shape[1]: cond_in = torch.cat([tensor, uncond]) if shared.batch_cond_uncond: - x_out = self.inner_model(x_in, sigma_in, cond=cond_in) + x_out = self.inner_model(x_in, sigma_in, cond={"c_crossattn": [cond_in], "c_concat": [image_cond_in]}) else: x_out = torch.zeros_like(x_in) 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=cond_in[a:b]) + x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond={"c_crossattn": [cond_in[a:b]], "c_concat": [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 for batch_offset in range(0, tensor.shape[0], batch_size): a = batch_offset b = min(a + batch_size, tensor.shape[0]) - x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond=tensor[a:b]) + x_out[a:b] = self.inner_model(x_in[a:b], sigma_in[a:b], cond={"c_crossattn": [tensor[a:b]], "c_concat": [image_cond_in[a:b]]}) - x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond=uncond) + x_out[-uncond.shape[0]:] = self.inner_model(x_in[-uncond.shape[0]:], sigma_in[-uncond.shape[0]:], cond={"c_crossattn": [uncond], "c_concat": [image_cond_in[-uncond.shape[0]:]]}) denoised_uncond = x_out[-uncond.shape[0]:] denoised = torch.clone(denoised_uncond) @@ -361,7 +377,7 @@ class KDiffusionSampler: return extra_params_kwargs - def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None): + def sample_img2img(self, p, x, noise, conditioning, unconditional_conditioning, steps=None, image_conditioning=None): steps, t_enc = setup_img2img_steps(p, steps) if p.sampler_noise_scheduler_override: @@ -389,11 +405,16 @@ class KDiffusionSampler: self.model_wrap_cfg.init_latent = x - samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, xi, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs)) + samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, xi, extra_args={ + 'cond': conditioning, + 'image_cond': image_conditioning, + 'uncond': unconditional_conditioning, + 'cond_scale': p.cfg_scale + }, disable=False, callback=self.callback_state, **extra_params_kwargs)) return samples - def sample(self, p, x, conditioning, unconditional_conditioning, steps=None): + def sample(self, p, x, conditioning, unconditional_conditioning, steps=None, image_conditioning = None): steps = steps or p.steps if p.sampler_noise_scheduler_override: @@ -414,7 +435,12 @@ class KDiffusionSampler: else: extra_params_kwargs['sigmas'] = sigmas - samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs)) + samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={ + 'cond': conditioning, + 'image_cond': image_conditioning, + 'uncond': unconditional_conditioning, + 'cond_scale': p.cfg_scale + }, disable=False, callback=self.callback_state, **extra_params_kwargs)) return samples -- cgit v1.2.3 From 0719c10bf1b817364a498ee11b90d30d3d527344 Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Wed, 19 Oct 2022 13:56:26 -0700 Subject: Fixed copying mistake --- modules/sd_hijack_inpainting.py | 79 +++++++++++++---------------------------- 1 file changed, 25 insertions(+), 54 deletions(-) (limited to 'modules') diff --git a/modules/sd_hijack_inpainting.py b/modules/sd_hijack_inpainting.py index 7e5670d6..d4d28d2e 100644 --- a/modules/sd_hijack_inpainting.py +++ b/modules/sd_hijack_inpainting.py @@ -19,63 +19,35 @@ from ldm.models.diffusion.ddim import DDIMSampler, noise_like # https://github.com/runwayml/stable-diffusion/blob/main/ldm/models/diffusion/ddim.py # ================================================================================================= @torch.no_grad() -def sample( - self, - S, - batch_size, - shape, - conditioning=None, - callback=None, - normals_sequence=None, - img_callback=None, - quantize_x0=False, - eta=0., - mask=None, - x0=None, - temperature=1., - noise_dropout=0., - score_corrector=None, - corrector_kwargs=None, - verbose=True, - x_T=None, - log_every_t=100, - unconditional_guidance_scale=1., - unconditional_conditioning=None, - # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ... - **kwargs - ): +def sample(self, + S, + batch_size, + shape, + conditioning=None, + callback=None, + normals_sequence=None, + img_callback=None, + quantize_x0=False, + eta=0., + mask=None, + x0=None, + temperature=1., + noise_dropout=0., + score_corrector=None, + corrector_kwargs=None, + verbose=True, + x_T=None, + log_every_t=100, + unconditional_guidance_scale=1., + unconditional_conditioning=None, + # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ... + **kwargs + ): if conditioning is not None: if isinstance(conditioning, dict): ctmp = conditioning[list(conditioning.keys())[0]] while isinstance(ctmp, list): - ctmp = elf.inpainting_fill == 2: - self.init_latent = self.init_latent * self.mask + create_random_tensors(self.init_latent.shape[1:], all_seeds[0:self.init_latent.shape[0]]) * self.nmask - elif self.inpainting_fill == 3: - self.init_latent = self.init_latent * self.mask - - if self.image_mask is not None: - conditioning_mask = np.array(self.image_mask.convert("L")) - conditioning_mask = conditioning_mask.astype(np.float32) / 255.0 - conditioning_mask = torch.from_numpy(conditioning_mask[None, None]) - - # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0 - conditioning_mask = torch.round(conditioning_mask) - else: - conditioning_mask = torch.ones(1, 1, *image.shape[-2:]) - - # Create another latent image, this time with a masked version of the original input. - conditioning_mask = conditioning_mask.to(image.device) - conditioning_image = image * (1.0 - conditioning_mask) - conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image)) - - # Create the concatenated conditioning tensor to be fed to `c_concat` - conditioning_mask = torch.nn.functional.interpolate(conditioning_mask, size=self.init_latent.shape[-2:]) - conditioning_mask = conditioning_mask.expand(conditioning_image.shape[0], -1, -1, -1) - self.image_conditioning = torch.cat([conditioning_mask, conditioning_image], dim=1) - self.image_conditioning = self.image_conditioning.to(shared.device).type(self.sd_model.dtype) - - def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): - x = create_random_tensors([opctmp[0] + ctmp = ctmp[0] cbs = ctmp.shape[0] if cbs != batch_size: print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}") @@ -106,7 +78,6 @@ def sample( ) return samples, intermediates - @torch.no_grad() def p_sample_ddim(self, x, c, t, index, repeat_noise=False, use_original_steps=False, quantize_denoised=False, temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None, -- cgit v1.2.3 From dde9f960727bfe151d418e43685a2881cf580a17 Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Wed, 19 Oct 2022 14:14:24 -0700 Subject: added support for ddim img2img --- modules/sd_samplers.py | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'modules') diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index 9d3cf289..d270e4df 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -208,6 +208,12 @@ class VanillaStableDiffusionSampler: self.init_latent = x self.step = 0 + # Wrap the conditioning models with additional image conditioning for inpainting model + if image_conditioning is not None: + conditioning = {"c_concat": [image_conditioning], "c_crossattn": [conditioning]} + unconditional_conditioning = {"c_concat": [image_conditioning], "c_crossattn": [unconditional_conditioning]} + + samples = self.launch_sampling(steps, lambda: self.sampler.decode(x1, conditioning, t_enc, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning)) return samples -- cgit v1.2.3 From c418467c03db916c3e5312e6ac4a67365e196dbd Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Wed, 19 Oct 2022 15:09:43 -0700 Subject: Don't compute latent mask if were not using it. Also added support for fixed highres_fix generation. --- modules/processing.py | 72 +++++++++++++++++++++++++++++++------------------- modules/sd_samplers.py | 4 +++ 2 files changed, 49 insertions(+), 27 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index a6c308f9..684e5833 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -541,12 +541,8 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.truncate_y = int(self.firstphase_height - firstphase_height_truncated) // opt_f - def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): - self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model) - - if not self.enable_hr: - x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) - + def create_dummy_mask(self, x): + if self.sampler.conditioning_key in {'hybrid', 'concat'}: # The "masked-image" in this case will just be all zeros since the entire image is masked. image_conditioning = torch.zeros(x.shape[0], 3, self.height, self.width, device=x.device) image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning)) @@ -555,11 +551,23 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): image_conditioning = torch.nn.functional.pad(image_conditioning, (0, 0, 0, 0, 1, 0), value=1.0) image_conditioning = image_conditioning.to(x.dtype) - samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=image_conditioning) + else: + # Dummy zero conditioning if we're not using inpainting model. + # Still takes up a bit of memory, but no encoder call. + image_conditioning = torch.zeros(x.shape[0], 5, x.shape[-2], x.shape[-1], dtype=x.dtype, device=x.device) + + return image_conditioning + + def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): + self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers, self.sampler_index, self.sd_model) + + if not self.enable_hr: + x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) + samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.create_dummy_mask(x)) return samples x = create_random_tensors([opt_C, self.firstphase_height // opt_f, self.firstphase_width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) - samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning) + samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.create_dummy_mask(x)) samples = samples[:, :, self.truncate_y//2:samples.shape[2]-self.truncate_y//2, self.truncate_x//2:samples.shape[3]-self.truncate_x//2] @@ -596,7 +604,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): x = None devices.torch_gc() - samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps) + samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps, image_conditioning=self.create_dummy_mask(samples)) return samples @@ -723,26 +731,36 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): elif self.inpainting_fill == 3: self.init_latent = self.init_latent * self.mask - if self.image_mask is not None: - conditioning_mask = np.array(self.image_mask.convert("L")) - conditioning_mask = conditioning_mask.astype(np.float32) / 255.0 - conditioning_mask = torch.from_numpy(conditioning_mask[None, None]) + conditioning_key = self.sampler.conditioning_key - # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0 - conditioning_mask = torch.round(conditioning_mask) + if conditioning_key in {'hybrid', 'concat'}: + if self.image_mask is not None: + conditioning_mask = np.array(self.image_mask.convert("L")) + conditioning_mask = conditioning_mask.astype(np.float32) / 255.0 + conditioning_mask = torch.from_numpy(conditioning_mask[None, None]) + + # Inpainting model uses a discretized mask as input, so we round to either 1.0 or 0.0 + conditioning_mask = torch.round(conditioning_mask) + else: + conditioning_mask = torch.ones(1, 1, *image.shape[-2:]) + + # Create another latent image, this time with a masked version of the original input. + conditioning_mask = conditioning_mask.to(image.device) + conditioning_image = image * (1.0 - conditioning_mask) + conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image)) + + # Create the concatenated conditioning tensor to be fed to `c_concat` + conditioning_mask = torch.nn.functional.interpolate(conditioning_mask, size=self.init_latent.shape[-2:]) + conditioning_mask = conditioning_mask.expand(conditioning_image.shape[0], -1, -1, -1) + self.image_conditioning = torch.cat([conditioning_mask, conditioning_image], dim=1) + self.image_conditioning = self.image_conditioning.to(shared.device).type(self.sd_model.dtype) else: - conditioning_mask = torch.ones(1, 1, *image.shape[-2:]) - - # Create another latent image, this time with a masked version of the original input. - conditioning_mask = conditioning_mask.to(image.device) - conditioning_image = image * (1.0 - conditioning_mask) - conditioning_image = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(conditioning_image)) - - # Create the concatenated conditioning tensor to be fed to `c_concat` - conditioning_mask = torch.nn.functional.interpolate(conditioning_mask, size=self.init_latent.shape[-2:]) - conditioning_mask = conditioning_mask.expand(conditioning_image.shape[0], -1, -1, -1) - self.image_conditioning = torch.cat([conditioning_mask, conditioning_image], dim=1) - self.image_conditioning = self.image_conditioning.to(shared.device).type(self.sd_model.dtype) + self.image_conditioning = torch.zeros( + self.init_latent.shape[0], 5, self.init_latent.shape[-2], self.init_latent.shape[-1], + dtype=self.init_latent.dtype, + device=self.init_latent.device + ) + def sample(self, conditioning, unconditional_conditioning, seeds, subseeds, subseed_strength): x = create_random_tensors([opt_C, self.height // opt_f, self.width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index d270e4df..c21be26e 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -117,6 +117,8 @@ class VanillaStableDiffusionSampler: self.config = None self.last_latent = None + self.conditioning_key = sd_model.model.conditioning_key + def number_of_needed_noises(self, p): return 0 @@ -328,6 +330,8 @@ class KDiffusionSampler: self.config = None self.last_latent = None + self.conditioning_key = sd_model.model.conditioning_key + def callback_state(self, d): step = d['i'] latent = d["denoised"] -- cgit v1.2.3 From d6ea5841374a28f3f6deb73abc251c8f0bcb240f Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:07:57 +0100 Subject: change html output --- modules/hypernetworks/hypernetwork.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 7d519cd9..73c1cb80 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -380,7 +380,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log Loss: {mean_loss:.7f}
Step: {hypernetwork.step}
Last prompt: {html.escape(entries[0].cond_text)}
-Last saved embedding: {html.escape(last_saved_file)}
+Last saved hypernetwork: {html.escape(last_saved_file)}
Last saved image: {html.escape(last_saved_image)}

""" -- cgit v1.2.3 From 166be3919b817cee5e702fd01c34afe9081b952c Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:09:40 +0100 Subject: allow overwrite old hn --- modules/hypernetworks/ui.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 08f75f15..f45345ea 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -10,9 +10,10 @@ from modules import sd_hijack, shared, devices from modules.hypernetworks import hypernetwork -def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm=False): +def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, add_layer_norm=False): fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") - assert not os.path.exists(fn), f"file {fn} already exists" + if not overwrite_old: + assert not os.path.exists(fn), f"file {fn} already exists" if type(layer_structure) == str: layer_structure = tuple(map(int, re.sub(r'\D', '', layer_structure))) -- cgit v1.2.3 From 0087079c2d487b67b06ffc30f36ce486a74e6318 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:10:59 +0100 Subject: allow overwrite old embedding --- modules/textual_inversion/textual_inversion.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index 3be69562..5776778b 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -153,7 +153,7 @@ class EmbeddingDatabase: return None, None -def create_embedding(name, num_vectors_per_token, init_text='*'): +def create_embedding(name, num_vectors_per_token, overwrite_old, init_text='*'): cond_model = shared.sd_model.cond_stage_model embedding_layer = cond_model.wrapped.transformer.text_model.embeddings @@ -165,7 +165,8 @@ def create_embedding(name, num_vectors_per_token, init_text='*'): vec[i] = embedded[i * int(embedded.shape[0]) // num_vectors_per_token] fn = os.path.join(shared.cmd_opts.embeddings_dir, f"{name}.pt") - assert not os.path.exists(fn), f"file {fn} already exists" + if not overwrite_old: + assert not os.path.exists(fn), f"file {fn} already exists" embedding = Embedding(vec, name) embedding.step = 0 -- cgit v1.2.3 From 632e8d660293081cadb145d8062e5aff0a4a8f0d Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:19:40 +0100 Subject: split learn rates --- modules/ui.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index cdb9d335..d07184ee 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1342,7 +1342,7 @@ def create_ui(wrap_gradio_gpu_call): _js="start_training_textual_inversion", inputs=[ train_embedding_name, - learn_rate, + embedding_learn_rate, batch_size, dataset_directory, log_directory, @@ -1367,7 +1367,7 @@ def create_ui(wrap_gradio_gpu_call): _js="start_training_textual_inversion", inputs=[ train_hypernetwork_name, - learn_rate, + hypernetwork_learn_rate, batch_size, dataset_directory, log_directory, -- cgit v1.2.3 From c3835ec85cbb44fa3c46fa871c622b6fee235c89 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:24:24 +0100 Subject: pass overwrite old flag --- modules/textual_inversion/ui.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/textual_inversion/ui.py b/modules/textual_inversion/ui.py index 36881e7a..e712284d 100644 --- a/modules/textual_inversion/ui.py +++ b/modules/textual_inversion/ui.py @@ -7,8 +7,8 @@ import modules.textual_inversion.preprocess from modules import sd_hijack, shared -def create_embedding(name, initialization_text, nvpt): - filename = modules.textual_inversion.textual_inversion.create_embedding(name, nvpt, init_text=initialization_text) +def create_embedding(name, initialization_text, nvpt, overwrite_old): + filename = modules.textual_inversion.textual_inversion.create_embedding(name, nvpt, overwrite_old, init_text=initialization_text) sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings() -- cgit v1.2.3 From 4d6b9f76a55fd0ac0f72634071032dd9c6efb409 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:27:16 +0100 Subject: reorder create_hypernetwork params --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index d07184ee..322c082b 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1307,9 +1307,9 @@ def create_ui(wrap_gradio_gpu_call): inputs=[ new_hypernetwork_name, new_hypernetwork_sizes, + overwrite_old_hypernetwork, new_hypernetwork_layer_structure, new_hypernetwork_add_layer_norm, - overwrite_old_hypernetwork, ], outputs=[ train_hypernetwork_name, -- cgit v1.2.3 From fbcce66601994f6ed370db36d9c238840fed6bd2 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:46:54 +0100 Subject: add existing caption file handling --- modules/textual_inversion/preprocess.py | 32 ++++++++++++++++++++++++-------- 1 file changed, 24 insertions(+), 8 deletions(-) (limited to 'modules') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 886cf0c3..5c43fe13 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -48,7 +48,7 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro shared.state.textinfo = "Preprocessing..." shared.state.job_count = len(files) - def save_pic_with_caption(image, index): + def save_pic_with_caption(image, index, existing_caption=None): caption = "" if process_caption: @@ -66,17 +66,26 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro basename = f"{index:05}-{subindex[0]}-{filename_part}" image.save(os.path.join(dst, f"{basename}.png")) + if preprocess_txt_action == 'prepend' and existing_caption: + caption = existing_caption + ' ' + caption + elif preprocess_txt_action == 'append' and existing_caption: + caption = caption + ' ' + existing_caption + elif preprocess_txt_action == 'copy' and existing_caption: + caption = existing_caption + + caption = caption.strip() + if len(caption) > 0: with open(os.path.join(dst, f"{basename}.txt"), "w", encoding="utf8") as file: file.write(caption) subindex[0] += 1 - def save_pic(image, index): + def save_pic(image, index, existing_caption=None): save_pic_with_caption(image, index) if process_flip: - save_pic_with_caption(ImageOps.mirror(image), index) + save_pic_with_caption(ImageOps.mirror(image), index, existing_caption=existing_caption) for index, imagefile in enumerate(tqdm.tqdm(files)): subindex = [0] @@ -86,6 +95,13 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro except Exception: continue + existing_caption = None + + try: + existing_caption = open(os.path.splitext(filename)[0] + '.txt', 'r').read() + except Exception as e: + print(e) + if shared.state.interrupted: break @@ -97,20 +113,20 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pro img = img.resize((width, height * img.height // img.width)) top = img.crop((0, 0, width, height)) - save_pic(top, index) + save_pic(top, index, existing_caption=existing_caption) bot = img.crop((0, img.height - height, width, img.height)) - save_pic(bot, index) + save_pic(bot, index, existing_caption=existing_caption) elif process_split and is_wide: img = img.resize((width * img.width // img.height, height)) left = img.crop((0, 0, width, height)) - save_pic(left, index) + save_pic(left, index, existing_caption=existing_caption) right = img.crop((img.width - width, 0, img.width, height)) - save_pic(right, index) + save_pic(right, index, existing_caption=existing_caption) else: img = images.resize_image(1, img, width, height) - save_pic(img, index) + save_pic(img, index, existing_caption=existing_caption) shared.state.nextjob() -- cgit v1.2.3 From ab353b141df8eee042b0964bcb645015dabf3459 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:48:07 +0100 Subject: link existing txt option --- modules/ui.py | 2 ++ 1 file changed, 2 insertions(+) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 322c082b..7f52ac0c 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1234,6 +1234,7 @@ def create_ui(wrap_gradio_gpu_call): process_dst = gr.Textbox(label='Destination directory') process_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512) process_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) + preprocess_txt_action = gr.Dropdown(label='Existing Caption txt Action', choices=['ignore', 'copy', 'prepend', 'append']) with gr.Row(): process_flip = gr.Checkbox(label='Create flipped copies') @@ -1326,6 +1327,7 @@ def create_ui(wrap_gradio_gpu_call): process_dst, process_width, process_height, + preprocess_txt_action, process_flip, process_split, process_caption, -- cgit v1.2.3 From 9b65c4ecf4f8eb6187ee721918adebe68e9bc631 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:49:23 +0100 Subject: pass preprocess_txt_action param --- modules/textual_inversion/preprocess.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 5c43fe13..3713bc89 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -11,7 +11,7 @@ if cmd_opts.deepdanbooru: import modules.deepbooru as deepbooru -def preprocess(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False): try: if process_caption: shared.interrogator.load() @@ -21,7 +21,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ db_opts[deepbooru.OPT_INCLUDE_RANKS] = False deepbooru.create_deepbooru_process(opts.interrogate_deepbooru_score_threshold, db_opts) - preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru) + preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru) finally: @@ -33,7 +33,7 @@ def preprocess(process_src, process_dst, process_width, process_height, process_ -def preprocess_work(process_src, process_dst, process_width, process_height, process_flip, process_split, process_caption, process_caption_deepbooru=False): +def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False): width = process_width height = process_height src = os.path.abspath(process_src) -- cgit v1.2.3 From 55d8c6cce6d3aef848b9f194adad2ce53064d8b7 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 00:53:29 +0100 Subject: default to ignore existing captions --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 7f52ac0c..bd5f1b05 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1234,7 +1234,7 @@ def create_ui(wrap_gradio_gpu_call): process_dst = gr.Textbox(label='Destination directory') process_width = gr.Slider(minimum=64, maximum=2048, step=64, label="Width", value=512) process_height = gr.Slider(minimum=64, maximum=2048, step=64, label="Height", value=512) - preprocess_txt_action = gr.Dropdown(label='Existing Caption txt Action', choices=['ignore', 'copy', 'prepend', 'append']) + preprocess_txt_action = gr.Dropdown(label='Existing Caption txt Action', value="ignore", choices=["ignore", "copy", "prepend", "append"]) with gr.Row(): process_flip = gr.Checkbox(label='Create flipped copies') -- cgit v1.2.3 From 6f98e89486f55b0e4657e96ce640cf1c4675d187 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Thu, 20 Oct 2022 00:10:45 +0000 Subject: update --- modules/hypernetworks/hypernetwork.py | 29 +++++++++++++++-------- modules/hypernetworks/ui.py | 3 ++- modules/ui.py | 43 +++++++++++++++++++---------------- 3 files changed, 44 insertions(+), 31 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 74300122..7d617680 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -22,16 +22,20 @@ from modules.textual_inversion.learn_schedule import LearnRateScheduler class HypernetworkModule(torch.nn.Module): multiplier = 1.0 - def __init__(self, dim, state_dict=None, layer_structure=None, add_layer_norm=False): + def __init__(self, dim, state_dict=None, layer_structure=None, add_layer_norm=False, activation_func=None): super().__init__() - assert layer_structure is not None, "layer_structure mut not be None" + assert layer_structure is not None, "layer_structure must not be None" assert layer_structure[0] == 1, "Multiplier Sequence should start with size 1!" assert layer_structure[-1] == 1, "Multiplier Sequence should end with size 1!" linears = [] for i in range(len(layer_structure) - 1): linears.append(torch.nn.Linear(int(dim * layer_structure[i]), int(dim * layer_structure[i+1]))) + if activation_func == "relu": + linears.append(torch.nn.ReLU()) + if activation_func == "leakyrelu": + linears.append(torch.nn.LeakyReLU()) if add_layer_norm: linears.append(torch.nn.LayerNorm(int(dim * layer_structure[i+1]))) @@ -42,8 +46,9 @@ class HypernetworkModule(torch.nn.Module): self.load_state_dict(state_dict) else: for layer in self.linear: - layer.weight.data.normal_(mean=0.0, std=0.01) - layer.bias.data.zero_() + if not "ReLU" in layer.__str__(): + layer.weight.data.normal_(mean=0.0, std=0.01) + layer.bias.data.zero_() self.to(devices.device) @@ -69,7 +74,8 @@ class HypernetworkModule(torch.nn.Module): def trainables(self): layer_structure = [] for layer in self.linear: - layer_structure += [layer.weight, layer.bias] + if not "ReLU" in layer.__str__(): + layer_structure += [layer.weight, layer.bias] return layer_structure @@ -81,7 +87,7 @@ class Hypernetwork: filename = None name = None - def __init__(self, name=None, enable_sizes=None, layer_structure=None, add_layer_norm=False): + def __init__(self, name=None, enable_sizes=None, layer_structure=None, add_layer_norm=False, activation_func=None): self.filename = None self.name = name self.layers = {} @@ -90,11 +96,12 @@ class Hypernetwork: self.sd_checkpoint_name = None self.layer_structure = layer_structure self.add_layer_norm = add_layer_norm + self.activation_func = activation_func for size in enable_sizes or []: self.layers[size] = ( - HypernetworkModule(size, None, self.layer_structure, self.add_layer_norm), - HypernetworkModule(size, None, self.layer_structure, self.add_layer_norm), + HypernetworkModule(size, None, self.layer_structure, self.add_layer_norm, self.activation_func), + HypernetworkModule(size, None, self.layer_structure, self.add_layer_norm, self.activation_func), ) def weights(self): @@ -117,6 +124,7 @@ class Hypernetwork: state_dict['name'] = self.name state_dict['layer_structure'] = self.layer_structure state_dict['is_layer_norm'] = self.add_layer_norm + state_dict['activation_func'] = self.activation_func state_dict['sd_checkpoint'] = self.sd_checkpoint state_dict['sd_checkpoint_name'] = self.sd_checkpoint_name @@ -131,12 +139,13 @@ class Hypernetwork: self.layer_structure = state_dict.get('layer_structure', [1, 2, 1]) self.add_layer_norm = state_dict.get('is_layer_norm', False) + self.activation_func = state_dict.get('activation_func', None) for size, sd in state_dict.items(): if type(size) == int: self.layers[size] = ( - HypernetworkModule(size, sd[0], self.layer_structure, self.add_layer_norm), - HypernetworkModule(size, sd[1], self.layer_structure, self.add_layer_norm), + HypernetworkModule(size, sd[0], self.layer_structure, self.add_layer_norm, self.activation_func), + HypernetworkModule(size, sd[1], self.layer_structure, self.add_layer_norm, self.activation_func), ) self.name = state_dict.get('name', self.name) diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 08f75f15..83f9547b 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -10,7 +10,7 @@ from modules import sd_hijack, shared, devices from modules.hypernetworks import hypernetwork -def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm=False): +def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm=False, activation_func=None): fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") assert not os.path.exists(fn), f"file {fn} already exists" @@ -22,6 +22,7 @@ def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm enable_sizes=[int(x) for x in enable_sizes], layer_structure=layer_structure, add_layer_norm=add_layer_norm, + activation_func=activation_func, ) hypernet.save(fn) diff --git a/modules/ui.py b/modules/ui.py index d2e24880..8751fa9c 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -5,43 +5,44 @@ import json import math import mimetypes import os +import platform import random +import subprocess as sp import sys import tempfile import time import traceback -import platform -import subprocess as sp from functools import partial, reduce +import gradio as gr +import gradio.routes +import gradio.utils import numpy as np +import piexif import torch from PIL import Image, PngImagePlugin -import piexif -import gradio as gr -import gradio.utils -import gradio.routes - -from modules import sd_hijack, sd_models, localization +from modules import localization, sd_hijack, sd_models from modules.paths import script_path -from modules.shared import opts, cmd_opts, restricted_opts +from modules.shared import cmd_opts, opts, restricted_opts + if cmd_opts.deepdanbooru: from modules.deepbooru import get_deepbooru_tags -import modules.shared as shared -from modules.sd_samplers import samplers, samplers_for_img2img -from modules.sd_hijack import model_hijack + +import modules.codeformer_model +import modules.generation_parameters_copypaste +import modules.gfpgan_model +import modules.hypernetworks.ui +import modules.images_history as img_his import modules.ldsr_model import modules.scripts -import modules.gfpgan_model -import modules.codeformer_model +import modules.shared as shared import modules.styles -import modules.generation_parameters_copypaste +import modules.textual_inversion.ui from modules import prompt_parser from modules.images import save_image -import modules.textual_inversion.ui -import modules.hypernetworks.ui -import modules.images_history as img_his +from modules.sd_hijack import model_hijack +from modules.sd_samplers import samplers, samplers_for_img2img # this is a fix for Windows users. Without it, javascript files will be served with text/html content-type and the browser will not show any UI mimetypes.init() @@ -268,8 +269,8 @@ def calc_time_left(progress, threshold, label, force_display): time_since_start = time.time() - shared.state.time_start eta = (time_since_start/progress) eta_relative = eta-time_since_start - if (eta_relative > threshold and progress > 0.02) or force_display: - return label + time.strftime('%H:%M:%S', time.gmtime(eta_relative)) + if (eta_relative > threshold and progress > 0.02) or force_display: + return label + time.strftime('%H:%M:%S', time.gmtime(eta_relative)) else: return "" @@ -1219,6 +1220,7 @@ def create_ui(wrap_gradio_gpu_call): new_hypernetwork_sizes = gr.CheckboxGroup(label="Modules", value=["768", "320", "640", "1280"], choices=["768", "320", "640", "1280"]) new_hypernetwork_layer_structure = gr.Textbox("1, 2, 1", label="Enter hypernetwork layer structure", placeholder="1st and last digit must be 1. ex:'1, 2, 1'") new_hypernetwork_add_layer_norm = gr.Checkbox(label="Add layer normalization") + new_hypernetwork_activation_func = gr.Dropdown(value="relu", label="Select activation function of hypernetwork", choices=["relu", "leakyrelu"]) with gr.Row(): with gr.Column(scale=3): @@ -1303,6 +1305,7 @@ def create_ui(wrap_gradio_gpu_call): new_hypernetwork_sizes, new_hypernetwork_layer_structure, new_hypernetwork_add_layer_norm, + new_hypernetwork_activation_func, ], outputs=[ train_hypernetwork_name, -- cgit v1.2.3 From ba469343e6a1c6e23e82acf5feb65c6101dacbb2 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Thu, 20 Oct 2022 00:17:04 +0000 Subject: align ui.py imports with upstream --- modules/ui.py | 37 ++++++++++++++++++------------------- 1 file changed, 18 insertions(+), 19 deletions(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 987b1d7d..913b23b4 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -5,44 +5,43 @@ import json import math import mimetypes import os -import platform import random -import subprocess as sp import sys import tempfile import time import traceback +import platform +import subprocess as sp from functools import partial, reduce -import gradio as gr -import gradio.routes -import gradio.utils import numpy as np -import piexif import torch from PIL import Image, PngImagePlugin +import piexif -from modules import localization, sd_hijack, sd_models -from modules.paths import script_path -from modules.shared import cmd_opts, opts, restricted_opts +import gradio as gr +import gradio.utils +import gradio.routes +from modules import sd_hijack, sd_models, localization +from modules.paths import script_path +from modules.shared import opts, cmd_opts, restricted_opts if cmd_opts.deepdanbooru: from modules.deepbooru import get_deepbooru_tags - -import modules.codeformer_model -import modules.generation_parameters_copypaste -import modules.gfpgan_model -import modules.hypernetworks.ui -import modules.images_history as img_his +import modules.shared as shared +from modules.sd_samplers import samplers, samplers_for_img2img +from modules.sd_hijack import model_hijack import modules.ldsr_model import modules.scripts -import modules.shared as shared +import modules.gfpgan_model +import modules.codeformer_model import modules.styles -import modules.textual_inversion.ui +import modules.generation_parameters_copypaste from modules import prompt_parser from modules.images import save_image -from modules.sd_hijack import model_hijack -from modules.sd_samplers import samplers, samplers_for_img2img +import modules.textual_inversion.ui +import modules.hypernetworks.ui +import modules.images_history as img_his # this is a fix for Windows users. Without it, javascript files will be served with text/html content-type and the browser will not show any UI mimetypes.init() -- cgit v1.2.3 From 858462f719c22ca9f24b94a41699653c34b5f4fb Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Thu, 20 Oct 2022 02:57:18 +0100 Subject: do caption copy for both flips --- modules/textual_inversion/preprocess.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/textual_inversion/preprocess.py b/modules/textual_inversion/preprocess.py index 3713bc89..6bba3852 100644 --- a/modules/textual_inversion/preprocess.py +++ b/modules/textual_inversion/preprocess.py @@ -82,7 +82,7 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre subindex[0] += 1 def save_pic(image, index, existing_caption=None): - save_pic_with_caption(image, index) + save_pic_with_caption(image, index, existing_caption=existing_caption) if process_flip: save_pic_with_caption(ImageOps.mirror(image), index, existing_caption=existing_caption) -- cgit v1.2.3 From aa7ff2a1972f3865883e10ba28c5414cdebe8e3b Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Wed, 19 Oct 2022 21:46:13 -0700 Subject: Fixed non-square highres fix generation --- modules/processing.py | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index 684e5833..3caac25e 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -541,10 +541,13 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.truncate_y = int(self.firstphase_height - firstphase_height_truncated) // opt_f - def create_dummy_mask(self, x): + def create_dummy_mask(self, x, first_phase: bool = False): if self.sampler.conditioning_key in {'hybrid', 'concat'}: + height = self.firstphase_height if first_phase else self.height + width = self.firstphase_width if first_phase else self.width + # The "masked-image" in this case will just be all zeros since the entire image is masked. - image_conditioning = torch.zeros(x.shape[0], 3, self.height, self.width, device=x.device) + image_conditioning = torch.zeros(x.shape[0], 3, height, width, device=x.device) image_conditioning = self.sd_model.get_first_stage_encoding(self.sd_model.encode_first_stage(image_conditioning)) # Add the fake full 1s mask to the first dimension. @@ -567,7 +570,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): return samples x = create_random_tensors([opt_C, self.firstphase_height // opt_f, self.firstphase_width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) - samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.create_dummy_mask(x)) + samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.create_dummy_mask(x, first_phase=True)) samples = samples[:, :, self.truncate_y//2:samples.shape[2]-self.truncate_y//2, self.truncate_x//2:samples.shape[3]-self.truncate_x//2] -- cgit v1.2.3 From 930b4c64f7dbce6918894d53538003e5959fd022 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Thu, 20 Oct 2022 08:18:02 +0300 Subject: allow float sizes for hypernet's layer_structure --- modules/hypernetworks/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 08f75f15..e0741d08 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -15,7 +15,7 @@ def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm assert not os.path.exists(fn), f"file {fn} already exists" if type(layer_structure) == str: - layer_structure = tuple(map(int, re.sub(r'\D', '', layer_structure))) + layer_structure = [float(x.strip()) for x in layer_structure.split(",")] hypernet = modules.hypernetworks.hypernetwork.Hypernetwork( name=name, -- cgit v1.2.3 From f8733ad08be08bafb40f4299785590e11f049e96 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Thu, 20 Oct 2022 11:07:37 +0000 Subject: add linear as a act func (option for doin nothing) --- modules/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/ui.py b/modules/ui.py index 913b23b4..716f14b8 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1224,7 +1224,7 @@ def create_ui(wrap_gradio_gpu_call): new_hypernetwork_sizes = gr.CheckboxGroup(label="Modules", value=["768", "320", "640", "1280"], choices=["768", "320", "640", "1280"]) new_hypernetwork_layer_structure = gr.Textbox("1, 2, 1", label="Enter hypernetwork layer structure", placeholder="1st and last digit must be 1. ex:'1, 2, 1'") new_hypernetwork_add_layer_norm = gr.Checkbox(label="Add layer normalization") - new_hypernetwork_activation_func = gr.Dropdown(value="relu", label="Select activation function of hypernetwork", choices=["relu", "leakyrelu"]) + new_hypernetwork_activation_func = gr.Dropdown(value="relu", label="Select activation function of hypernetwork", choices=["linear", "relu", "leakyrelu"]) with gr.Row(): with gr.Column(scale=3): -- cgit v1.2.3 From 92a17a7a4a13fceb3c3e25a2e854b2a7dd6eb5df Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Thu, 20 Oct 2022 09:45:03 -0700 Subject: Made dummy latents smaller. Minor code cleanups --- modules/processing.py | 7 ++++--- modules/sd_samplers.py | 6 ++++-- 2 files changed, 8 insertions(+), 5 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index 3caac25e..539cde38 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -557,7 +557,8 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): else: # Dummy zero conditioning if we're not using inpainting model. # Still takes up a bit of memory, but no encoder call. - image_conditioning = torch.zeros(x.shape[0], 5, x.shape[-2], x.shape[-1], dtype=x.dtype, device=x.device) + # Pretty sure we can just make this a 1x1 image since its not going to be used besides its batch size. + image_conditioning = torch.zeros(x.shape[0], 5, 1, 1, dtype=x.dtype, device=x.device) return image_conditioning @@ -759,8 +760,8 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): self.image_conditioning = self.image_conditioning.to(shared.device).type(self.sd_model.dtype) else: self.image_conditioning = torch.zeros( - self.init_latent.shape[0], 5, self.init_latent.shape[-2], self.init_latent.shape[-1], - dtype=self.init_latent.dtype, + self.init_latent.shape[0], 5, 1, 1, + dtype=self.init_latent.dtype, device=self.init_latent.device ) diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index c21be26e..cc682593 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -138,7 +138,7 @@ class VanillaStableDiffusionSampler: if self.stop_at is not None and self.step > self.stop_at: raise InterruptedException - # Have to unwrap the inpainting conditioning here to perform pre-preocessing + # Have to unwrap the inpainting conditioning here to perform pre-processing image_conditioning = None if isinstance(cond, dict): image_conditioning = cond["c_concat"][0] @@ -146,7 +146,7 @@ class VanillaStableDiffusionSampler: unconditional_conditioning = unconditional_conditioning["c_crossattn"][0] conds_list, tensor = prompt_parser.reconstruct_multicond_batch(cond, self.step) - unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step) + unconditional_conditioning = prompt_parser.reconstruct_cond_batch(unconditional_conditioning, self.step) assert all([len(conds) == 1 for conds in conds_list]), 'composition via AND is not supported for DDIM/PLMS samplers' cond = tensor @@ -165,6 +165,8 @@ class VanillaStableDiffusionSampler: img_orig = self.sampler.model.q_sample(self.init_latent, ts) x_dec = img_orig * self.mask + self.nmask * x_dec + # Wrap the image conditioning back up since the DDIM code can accept the dict directly. + # Note that they need to be lists because it just concatenates them later. if image_conditioning is not None: cond = {"c_concat": [image_conditioning], "c_crossattn": [cond]} unconditional_conditioning = {"c_concat": [image_conditioning], "c_crossattn": [unconditional_conditioning]} -- cgit v1.2.3 From d1cb08bfb221cd1b0cfc6078162b4e206ea80a5c Mon Sep 17 00:00:00 2001 From: Vladimir Repin <32306715+mezotaken@users.noreply.github.com> Date: Thu, 20 Oct 2022 22:49:06 +0300 Subject: fix skip and interrupt for highres. fix option --- modules/processing.py | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index bcb0c32c..6324ca91 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -587,9 +587,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): x = None devices.torch_gc() - samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps) - - return samples + return self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps) or samples class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): -- cgit v1.2.3 From 708c3a7bd8ce68cbe1aa7c268e5a4b1980affc9f Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Thu, 20 Oct 2022 13:28:43 -0700 Subject: Added PLMS hijack and made sure to always replace methods --- modules/sd_hijack_inpainting.py | 163 ++++++++++++++++++++++++++++++++++++++-- modules/sd_models.py | 3 +- 2 files changed, 157 insertions(+), 9 deletions(-) (limited to 'modules') diff --git a/modules/sd_hijack_inpainting.py b/modules/sd_hijack_inpainting.py index d4d28d2e..43938071 100644 --- a/modules/sd_hijack_inpainting.py +++ b/modules/sd_hijack_inpainting.py @@ -1,16 +1,14 @@ import torch -import numpy as np -from tqdm import tqdm -from einops import rearrange, repeat +from einops import repeat from omegaconf import ListConfig -from types import MethodType - import ldm.models.diffusion.ddpm import ldm.models.diffusion.ddim +import ldm.models.diffusion.plms from ldm.models.diffusion.ddpm import LatentDiffusion +from ldm.models.diffusion.plms import PLMSSampler from ldm.models.diffusion.ddim import DDIMSampler, noise_like # ================================================================================================= @@ -19,7 +17,7 @@ from ldm.models.diffusion.ddim import DDIMSampler, noise_like # https://github.com/runwayml/stable-diffusion/blob/main/ldm/models/diffusion/ddim.py # ================================================================================================= @torch.no_grad() -def sample(self, +def sample_ddim(self, S, batch_size, shape, @@ -132,6 +130,153 @@ def p_sample_ddim(self, x, c, t, index, repeat_noise=False, use_original_steps=F return x_prev, pred_x0 +# ================================================================================================= +# Monkey patch PLMSSampler methods. +# This one was not actually patched correctly in the RunwayML repo, but we can replicate the changes. +# Adapted from: +# https://github.com/CompVis/stable-diffusion/blob/main/ldm/models/diffusion/plms.py +# ================================================================================================= +@torch.no_grad() +def sample_plms(self, + S, + batch_size, + shape, + conditioning=None, + callback=None, + normals_sequence=None, + img_callback=None, + quantize_x0=False, + eta=0., + mask=None, + x0=None, + temperature=1., + noise_dropout=0., + score_corrector=None, + corrector_kwargs=None, + verbose=True, + x_T=None, + log_every_t=100, + unconditional_guidance_scale=1., + unconditional_conditioning=None, + # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ... + **kwargs + ): + if conditioning is not None: + if isinstance(conditioning, dict): + ctmp = conditioning[list(conditioning.keys())[0]] + while isinstance(ctmp, list): + ctmp = ctmp[0] + cbs = ctmp.shape[0] + if cbs != batch_size: + print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}") + else: + if conditioning.shape[0] != batch_size: + print(f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}") + + self.make_schedule(ddim_num_steps=S, ddim_eta=eta, verbose=verbose) + # sampling + C, H, W = shape + size = (batch_size, C, H, W) + print(f'Data shape for PLMS sampling is {size}') + + samples, intermediates = self.plms_sampling(conditioning, size, + callback=callback, + img_callback=img_callback, + quantize_denoised=quantize_x0, + mask=mask, x0=x0, + ddim_use_original_steps=False, + noise_dropout=noise_dropout, + temperature=temperature, + score_corrector=score_corrector, + corrector_kwargs=corrector_kwargs, + x_T=x_T, + log_every_t=log_every_t, + unconditional_guidance_scale=unconditional_guidance_scale, + unconditional_conditioning=unconditional_conditioning, + ) + return samples, intermediates + + +@torch.no_grad() +def p_sample_plms(self, x, c, t, index, repeat_noise=False, use_original_steps=False, quantize_denoised=False, + temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None, + unconditional_guidance_scale=1., unconditional_conditioning=None, old_eps=None, t_next=None): + b, *_, device = *x.shape, x.device + + def get_model_output(x, t): + if unconditional_conditioning is None or unconditional_guidance_scale == 1.: + e_t = self.model.apply_model(x, t, c) + else: + x_in = torch.cat([x] * 2) + t_in = torch.cat([t] * 2) + + if isinstance(c, dict): + assert isinstance(unconditional_conditioning, dict) + c_in = dict() + for k in c: + if isinstance(c[k], list): + c_in[k] = [ + torch.cat([unconditional_conditioning[k][i], c[k][i]]) + for i in range(len(c[k])) + ] + else: + c_in[k] = torch.cat([unconditional_conditioning[k], c[k]]) + else: + c_in = torch.cat([unconditional_conditioning, c]) + + e_t_uncond, e_t = self.model.apply_model(x_in, t_in, c_in).chunk(2) + e_t = e_t_uncond + unconditional_guidance_scale * (e_t - e_t_uncond) + + if score_corrector is not None: + assert self.model.parameterization == "eps" + e_t = score_corrector.modify_score(self.model, e_t, x, t, c, **corrector_kwargs) + + return e_t + + alphas = self.model.alphas_cumprod if use_original_steps else self.ddim_alphas + alphas_prev = self.model.alphas_cumprod_prev if use_original_steps else self.ddim_alphas_prev + sqrt_one_minus_alphas = self.model.sqrt_one_minus_alphas_cumprod if use_original_steps else self.ddim_sqrt_one_minus_alphas + sigmas = self.model.ddim_sigmas_for_original_num_steps if use_original_steps else self.ddim_sigmas + + def get_x_prev_and_pred_x0(e_t, index): + # select parameters corresponding to the currently considered timestep + a_t = torch.full((b, 1, 1, 1), alphas[index], device=device) + a_prev = torch.full((b, 1, 1, 1), alphas_prev[index], device=device) + sigma_t = torch.full((b, 1, 1, 1), sigmas[index], device=device) + sqrt_one_minus_at = torch.full((b, 1, 1, 1), sqrt_one_minus_alphas[index],device=device) + + # current prediction for x_0 + pred_x0 = (x - sqrt_one_minus_at * e_t) / a_t.sqrt() + if quantize_denoised: + pred_x0, _, *_ = self.model.first_stage_model.quantize(pred_x0) + # direction pointing to x_t + dir_xt = (1. - a_prev - sigma_t**2).sqrt() * e_t + noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature + if noise_dropout > 0.: + noise = torch.nn.functional.dropout(noise, p=noise_dropout) + x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise + return x_prev, pred_x0 + + e_t = get_model_output(x, t) + if len(old_eps) == 0: + # Pseudo Improved Euler (2nd order) + x_prev, pred_x0 = get_x_prev_and_pred_x0(e_t, index) + e_t_next = get_model_output(x_prev, t_next) + e_t_prime = (e_t + e_t_next) / 2 + elif len(old_eps) == 1: + # 2nd order Pseudo Linear Multistep (Adams-Bashforth) + e_t_prime = (3 * e_t - old_eps[-1]) / 2 + elif len(old_eps) == 2: + # 3nd order Pseudo Linear Multistep (Adams-Bashforth) + e_t_prime = (23 * e_t - 16 * old_eps[-1] + 5 * old_eps[-2]) / 12 + elif len(old_eps) >= 3: + # 4nd order Pseudo Linear Multistep (Adams-Bashforth) + e_t_prime = (55 * e_t - 59 * old_eps[-1] + 37 * old_eps[-2] - 9 * old_eps[-3]) / 24 + + x_prev, pred_x0 = get_x_prev_and_pred_x0(e_t_prime, index) + + return x_prev, pred_x0, e_t + # ================================================================================================= # Monkey patch LatentInpaintDiffusion to load the checkpoint with a proper config. # Adapted from: @@ -175,5 +320,9 @@ def should_hijack_inpainting(checkpoint_info): def do_inpainting_hijack(): ldm.models.diffusion.ddpm.get_unconditional_conditioning = get_unconditional_conditioning ldm.models.diffusion.ddpm.LatentInpaintDiffusion = LatentInpaintDiffusion + ldm.models.diffusion.ddim.DDIMSampler.p_sample_ddim = p_sample_ddim - ldm.models.diffusion.ddim.DDIMSampler.sample = sample \ No newline at end of file + ldm.models.diffusion.ddim.DDIMSampler.sample = sample_ddim + + ldm.models.diffusion.plms.PLMSSampler.p_sample_plms = p_sample_plms + ldm.models.diffusion.plms.PLMSSampler.sample = sample_plms \ No newline at end of file diff --git a/modules/sd_models.py b/modules/sd_models.py index 47836d25..7072db08 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -214,8 +214,6 @@ def load_model(): sd_config = OmegaConf.load(checkpoint_info.config) if should_hijack_inpainting(checkpoint_info): - do_inpainting_hijack() - # Hardcoded config for now... sd_config.model.target = "ldm.models.diffusion.ddpm.LatentInpaintDiffusion" sd_config.model.params.use_ema = False @@ -225,6 +223,7 @@ def load_model(): # Create a "fake" config with a different name so that we know to unload it when switching models. checkpoint_info = checkpoint_info._replace(config=checkpoint_info.config.replace(".yaml", "-inpainting.yaml")) + do_inpainting_hijack() sd_model = instantiate_from_config(sd_config.model) load_model_weights(sd_model, checkpoint_info) -- cgit v1.2.3 From d23a46ceaa76af2847f11172f32c92665c268b1b Mon Sep 17 00:00:00 2001 From: Vladimir Repin <32306715+mezotaken@users.noreply.github.com> Date: Thu, 20 Oct 2022 23:49:14 +0300 Subject: Different approach to skip/interrupt with highres fix --- modules/processing.py | 4 +++- modules/sd_samplers.py | 4 ++++ 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index 6324ca91..bcb0c32c 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -587,7 +587,9 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): x = None devices.torch_gc() - return self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps) or samples + samples = self.sampler.sample_img2img(self, samples, noise, conditioning, unconditional_conditioning, steps=self.steps) + + return samples class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index b58e810b..7ff77c01 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -196,6 +196,7 @@ class VanillaStableDiffusionSampler: x1 = self.sampler.stochastic_encode(x, torch.tensor([t_enc] * int(x.shape[0])).to(shared.device), noise=noise) self.init_latent = x + self.last_latent = x self.step = 0 samples = self.launch_sampling(steps, lambda: self.sampler.decode(x1, conditioning, t_enc, unconditional_guidance_scale=p.cfg_scale, unconditional_conditioning=unconditional_conditioning)) @@ -206,6 +207,7 @@ class VanillaStableDiffusionSampler: self.initialize(p) self.init_latent = None + self.last_latent = x self.step = 0 steps = steps or p.steps @@ -388,6 +390,7 @@ class KDiffusionSampler: extra_params_kwargs['sigmas'] = sigma_sched self.model_wrap_cfg.init_latent = x + self.last_latent = x samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, xi, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs)) @@ -414,6 +417,7 @@ class KDiffusionSampler: else: extra_params_kwargs['sigmas'] = sigmas + self.last_latent = x samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args={'cond': conditioning, 'uncond': unconditional_conditioning, 'cond_scale': p.cfg_scale}, disable=False, callback=self.callback_state, **extra_params_kwargs)) return samples -- cgit v1.2.3 From 49533eed9e3aad19e9868ee140708baec4fd44be Mon Sep 17 00:00:00 2001 From: random_thoughtss Date: Thu, 20 Oct 2022 16:01:27 -0700 Subject: XY grid correctly re-assignes model when config changes --- modules/sd_models.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/sd_models.py b/modules/sd_models.py index 7072db08..fea84630 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -204,9 +204,9 @@ def load_model_weights(model, checkpoint_info): model.sd_checkpoint_info = checkpoint_info -def load_model(): +def load_model(checkpoint_info=None): from modules import lowvram, sd_hijack - checkpoint_info = select_checkpoint() + checkpoint_info = checkpoint_info or select_checkpoint() if checkpoint_info.config != shared.cmd_opts.config: print(f"Loading config from: {checkpoint_info.config}") @@ -249,7 +249,7 @@ def reload_model_weights(sd_model, info=None): if sd_model.sd_checkpoint_info.config != checkpoint_info.config or should_hijack_inpainting(checkpoint_info) != should_hijack_inpainting(sd_model.sd_checkpoint_info): checkpoints_loaded.clear() - shared.sd_model = load_model() + shared.sd_model = load_model(checkpoint_info) return shared.sd_model if shared.cmd_opts.lowvram or shared.cmd_opts.medvram: -- cgit v1.2.3 From 45872181902ada06267e2de601586d512cf5df1a Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 09:00:39 +0300 Subject: updated readme and some small stylistic changes to code --- modules/processing.py | 14 ++++++-------- modules/sd_hijack_inpainting.py | 3 +++ 2 files changed, 9 insertions(+), 8 deletions(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index 539cde38..21786968 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -540,11 +540,10 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): self.truncate_x = int(self.firstphase_width - firstphase_width_truncated) // opt_f self.truncate_y = int(self.firstphase_height - firstphase_height_truncated) // opt_f - - def create_dummy_mask(self, x, first_phase: bool = False): + def create_dummy_mask(self, x, width=None, height=None): if self.sampler.conditioning_key in {'hybrid', 'concat'}: - height = self.firstphase_height if first_phase else self.height - width = self.firstphase_width if first_phase else self.width + height = height or self.height + width = width or self.width # The "masked-image" in this case will just be all zeros since the entire image is masked. image_conditioning = torch.zeros(x.shape[0], 3, height, width, device=x.device) @@ -571,7 +570,7 @@ class StableDiffusionProcessingTxt2Img(StableDiffusionProcessing): return samples x = create_random_tensors([opt_C, self.firstphase_height // opt_f, self.firstphase_width // opt_f], seeds=seeds, subseeds=subseeds, subseed_strength=self.subseed_strength, seed_resize_from_h=self.seed_resize_from_h, seed_resize_from_w=self.seed_resize_from_w, p=self) - samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.create_dummy_mask(x, first_phase=True)) + samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.create_dummy_mask(x, self.firstphase_width, self.firstphase_height)) samples = samples[:, :, self.truncate_y//2:samples.shape[2]-self.truncate_y//2, self.truncate_x//2:samples.shape[3]-self.truncate_x//2] @@ -634,6 +633,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): self.inpainting_mask_invert = inpainting_mask_invert self.mask = None self.nmask = None + self.image_conditioning = None def init(self, all_prompts, all_seeds, all_subseeds): self.sampler = sd_samplers.create_sampler_with_index(sd_samplers.samplers_for_img2img, self.sampler_index, self.sd_model) @@ -735,9 +735,7 @@ class StableDiffusionProcessingImg2Img(StableDiffusionProcessing): elif self.inpainting_fill == 3: self.init_latent = self.init_latent * self.mask - conditioning_key = self.sampler.conditioning_key - - if conditioning_key in {'hybrid', 'concat'}: + if self.sampler.conditioning_key in {'hybrid', 'concat'}: if self.image_mask is not None: conditioning_mask = np.array(self.image_mask.convert("L")) conditioning_mask = conditioning_mask.astype(np.float32) / 255.0 diff --git a/modules/sd_hijack_inpainting.py b/modules/sd_hijack_inpainting.py index 43938071..fd92a335 100644 --- a/modules/sd_hijack_inpainting.py +++ b/modules/sd_hijack_inpainting.py @@ -301,6 +301,7 @@ def get_unconditional_conditioning(self, batch_size, null_label=None): c = repeat(c, "1 ... -> b ...", b=batch_size).to(self.device) return c + class LatentInpaintDiffusion(LatentDiffusion): def __init__( self, @@ -314,9 +315,11 @@ class LatentInpaintDiffusion(LatentDiffusion): assert self.masked_image_key in concat_keys self.concat_keys = concat_keys + def should_hijack_inpainting(checkpoint_info): return str(checkpoint_info.filename).endswith("inpainting.ckpt") and not checkpoint_info.config.endswith("inpainting.yaml") + def do_inpainting_hijack(): ldm.models.diffusion.ddpm.get_unconditional_conditioning = get_unconditional_conditioning ldm.models.diffusion.ddpm.LatentInpaintDiffusion = LatentInpaintDiffusion -- cgit v1.2.3 From 74088c2a06a975092806362aede22f82716cb011 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Thu, 20 Oct 2022 08:18:02 +0300 Subject: allow float sizes for hypernet's layer_structure --- modules/hypernetworks/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 08f75f15..e0741d08 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -15,7 +15,7 @@ def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm assert not os.path.exists(fn), f"file {fn} already exists" if type(layer_structure) == str: - layer_structure = tuple(map(int, re.sub(r'\D', '', layer_structure))) + layer_structure = [float(x.strip()) for x in layer_structure.split(",")] hypernet = modules.hypernetworks.hypernetwork.Hypernetwork( name=name, -- cgit v1.2.3 From 60872c5b404114336f9ca0c671ba88fa4a8201c9 Mon Sep 17 00:00:00 2001 From: winterspringsummer Date: Thu, 20 Oct 2022 19:10:32 +0900 Subject: Fixed path issue while extras batch processing --- modules/extras.py | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/extras.py b/modules/extras.py index b853fa5b..f9796624 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -118,10 +118,14 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ while len(cached_images) > 2: del cached_images[next(iter(cached_images.keys()))] - - images.save_image(image, path=outpath, basename="", seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, - no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, - forced_filename=image_name if opts.use_original_name_batch else None) + + if opts.use_original_name_batch and image_name != None: + basename = os.path.splitext(os.path.basename(image_name))[0] + else: + basename = '' + + images.save_image(image, path=outpath, basename=basename, seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, + no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=None) if opts.enable_pnginfo: image.info = existing_pnginfo -- cgit v1.2.3 From fb5a8cf0d9ed027ea3aa2e5422c946d8e6e72efe Mon Sep 17 00:00:00 2001 From: winterspringsummer Date: Thu, 20 Oct 2022 21:31:29 +0900 Subject: Added try except to extras batch from directory --- modules/extras.py | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'modules') diff --git a/modules/extras.py b/modules/extras.py index f9796624..0d817cf9 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -41,7 +41,10 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ return outputs, "Please select an input directory.", '' image_list = [file for file in [os.path.join(input_dir, x) for x in os.listdir(input_dir)] if os.path.isfile(file)] for img in image_list: - image = Image.open(img) + try: + image = Image.open(img) + except Exception: + continue imageArr.append(image) imageNameArr.append(img) else: @@ -122,10 +125,10 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ if opts.use_original_name_batch and image_name != None: basename = os.path.splitext(os.path.basename(image_name))[0] else: - basename = '' + basename = None - images.save_image(image, path=outpath, basename=basename, seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, - no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=None) + images.save_image(image, path=outpath, basename='', seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, + no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=basename) if opts.enable_pnginfo: image.info = existing_pnginfo -- cgit v1.2.3 From a13c3bed3cec27afe3c015d3d62db36e25b10d1f Mon Sep 17 00:00:00 2001 From: winterspringsummer Date: Thu, 20 Oct 2022 21:43:27 +0900 Subject: Fixed path issue while extras batch processing --- modules/extras.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/extras.py b/modules/extras.py index 0d817cf9..ac85142c 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -125,10 +125,10 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ if opts.use_original_name_batch and image_name != None: basename = os.path.splitext(os.path.basename(image_name))[0] else: - basename = None + basename = '' - images.save_image(image, path=outpath, basename='', seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, - no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=basename) + images.save_image(image, path=outpath, basename=basename, seed=None, prompt=None, extension=opts.samples_format, info=info, short_filename=True, + no_prompt=True, grid=False, pnginfo_section_name="extras", existing_info=existing_pnginfo, forced_filename=None) if opts.enable_pnginfo: image.info = existing_pnginfo -- cgit v1.2.3 From 9d71eef02e7395e179b8d5e61e6d91ddd8928d2e Mon Sep 17 00:00:00 2001 From: winterspringsummer Date: Fri, 21 Oct 2022 09:23:13 +0900 Subject: sort file list in alphabetical ordering in extras --- modules/extras.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/extras.py b/modules/extras.py index ac85142c..22c5a1c1 100644 --- a/modules/extras.py +++ b/modules/extras.py @@ -39,7 +39,7 @@ def run_extras(extras_mode, resize_mode, image, image_folder, input_dir, output_ if input_dir == '': return outputs, "Please select an input directory.", '' - image_list = [file for file in [os.path.join(input_dir, x) for x in os.listdir(input_dir)] if os.path.isfile(file)] + image_list = [file for file in [os.path.join(input_dir, x) for x in sorted(os.listdir(input_dir))] if os.path.isfile(file)] for img in image_list: try: image = Image.open(img) -- cgit v1.2.3 From c23f666dba2b484d521d2dc4be91cf9e09312647 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 09:47:43 +0300 Subject: a more strict check for activation type and a more reasonable check for type of layer in hypernets --- modules/hypernetworks/hypernetwork.py | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 7d617680..84e7e350 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -32,10 +32,16 @@ class HypernetworkModule(torch.nn.Module): linears = [] for i in range(len(layer_structure) - 1): linears.append(torch.nn.Linear(int(dim * layer_structure[i]), int(dim * layer_structure[i+1]))) + if activation_func == "relu": linears.append(torch.nn.ReLU()) - if activation_func == "leakyrelu": + elif activation_func == "leakyrelu": linears.append(torch.nn.LeakyReLU()) + elif activation_func == 'linear' or activation_func is None: + pass + else: + raise RuntimeError(f'hypernetwork uses an unsupported activation function: {activation_func}') + if add_layer_norm: linears.append(torch.nn.LayerNorm(int(dim * layer_structure[i+1]))) @@ -46,7 +52,7 @@ class HypernetworkModule(torch.nn.Module): self.load_state_dict(state_dict) else: for layer in self.linear: - if not "ReLU" in layer.__str__(): + if type(layer) == torch.nn.Linear: layer.weight.data.normal_(mean=0.0, std=0.01) layer.bias.data.zero_() @@ -74,7 +80,7 @@ class HypernetworkModule(torch.nn.Module): def trainables(self): layer_structure = [] for layer in self.linear: - if not "ReLU" in layer.__str__(): + if type(layer) == torch.nn.Linear: layer_structure += [layer.weight, layer.bias] return layer_structure -- cgit v1.2.3 From 7157e5d064741fa57ca81a2c6432a651f21ee82f Mon Sep 17 00:00:00 2001 From: Patryk Wychowaniec Date: Thu, 20 Oct 2022 19:22:59 +0200 Subject: interrogate: Fix CLIP-interrogation on CPU Currently, trying to perform CLIP interrogation on a CPU fails, saying: ``` RuntimeError: "slow_conv2d_cpu" not implemented for 'Half' ``` This merge request fixes this issue by detecting whether the target device is CPU and, if so, force-enabling `--no-half` and passing `device="cpu"` to `clip.load()` (which then does some extra tricks to ensure it works correctly on CPU). --- modules/interrogate.py | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'modules') diff --git a/modules/interrogate.py b/modules/interrogate.py index 64b91eb4..65b05d34 100644 --- a/modules/interrogate.py +++ b/modules/interrogate.py @@ -28,9 +28,11 @@ class InterrogateModels: clip_preprocess = None categories = None dtype = None + running_on_cpu = None def __init__(self, content_dir): self.categories = [] + self.running_on_cpu = devices.device_interrogate == torch.device("cpu") if os.path.exists(content_dir): for filename in os.listdir(content_dir): @@ -53,7 +55,11 @@ class InterrogateModels: def load_clip_model(self): import clip - model, preprocess = clip.load(clip_model_name) + if self.running_on_cpu: + model, preprocess = clip.load(clip_model_name, device="cpu") + else: + model, preprocess = clip.load(clip_model_name) + model.eval() model = model.to(devices.device_interrogate) @@ -62,14 +68,14 @@ class InterrogateModels: def load(self): if self.blip_model is None: self.blip_model = self.load_blip_model() - if not shared.cmd_opts.no_half: + if not shared.cmd_opts.no_half and not self.running_on_cpu: self.blip_model = self.blip_model.half() self.blip_model = self.blip_model.to(devices.device_interrogate) if self.clip_model is None: self.clip_model, self.clip_preprocess = self.load_clip_model() - if not shared.cmd_opts.no_half: + if not shared.cmd_opts.no_half and not self.running_on_cpu: self.clip_model = self.clip_model.half() self.clip_model = self.clip_model.to(devices.device_interrogate) -- cgit v1.2.3 From b69c37d25e4ffc56e8f8c247fa2c38b4648cefb7 Mon Sep 17 00:00:00 2001 From: guaneec Date: Thu, 20 Oct 2022 22:21:12 +0800 Subject: Allow datasets with only 1 image in TI --- modules/textual_inversion/dataset.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py index 23bb4b6a..5b1c5002 100644 --- a/modules/textual_inversion/dataset.py +++ b/modules/textual_inversion/dataset.py @@ -83,7 +83,7 @@ class PersonalizedBase(Dataset): self.dataset.append(entry) - assert len(self.dataset) > 1, "No images have been found in the dataset." + assert len(self.dataset) > 0, "No images have been found in the dataset." self.length = len(self.dataset) * repeats // batch_size self.initial_indexes = np.arange(len(self.dataset)) @@ -91,7 +91,7 @@ class PersonalizedBase(Dataset): self.shuffle() def shuffle(self): - self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])] + self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0]).numpy()] def create_text(self, filename_text): text = random.choice(self.lines) -- cgit v1.2.3 From 5245c7a4935f67b677da0f5a1fc2b74c074aa0e2 Mon Sep 17 00:00:00 2001 From: timntorres Date: Wed, 19 Oct 2022 12:21:32 -0700 Subject: Issue #2921-Give PNG info to Hypernet previews. --- modules/hypernetworks/hypernetwork.py | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 84e7e350..68c8f26d 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -256,6 +256,9 @@ def stack_conds(conds): def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height): + # images is required here to give training previews their infotext. Importing this at the very top causes a circular dependency. + from modules import images + assert hypernetwork_name, 'hypernetwork not selected' path = shared.hypernetworks.get(hypernetwork_name, None) @@ -298,6 +301,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log last_saved_file = "" last_saved_image = "" + forced_filename = "" ititial_step = hypernetwork.step or 0 if ititial_step > steps: @@ -345,7 +349,8 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log }) if hypernetwork.step > 0 and images_dir is not None and hypernetwork.step % create_image_every == 0: - last_saved_image = os.path.join(images_dir, f'{hypernetwork_name}-{hypernetwork.step}.png') + forced_filename = f'{hypernetwork_name}-{hypernetwork.step}' + last_saved_image = os.path.join(images_dir, forced_filename) optimizer.zero_grad() shared.sd_model.cond_stage_model.to(devices.device) @@ -381,7 +386,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log if image is not None: shared.state.current_image = image - image.save(last_saved_image) + last_saved_image, last_text_info = images.save_image(image, images_dir, "", p.seed, p.prompt, shared.opts.samples_format, processed.infotexts[0], p=p, forced_filename=forced_filename) last_saved_image += f", prompt: {preview_text}" shared.state.job_no = hypernetwork.step -- cgit v1.2.3 From 6014fb8afbe05c8d02fffe7a36a2e48128713bd2 Mon Sep 17 00:00:00 2001 From: timntorres Date: Wed, 19 Oct 2022 12:22:23 -0700 Subject: Do nothing if image file already exists. --- modules/images.py | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/images.py b/modules/images.py index b9589563..550e53ae 100644 --- a/modules/images.py +++ b/modules/images.py @@ -416,7 +416,11 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i dirname = apply_filename_pattern(opts.directories_filename_pattern or "[prompt_words]", p, seed, prompt).strip('\\ /') path = os.path.join(path, dirname) - os.makedirs(path, exist_ok=True) + try: + os.makedirs(path, exist_ok=True) + except FileExistsError: + # If the file already exists, continue and allow said file to be overwritten. + pass if forced_filename is None: basecount = get_next_sequence_number(path, basename) -- cgit v1.2.3 From 4ff274e1e35bb642687253ce744d2cfa738ab293 Mon Sep 17 00:00:00 2001 From: timntorres Date: Wed, 19 Oct 2022 12:32:22 -0700 Subject: Revise comments. --- modules/hypernetworks/hypernetwork.py | 2 +- modules/images.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 68c8f26d..3f96361c 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -256,7 +256,7 @@ def stack_conds(conds): def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log_directory, training_width, training_height, steps, create_image_every, save_hypernetwork_every, template_file, preview_from_txt2img, preview_prompt, preview_negative_prompt, preview_steps, preview_sampler_index, preview_cfg_scale, preview_seed, preview_width, preview_height): - # images is required here to give training previews their infotext. Importing this at the very top causes a circular dependency. + # images allows training previews to have infotext. Importing it at the top causes a circular import problem. from modules import images assert hypernetwork_name, 'hypernetwork not selected' diff --git a/modules/images.py b/modules/images.py index 550e53ae..b8834e3c 100644 --- a/modules/images.py +++ b/modules/images.py @@ -419,7 +419,7 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i try: os.makedirs(path, exist_ok=True) except FileExistsError: - # If the file already exists, continue and allow said file to be overwritten. + # If the file already exists, allow said file to be overwritten. pass if forced_filename is None: -- cgit v1.2.3 From 2273e752fb3e578f1047f6d38b96330b07bf61a9 Mon Sep 17 00:00:00 2001 From: timntorres Date: Wed, 19 Oct 2022 14:23:48 -0700 Subject: Remove redundant try/except. --- modules/images.py | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'modules') diff --git a/modules/images.py b/modules/images.py index b8834e3c..b9589563 100644 --- a/modules/images.py +++ b/modules/images.py @@ -416,11 +416,7 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i dirname = apply_filename_pattern(opts.directories_filename_pattern or "[prompt_words]", p, seed, prompt).strip('\\ /') path = os.path.join(path, dirname) - try: - os.makedirs(path, exist_ok=True) - except FileExistsError: - # If the file already exists, allow said file to be overwritten. - pass + os.makedirs(path, exist_ok=True) if forced_filename is None: basecount = get_next_sequence_number(path, basename) -- cgit v1.2.3 From 03a1e288c4973dd2dff57a97469b40f146b6fccf Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 10:13:24 +0300 Subject: turns out LayerNorm also has weight and bias and needs to be pre-multiplied and trained for hypernets --- modules/hypernetworks/hypernetwork.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 3274a802..b1a5d0c7 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -52,7 +52,7 @@ class HypernetworkModule(torch.nn.Module): self.load_state_dict(state_dict) else: for layer in self.linear: - if type(layer) == torch.nn.Linear: + if type(layer) == torch.nn.Linear or type(layer) == torch.nn.LayerNorm: layer.weight.data.normal_(mean=0.0, std=0.01) layer.bias.data.zero_() @@ -80,7 +80,7 @@ class HypernetworkModule(torch.nn.Module): def trainables(self): layer_structure = [] for layer in self.linear: - if type(layer) == torch.nn.Linear: + if type(layer) == torch.nn.Linear or type(layer) == torch.nn.LayerNorm: layer_structure += [layer.weight, layer.bias] return layer_structure -- cgit v1.2.3 From bf30673f5132c8f28357b31224c54331e788d3e7 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 10:19:25 +0300 Subject: Fix Hypernet infotext string split bug for PR #3283 --- modules/processing.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/processing.py b/modules/processing.py index 21786968..d1deffa9 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -304,7 +304,7 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration "Size": f"{p.width}x{p.height}", "Model hash": getattr(p, 'sd_model_hash', None if not opts.add_model_hash_to_info or not shared.sd_model.sd_model_hash else shared.sd_model.sd_model_hash), "Model": (None if not opts.add_model_name_to_info or not shared.sd_model.sd_checkpoint_info.model_name else shared.sd_model.sd_checkpoint_info.model_name.replace(',', '').replace(':', '')), - "Hypernet": (None if shared.loaded_hypernetwork is None else shared.loaded_hypernetwork.filename.split('\\')[-1].split('.')[0]), + "Hypernet": (None if shared.loaded_hypernetwork is None else os.path.splitext(os.path.basename(shared.loaded_hypernetwork.filename))[0]), "Batch size": (None if p.batch_size < 2 else p.batch_size), "Batch pos": (None if p.batch_size < 2 else position_in_batch), "Variation seed": (None if p.subseed_strength == 0 else all_subseeds[index]), -- cgit v1.2.3 From df5706409386cc2e88718bd9101045587c39f8bb Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 16:10:51 +0300 Subject: do not load aesthetic clip model until it's needed add refresh button for aesthetic embeddings add aesthetic params to images' infotext --- modules/aesthetic_clip.py | 40 +++++++++++++++++++---- modules/generation_parameters_copypaste.py | 18 +++++++++-- modules/img2img.py | 5 +-- modules/processing.py | 4 +-- modules/sd_models.py | 3 -- modules/txt2img.py | 4 +-- modules/ui.py | 52 ++++++++++++++++++++---------- 7 files changed, 88 insertions(+), 38 deletions(-) (limited to 'modules') diff --git a/modules/aesthetic_clip.py b/modules/aesthetic_clip.py index 34efa931..8c828541 100644 --- a/modules/aesthetic_clip.py +++ b/modules/aesthetic_clip.py @@ -40,6 +40,8 @@ def iter_to_batched(iterable, n=1): def create_ui(): + import modules.ui + with gr.Group(): with gr.Accordion("Open for Clip Aesthetic!", open=False): with gr.Row(): @@ -55,6 +57,8 @@ def create_ui(): label="Aesthetic imgs embedding", value="None") + modules.ui.create_refresh_button(aesthetic_imgs, shared.update_aesthetic_embeddings, lambda: {"choices": sorted(shared.aesthetic_embeddings.keys())}, "refresh_aesthetic_embeddings") + with gr.Row(): aesthetic_imgs_text = gr.Textbox(label='Aesthetic text for imgs', placeholder="This text is used to rotate the feature space of the imgs embs", @@ -66,11 +70,21 @@ def create_ui(): return aesthetic_weight, aesthetic_steps, aesthetic_lr, aesthetic_slerp, aesthetic_imgs, aesthetic_imgs_text, aesthetic_slerp_angle, aesthetic_text_negative +aesthetic_clip_model = None + + +def aesthetic_clip(): + global aesthetic_clip_model + + if aesthetic_clip_model is None or aesthetic_clip_model.name_or_path != shared.sd_model.cond_stage_model.wrapped.transformer.name_or_path: + aesthetic_clip_model = CLIPModel.from_pretrained(shared.sd_model.cond_stage_model.wrapped.transformer.name_or_path) + aesthetic_clip_model.cpu() + + return aesthetic_clip_model + + def generate_imgs_embd(name, folder, batch_size): - # clipModel = CLIPModel.from_pretrained( - # shared.sd_model.cond_stage_model.clipModel.name_or_path - # ) - model = shared.clip_model.to(device) + model = aesthetic_clip().to(device) processor = CLIPProcessor.from_pretrained(model.name_or_path) with torch.no_grad(): @@ -91,7 +105,7 @@ def generate_imgs_embd(name, folder, batch_size): path = str(Path(shared.cmd_opts.aesthetic_embeddings_dir) / f"{name}.pt") torch.save(embs, path) - model = model.cpu() + model.cpu() del processor del embs gc.collect() @@ -132,7 +146,7 @@ class AestheticCLIP: self.image_embs = None self.load_image_embs(None) - def set_aesthetic_params(self, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, image_embs_name=None, + def set_aesthetic_params(self, p, aesthetic_lr=0, aesthetic_weight=0, aesthetic_steps=0, image_embs_name=None, aesthetic_slerp=True, aesthetic_imgs_text="", aesthetic_slerp_angle=0.15, aesthetic_text_negative=False): @@ -145,6 +159,18 @@ class AestheticCLIP: self.aesthetic_steps = aesthetic_steps self.load_image_embs(image_embs_name) + if self.image_embs_name is not None: + p.extra_generation_params.update({ + "Aesthetic LR": aesthetic_lr, + "Aesthetic weight": aesthetic_weight, + "Aesthetic steps": aesthetic_steps, + "Aesthetic embedding": self.image_embs_name, + "Aesthetic slerp": aesthetic_slerp, + "Aesthetic text": aesthetic_imgs_text, + "Aesthetic text negative": aesthetic_text_negative, + "Aesthetic slerp angle": aesthetic_slerp_angle, + }) + def set_skip(self, skip): self.skip = skip @@ -168,7 +194,7 @@ class AestheticCLIP: tokens = torch.asarray(remade_batch_tokens).to(device) - model = copy.deepcopy(shared.clip_model).to(device) + model = copy.deepcopy(aesthetic_clip()).to(device) model.requires_grad_(True) if self.aesthetic_imgs_text is not None and len(self.aesthetic_imgs_text) > 0: text_embs_2 = model.get_text_features( diff --git a/modules/generation_parameters_copypaste.py b/modules/generation_parameters_copypaste.py index 0f041449..f73647da 100644 --- a/modules/generation_parameters_copypaste.py +++ b/modules/generation_parameters_copypaste.py @@ -4,13 +4,22 @@ import gradio as gr from modules.shared import script_path from modules import shared -re_param_code = r"\s*([\w ]+):\s*([^,]+)(?:,|$)" +re_param_code = r'\s*([\w ]+):\s*("(?:\\|\"|[^\"])+"|[^,]*)(?:,|$)' re_param = re.compile(re_param_code) re_params = re.compile(r"^(?:" + re_param_code + "){3,}$") re_imagesize = re.compile(r"^(\d+)x(\d+)$") type_of_gr_update = type(gr.update()) +def quote(text): + if ',' not in str(text): + return text + + text = str(text) + text = text.replace('\\', '\\\\') + text = text.replace('"', '\\"') + return f'"{text}"' + def parse_generation_parameters(x: str): """parses generation parameters string, the one you see in text field under the picture in UI: ``` @@ -83,7 +92,12 @@ def connect_paste(button, paste_fields, input_comp, js=None): else: try: valtype = type(output.value) - val = valtype(v) + + if valtype == bool and v == "False": + val = False + else: + val = valtype(v) + res.append(gr.update(value=val)) except Exception: res.append(gr.update()) diff --git a/modules/img2img.py b/modules/img2img.py index bc7c66bc..eea5199b 100644 --- a/modules/img2img.py +++ b/modules/img2img.py @@ -109,10 +109,7 @@ def img2img(mode: int, prompt: str, negative_prompt: str, prompt_style: str, pro inpainting_mask_invert=inpainting_mask_invert, ) - shared.aesthetic_clip.set_aesthetic_params(float(aesthetic_lr), float(aesthetic_weight), int(aesthetic_steps), - aesthetic_imgs, aesthetic_slerp, aesthetic_imgs_text, - aesthetic_slerp_angle, - aesthetic_text_negative) + shared.aesthetic_clip.set_aesthetic_params(p, float(aesthetic_lr), float(aesthetic_weight), int(aesthetic_steps), aesthetic_imgs, aesthetic_slerp, aesthetic_imgs_text, aesthetic_slerp_angle, aesthetic_text_negative) if shared.cmd_opts.enable_console_prompts: print(f"\nimg2img: {prompt}", file=shared.progress_print_out) diff --git a/modules/processing.py b/modules/processing.py index d1deffa9..f0852cd5 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -12,7 +12,7 @@ from skimage import exposure from typing import Any, Dict, List, Optional import modules.sd_hijack -from modules import devices, prompt_parser, masking, sd_samplers, lowvram +from modules import devices, prompt_parser, masking, sd_samplers, lowvram, generation_parameters_copypaste from modules.sd_hijack import model_hijack from modules.shared import opts, cmd_opts, state import modules.shared as shared @@ -318,7 +318,7 @@ def create_infotext(p, all_prompts, all_seeds, all_subseeds, comments, iteration generation_params.update(p.extra_generation_params) - generation_params_text = ", ".join([k if k == v else f'{k}: {v}' for k, v in generation_params.items() if v is not 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]) negative_prompt_text = "\nNegative prompt: " + p.negative_prompt if p.negative_prompt else "" diff --git a/modules/sd_models.py b/modules/sd_models.py index 05a1df28..b1c91b0d 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -234,9 +234,6 @@ def load_model(checkpoint_info=None): sd_hijack.model_hijack.hijack(sd_model) - if shared.clip_model is None or shared.clip_model.transformer.name_or_path != sd_model.cond_stage_model.wrapped.transformer.name_or_path: - shared.clip_model = CLIPModel.from_pretrained(sd_model.cond_stage_model.wrapped.transformer.name_or_path) - sd_model.eval() print(f"Model loaded.") diff --git a/modules/txt2img.py b/modules/txt2img.py index 32ed1d8d..1761cfa2 100644 --- a/modules/txt2img.py +++ b/modules/txt2img.py @@ -36,9 +36,7 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: firstphase_height=firstphase_height if enable_hr else None, ) - shared.aesthetic_clip.set_aesthetic_params(float(aesthetic_lr), float(aesthetic_weight), int(aesthetic_steps), - aesthetic_imgs, aesthetic_slerp, aesthetic_imgs_text, aesthetic_slerp_angle, - aesthetic_text_negative) + shared.aesthetic_clip.set_aesthetic_params(p, float(aesthetic_lr), float(aesthetic_weight), int(aesthetic_steps), aesthetic_imgs, aesthetic_slerp, aesthetic_imgs_text, aesthetic_slerp_angle, aesthetic_text_negative) if cmd_opts.enable_console_prompts: print(f"\ntxt2img: {prompt}", file=shared.progress_print_out) diff --git a/modules/ui.py b/modules/ui.py index 381ca925..0d020de6 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -597,27 +597,29 @@ def apply_setting(key, value): return value -def create_ui(wrap_gradio_gpu_call): - import modules.img2img - import modules.txt2img +def create_refresh_button(refresh_component, refresh_method, refreshed_args, elem_id): + def refresh(): + refresh_method() + args = refreshed_args() if callable(refreshed_args) else refreshed_args - def create_refresh_button(refresh_component, refresh_method, refreshed_args, elem_id): - def refresh(): - refresh_method() - args = refreshed_args() if callable(refreshed_args) else refreshed_args + for k, v in args.items(): + setattr(refresh_component, k, v) - for k, v in args.items(): - setattr(refresh_component, k, v) + return gr.update(**(args or {})) - return gr.update(**(args or {})) + refresh_button = gr.Button(value=refresh_symbol, elem_id=elem_id) + refresh_button.click( + fn=refresh, + inputs=[], + outputs=[refresh_component] + ) + return refresh_button + + +def create_ui(wrap_gradio_gpu_call): + import modules.img2img + import modules.txt2img - refresh_button = gr.Button(value=refresh_symbol, elem_id=elem_id) - refresh_button.click( - fn = refresh, - inputs = [], - outputs = [refresh_component] - ) - return refresh_button with gr.Blocks(analytics_enabled=False) as txt2img_interface: txt2img_prompt, roll, txt2img_prompt_style, txt2img_negative_prompt, txt2img_prompt_style2, submit, _, _, txt2img_prompt_style_apply, txt2img_save_style, txt2img_paste, token_counter, token_button = create_toprow(is_img2img=False) @@ -802,6 +804,14 @@ def create_ui(wrap_gradio_gpu_call): (hr_options, lambda d: gr.Row.update(visible="Denoising strength" in d)), (firstphase_width, "First pass size-1"), (firstphase_height, "First pass size-2"), + (aesthetic_lr, "Aesthetic LR"), + (aesthetic_weight, "Aesthetic weight"), + (aesthetic_steps, "Aesthetic steps"), + (aesthetic_imgs, "Aesthetic embedding"), + (aesthetic_slerp, "Aesthetic slerp"), + (aesthetic_imgs_text, "Aesthetic text"), + (aesthetic_text_negative, "Aesthetic text negative"), + (aesthetic_slerp_angle, "Aesthetic slerp angle"), ] txt2img_preview_params = [ @@ -1077,6 +1087,14 @@ def create_ui(wrap_gradio_gpu_call): (seed_resize_from_w, "Seed resize from-1"), (seed_resize_from_h, "Seed resize from-2"), (denoising_strength, "Denoising strength"), + (aesthetic_lr_im, "Aesthetic LR"), + (aesthetic_weight_im, "Aesthetic weight"), + (aesthetic_steps_im, "Aesthetic steps"), + (aesthetic_imgs_im, "Aesthetic embedding"), + (aesthetic_slerp_im, "Aesthetic slerp"), + (aesthetic_imgs_text_im, "Aesthetic text"), + (aesthetic_text_negative_im, "Aesthetic text negative"), + (aesthetic_slerp_angle_im, "Aesthetic slerp angle"), ] token_button.click(fn=update_token_counter, inputs=[img2img_prompt, steps], outputs=[token_counter]) -- cgit v1.2.3 From 9286fe53de2eef91f13cc3ad5938ddf67ecc8413 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Fri, 21 Oct 2022 16:38:06 +0300 Subject: make aestetic embedding ciompatible with prompts longer than 75 tokens --- modules/sd_hijack.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules') diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 36198a3c..1f8587d1 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -332,8 +332,8 @@ class FrozenCLIPEmbedderWithCustomWords(torch.nn.Module): multipliers.append([1.0] * 75) z1 = self.process_tokens(tokens, multipliers) + z1 = shared.aesthetic_clip(z1, remade_batch_tokens) z = z1 if z is None else torch.cat((z, z1), axis=-2) - z = shared.aesthetic_clip(z, remade_batch_tokens) remade_batch_tokens = rem_tokens batch_multipliers = rem_multipliers -- cgit v1.2.3 From d0ea471b0cdaede163c6e7f6fae8535f5c3cd226 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Fri, 21 Oct 2022 14:04:41 +0100 Subject: Use opts in textual_inversion image_embedding.py for dynamic fonts --- modules/textual_inversion/image_embedding.py | 1 + 1 file changed, 1 insertion(+) (limited to 'modules') diff --git a/modules/textual_inversion/image_embedding.py b/modules/textual_inversion/image_embedding.py index 898ce3b3..c50b1e7b 100644 --- a/modules/textual_inversion/image_embedding.py +++ b/modules/textual_inversion/image_embedding.py @@ -5,6 +5,7 @@ import zlib from PIL import Image, PngImagePlugin, ImageDraw, ImageFont from fonts.ttf import Roboto import torch +from modules.shared import opts class EmbeddingEncoder(json.JSONEncoder): -- cgit v1.2.3 From 306e2ff6ab8f4c7e94ab55f4f08ab8f94d73d287 Mon Sep 17 00:00:00 2001 From: DepFA <35278260+dfaker@users.noreply.github.com> Date: Fri, 21 Oct 2022 14:47:21 +0100 Subject: Update image_embedding.py --- modules/textual_inversion/image_embedding.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'modules') diff --git a/modules/textual_inversion/image_embedding.py b/modules/textual_inversion/image_embedding.py index c50b1e7b..ea653806 100644 --- a/modules/textual_inversion/image_embedding.py +++ b/modules/textual_inversion/image_embedding.py @@ -134,7 +134,7 @@ def caption_image_overlay(srcimage, title, footerLeft, footerMid, footerRight, t from math import cos image = srcimage.copy() - + fontsize = 32 if textfont is None: try: textfont = ImageFont.truetype(opts.font or Roboto, fontsize) @@ -151,7 +151,7 @@ def caption_image_overlay(srcimage, title, footerLeft, footerMid, footerRight, t image = Image.alpha_composite(image.convert('RGBA'), gradient.resize(image.size)) draw = ImageDraw.Draw(image) - fontsize = 32 + font = ImageFont.truetype(textfont, fontsize) padding = 10 -- cgit v1.2.3