diff options
28 files changed, 585 insertions, 130 deletions
diff --git a/configs/alt-diffusion-inference.yaml b/configs/alt-diffusion-inference.yaml new file mode 100644 index 00000000..cfbee72d --- /dev/null +++ b/configs/alt-diffusion-inference.yaml @@ -0,0 +1,72 @@ +model: + base_learning_rate: 1.0e-04 + target: ldm.models.diffusion.ddpm.LatentDiffusion + params: + linear_start: 0.00085 + linear_end: 0.0120 + num_timesteps_cond: 1 + log_every_t: 200 + timesteps: 1000 + first_stage_key: "jpg" + cond_stage_key: "txt" + image_size: 64 + channels: 4 + cond_stage_trainable: false # Note: different from the one we trained before + conditioning_key: crossattn + monitor: val/loss_simple_ema + scale_factor: 0.18215 + use_ema: False + + scheduler_config: # 10000 warmup steps + target: ldm.lr_scheduler.LambdaLinearScheduler + params: + warm_up_steps: [ 10000 ] + cycle_lengths: [ 10000000000000 ] # incredibly large number to prevent corner cases + f_start: [ 1.e-6 ] + f_max: [ 1. ] + f_min: [ 1. ] + + unet_config: + target: ldm.modules.diffusionmodules.openaimodel.UNetModel + params: + image_size: 32 # unused + in_channels: 4 + out_channels: 4 + model_channels: 320 + attention_resolutions: [ 4, 2, 1 ] + num_res_blocks: 2 + channel_mult: [ 1, 2, 4, 4 ] + num_heads: 8 + use_spatial_transformer: True + transformer_depth: 1 + context_dim: 768 + use_checkpoint: True + legacy: False + + first_stage_config: + target: ldm.models.autoencoder.AutoencoderKL + params: + embed_dim: 4 + monitor: val/rec_loss + ddconfig: + double_z: true + z_channels: 4 + resolution: 256 + in_channels: 3 + out_ch: 3 + ch: 128 + ch_mult: + - 1 + - 2 + - 4 + - 4 + num_res_blocks: 2 + attn_resolutions: [] + dropout: 0.0 + lossconfig: + target: torch.nn.Identity + + cond_stage_config: + target: modules.xlmr.BertSeriesModelWithTransformation + params: + name: "XLMR-Large"
\ No newline at end of file diff --git a/v1-inference.yaml b/configs/v1-inference.yaml index d4effe56..d4effe56 100644 --- a/v1-inference.yaml +++ b/configs/v1-inference.yaml diff --git a/extensions-builtin/roll-artist/scripts/roll-artist.py b/extensions-builtin/roll-artist/scripts/roll-artist.py new file mode 100644 index 00000000..c3bc1fd0 --- /dev/null +++ b/extensions-builtin/roll-artist/scripts/roll-artist.py @@ -0,0 +1,50 @@ +import random
+
+from modules import script_callbacks, shared
+import gradio as gr
+
+art_symbol = '\U0001f3a8' # 🎨
+global_prompt = None
+related_ids = {"txt2img_prompt", "txt2img_clear_prompt", "img2img_prompt", "img2img_clear_prompt" }
+
+
+def roll_artist(prompt):
+ allowed_cats = set([x for x in shared.artist_db.categories() if len(shared.opts.random_artist_categories)==0 or x in shared.opts.random_artist_categories])
+ artist = random.choice([x for x in shared.artist_db.artists if x.category in allowed_cats])
+
+ return prompt + ", " + artist.name if prompt != '' else artist.name
+
+
+def add_roll_button(prompt):
+ roll = gr.Button(value=art_symbol, elem_id="roll", visible=len(shared.artist_db.artists) > 0)
+
+ roll.click(
+ fn=roll_artist,
+ _js="update_txt2img_tokens",
+ inputs=[
+ prompt,
+ ],
+ outputs=[
+ prompt,
+ ]
+ )
+
+
+def after_component(component, **kwargs):
+ global global_prompt
+
+ elem_id = kwargs.get('elem_id', None)
+ if elem_id not in related_ids:
+ return
+
+ if elem_id == "txt2img_prompt":
+ global_prompt = component
+ elif elem_id == "txt2img_clear_prompt":
+ add_roll_button(global_prompt)
+ elif elem_id == "img2img_prompt":
+ global_prompt = component
+ elif elem_id == "img2img_clear_prompt":
+ add_roll_button(global_prompt)
+
+
+script_callbacks.on_after_component(after_component)
diff --git a/modules/api/api.py b/modules/api/api.py index 1ceba75d..11daff0d 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -121,7 +121,6 @@ class Api: def text2imgapi(self, txt2imgreq: StableDiffusionTxt2ImgProcessingAPI): populate = txt2imgreq.copy(update={ # Override __init__ params - "sd_model": shared.sd_model, "sampler_name": validate_sampler_name(txt2imgreq.sampler_name or txt2imgreq.sampler_index), "do_not_save_samples": True, "do_not_save_grid": True @@ -129,15 +128,14 @@ class Api: ) if populate.sampler_name: populate.sampler_index = None # prevent a warning later on - p = StableDiffusionProcessingTxt2Img(**vars(populate)) - # Override object param - - shared.state.begin() with self.queue_lock: + p = StableDiffusionProcessingTxt2Img(sd_model=shared.sd_model, **vars(populate)) + + shared.state.begin() processed = process_images(p) + shared.state.end() - shared.state.end() b64images = list(map(encode_pil_to_base64, processed.images)) @@ -153,7 +151,6 @@ class Api: mask = decode_base64_to_image(mask) populate = img2imgreq.copy(update={ # Override __init__ params - "sd_model": shared.sd_model, "sampler_name": validate_sampler_name(img2imgreq.sampler_name or img2imgreq.sampler_index), "do_not_save_samples": True, "do_not_save_grid": True, @@ -165,16 +162,14 @@ class Api: args = vars(populate) args.pop('include_init_images', None) # this is meant to be done by "exclude": True in model, but it's for a reason that I cannot determine. - p = StableDiffusionProcessingImg2Img(**args) - - p.init_images = [decode_base64_to_image(x) for x in init_images] - - shared.state.begin() with self.queue_lock: - processed = process_images(p) + p = StableDiffusionProcessingImg2Img(sd_model=shared.sd_model, **args) + p.init_images = [decode_base64_to_image(x) for x in init_images] - shared.state.end() + shared.state.begin() + processed = process_images(p) + shared.state.end() b64images = list(map(encode_pil_to_base64, processed.images)) diff --git a/modules/generation_parameters_copypaste.py b/modules/generation_parameters_copypaste.py index fbd91300..54b3372d 100644 --- a/modules/generation_parameters_copypaste.py +++ b/modules/generation_parameters_copypaste.py @@ -38,7 +38,7 @@ def quote(text): def image_from_url_text(filedata):
if type(filedata) == dict and filedata["is_file"]:
filename = filedata["name"]
- is_in_right_dir = any(Path(temp_dir).resolve() in Path(filename).resolve().parents for temp_dir in shared.demo.temp_dirs)
+ is_in_right_dir = any([filename in fileset for fileset in shared.demo.temp_file_sets])
assert is_in_right_dir, 'trying to open image file outside of allowed directories'
return Image.open(filename)
diff --git a/modules/images.py b/modules/images.py index 5afd3891..719aaf3b 100644 --- a/modules/images.py +++ b/modules/images.py @@ -526,6 +526,9 @@ def save_image(image, path, basename, seed=None, prompt=None, extension='png', i image_to_save.save(temp_file_path, format=image_format, quality=opts.jpeg_quality, pnginfo=pnginfo_data)
elif extension.lower() in (".jpg", ".jpeg", ".webp"):
+ if image_to_save.mode == 'RGBA':
+ image_to_save = image_to_save.convert("RGB")
+
image_to_save.save(temp_file_path, format=image_format, quality=opts.jpeg_quality)
if opts.enable_pnginfo and info is not None:
diff --git a/modules/img2img.py b/modules/img2img.py index 81da4b13..ca58b5d8 100644 --- a/modules/img2img.py +++ b/modules/img2img.py @@ -162,4 +162,4 @@ def img2img(mode: int, prompt: str, negative_prompt: str, prompt_style: str, pro if opts.do_not_show_images:
processed.images = []
- return processed.images, generation_info_js, plaintext_to_html(processed.info)
+ return processed.images, generation_info_js, plaintext_to_html(processed.info), plaintext_to_html(processed.comments)
diff --git a/modules/interrogate.py b/modules/interrogate.py index 46935210..6f761c5a 100644 --- a/modules/interrogate.py +++ b/modules/interrogate.py @@ -135,7 +135,7 @@ class InterrogateModels: return caption[0]
def interrogate(self, pil_image):
- res = None
+ res = ""
try:
diff --git a/modules/memmon.py b/modules/memmon.py index 9fb9b687..a7060f58 100644 --- a/modules/memmon.py +++ b/modules/memmon.py @@ -71,10 +71,13 @@ class MemUsageMonitor(threading.Thread): def read(self): if not self.disabled: free, total = torch.cuda.mem_get_info() + self.data["free"] = free self.data["total"] = total torch_stats = torch.cuda.memory_stats(self.device) + self.data["active"] = torch_stats["active.all.current"] self.data["active_peak"] = torch_stats["active_bytes.all.peak"] + self.data["reserved"] = torch_stats["reserved_bytes.all.current"] self.data["reserved_peak"] = torch_stats["reserved_bytes.all.peak"] self.data["system_peak"] = total - self.data["min_free"] diff --git a/modules/processing.py b/modules/processing.py index 0a9a8f95..42dc19ea 100644 --- a/modules/processing.py +++ b/modules/processing.py @@ -239,7 +239,7 @@ class StableDiffusionProcessing(): class Processed:
- def __init__(self, p: StableDiffusionProcessing, images_list, seed=-1, info="", subseed=None, all_prompts=None, all_negative_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_negative_prompts=None, all_seeds=None, all_subseeds=None, index_of_first_image=0, infotexts=None, comments=""):
self.images = images_list
self.prompt = p.prompt
self.negative_prompt = p.negative_prompt
@@ -247,6 +247,7 @@ class Processed: self.subseed = subseed
self.subseed_strength = p.subseed_strength
self.info = info
+ self.comments = comments
self.width = p.width
self.height = p.height
self.sampler_name = p.sampler_name
@@ -646,7 +647,7 @@ def process_images_inner(p: StableDiffusionProcessing) -> Processed: devices.torch_gc()
- res = Processed(p, output_images, p.all_seeds[0], infotext() + "".join(["\n\n" + x for x in comments]), subseed=p.all_subseeds[0], index_of_first_image=index_of_first_image, infotexts=infotexts)
+ res = Processed(p, output_images, p.all_seeds[0], infotext(), comments="".join(["\n\n" + x for x in comments]), subseed=p.all_subseeds[0], index_of_first_image=index_of_first_image, infotexts=infotexts)
if p.scripts is not None:
p.scripts.postprocess(p, res)
diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index 690a9ec2..edcbaf52 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -5,7 +5,7 @@ import modules.textual_inversion.textual_inversion from modules import devices, sd_hijack_optimizations, shared, sd_hijack_checkpoint
from modules.hypernetworks import hypernetwork
from modules.shared import cmd_opts
-from modules import sd_hijack_clip, sd_hijack_open_clip, sd_hijack_unet
+from modules import sd_hijack_clip, sd_hijack_open_clip, sd_hijack_unet, sd_hijack_xlmr, xlmr
from modules.sd_hijack_optimizations import invokeAI_mps_available
@@ -68,6 +68,7 @@ def fix_checkpoint(): ldm.modules.diffusionmodules.openaimodel.ResBlock.forward = sd_hijack_checkpoint.ResBlock_forward
ldm.modules.diffusionmodules.openaimodel.AttentionBlock.forward = sd_hijack_checkpoint.AttentionBlock_forward
+
class StableDiffusionModelHijack:
fixes = None
comments = []
@@ -78,17 +79,25 @@ class StableDiffusionModelHijack: embedding_db = modules.textual_inversion.textual_inversion.EmbeddingDatabase(cmd_opts.embeddings_dir)
def hijack(self, m):
- if type(m.cond_stage_model) == ldm.modules.encoders.modules.FrozenCLIPEmbedder:
+
+ if type(m.cond_stage_model) == xlmr.BertSeriesModelWithTransformation:
+ model_embeddings = m.cond_stage_model.roberta.embeddings
+ model_embeddings.token_embedding = EmbeddingsWithFixes(model_embeddings.word_embeddings, self)
+ m.cond_stage_model = sd_hijack_xlmr.FrozenXLMREmbedderWithCustomWords(m.cond_stage_model, self)
+
+ elif type(m.cond_stage_model) == ldm.modules.encoders.modules.FrozenCLIPEmbedder:
model_embeddings = m.cond_stage_model.transformer.text_model.embeddings
model_embeddings.token_embedding = EmbeddingsWithFixes(model_embeddings.token_embedding, self)
m.cond_stage_model = sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords(m.cond_stage_model, self)
+
elif type(m.cond_stage_model) == ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder:
m.cond_stage_model.model.token_embedding = EmbeddingsWithFixes(m.cond_stage_model.model.token_embedding, self)
m.cond_stage_model = sd_hijack_open_clip.FrozenOpenCLIPEmbedderWithCustomWords(m.cond_stage_model, self)
- self.clip = m.cond_stage_model
-
apply_optimizations()
+
+ self.clip = m.cond_stage_model
+
fix_checkpoint()
def flatten(el):
@@ -101,7 +110,11 @@ class StableDiffusionModelHijack: self.layers = flatten(m)
def undo_hijack(self, m):
- if type(m.cond_stage_model) == sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords:
+
+ if type(m.cond_stage_model) == xlmr.BertSeriesModelWithTransformation:
+ m.cond_stage_model = m.cond_stage_model.wrapped
+
+ elif type(m.cond_stage_model) == sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords:
m.cond_stage_model = m.cond_stage_model.wrapped
model_embeddings = m.cond_stage_model.transformer.text_model.embeddings
@@ -129,8 +142,8 @@ class StableDiffusionModelHijack: def tokenize(self, text):
_, remade_batch_tokens, _, _, _, token_count = self.clip.process_text([text])
- return remade_batch_tokens[0], token_count, sd_hijack_clip.get_target_prompt_token_count(token_count)
+ return remade_batch_tokens[0], token_count, sd_hijack_clip.get_target_prompt_token_count(token_count)
class EmbeddingsWithFixes(torch.nn.Module):
diff --git a/modules/sd_hijack_clip.py b/modules/sd_hijack_clip.py index b451d1cf..ca92b142 100644 --- a/modules/sd_hijack_clip.py +++ b/modules/sd_hijack_clip.py @@ -5,7 +5,6 @@ import torch from modules import prompt_parser, devices
from modules.shared import opts
-
def get_target_prompt_token_count(token_count):
return math.ceil(max(token_count, 1) / 75) * 75
@@ -254,10 +253,13 @@ class FrozenCLIPEmbedderWithCustomWords(FrozenCLIPEmbedderWithCustomWordsBase): def __init__(self, wrapped, hijack):
super().__init__(wrapped, hijack)
self.tokenizer = wrapped.tokenizer
- self.comma_token = [v for k, v in self.tokenizer.get_vocab().items() if k == ',</w>'][0]
+
+ vocab = self.tokenizer.get_vocab()
+
+ self.comma_token = vocab.get(',</w>', None)
self.token_mults = {}
- 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 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:
@@ -296,6 +298,6 @@ class FrozenCLIPEmbedderWithCustomWords(FrozenCLIPEmbedderWithCustomWordsBase): def encode_embedding_init_text(self, init_text, nvpt):
embedding_layer = self.wrapped.transformer.text_model.embeddings
ids = self.wrapped.tokenizer(init_text, max_length=nvpt, return_tensors="pt", add_special_tokens=False)["input_ids"]
- embedded = embedding_layer.token_embedding.wrapped(ids.to(devices.device)).squeeze(0)
+ embedded = embedding_layer.token_embedding.wrapped(ids.to(embedding_layer.token_embedding.wrapped.weight.device)).squeeze(0)
return embedded
diff --git a/modules/sd_hijack_inpainting.py b/modules/sd_hijack_inpainting.py index bb5499b3..06b75772 100644 --- a/modules/sd_hijack_inpainting.py +++ b/modules/sd_hijack_inpainting.py @@ -178,7 +178,7 @@ def sample_plms(self, # sampling C, H, W = shape size = (batch_size, C, H, W) - print(f'Data shape for PLMS sampling is {size}') + # print(f'Data shape for PLMS sampling is {size}') # remove unnecessary message samples, intermediates = self.plms_sampling(conditioning, size, callback=callback, diff --git a/modules/sd_hijack_xlmr.py b/modules/sd_hijack_xlmr.py new file mode 100644 index 00000000..4ac51c38 --- /dev/null +++ b/modules/sd_hijack_xlmr.py @@ -0,0 +1,34 @@ +import open_clip.tokenizer
+import torch
+
+from modules import sd_hijack_clip, devices
+from modules.shared import opts
+
+
+class FrozenXLMREmbedderWithCustomWords(sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords):
+ def __init__(self, wrapped, hijack):
+ super().__init__(wrapped, hijack)
+
+ self.id_start = wrapped.config.bos_token_id
+ self.id_end = wrapped.config.eos_token_id
+ self.id_pad = wrapped.config.pad_token_id
+
+ self.comma_token = self.tokenizer.get_vocab().get(',', None) # alt diffusion doesn't have </w> bits for comma
+
+ def encode_with_transformers(self, tokens):
+ # there's no CLIP Skip here because all hidden layers have size of 1024 and the last one uses a
+ # trained layer to transform those 1024 into 768 for unet; so you can't choose which transformer
+ # layer to work with - you have to use the last
+
+ attention_mask = (tokens != self.id_pad).to(device=tokens.device, dtype=torch.int64)
+ features = self.wrapped(input_ids=tokens, attention_mask=attention_mask)
+ z = features['projection_state']
+
+ return z
+
+ def encode_embedding_init_text(self, init_text, nvpt):
+ embedding_layer = self.wrapped.roberta.embeddings
+ ids = self.wrapped.tokenizer(init_text, max_length=nvpt, return_tensors="pt", add_special_tokens=False)["input_ids"]
+ embedded = embedding_layer.token_embedding.wrapped(ids.to(devices.device)).squeeze(0)
+
+ return embedded
diff --git a/modules/sd_models.py b/modules/sd_models.py index ecdd91c5..ebd4dff7 100644 --- a/modules/sd_models.py +++ b/modules/sd_models.py @@ -325,6 +325,9 @@ def load_model(checkpoint_info=None): script_callbacks.model_loaded_callback(sd_model)
print("Model loaded.")
+
+ sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings(force_reload = True) # Reload embeddings after model load as they may or may not fit the model
+
return sd_model
diff --git a/modules/sd_samplers.py b/modules/sd_samplers.py index 177b5338..e904d860 100644 --- a/modules/sd_samplers.py +++ b/modules/sd_samplers.py @@ -465,7 +465,9 @@ class KDiffusionSampler: if p.sampler_noise_scheduler_override:
sigmas = p.sampler_noise_scheduler_override(steps)
elif self.config is not None and self.config.options.get('scheduler', None) == 'karras':
- sigmas = k_diffusion.sampling.get_sigmas_karras(n=steps, sigma_min=0.1, sigma_max=10, device=shared.device)
+ sigma_min, sigma_max = (0.1, 10) if opts.use_old_karras_scheduler_sigmas else (self.model_wrap.sigmas[0].item(), self.model_wrap.sigmas[-1].item())
+
+ sigmas = k_diffusion.sampling.get_sigmas_karras(n=steps, sigma_min=sigma_min, sigma_max=sigma_max, device=shared.device)
else:
sigmas = self.model_wrap.get_sigmas(steps)
diff --git a/modules/sd_vae.py b/modules/sd_vae.py index 3856418e..ac71d62d 100644 --- a/modules/sd_vae.py +++ b/modules/sd_vae.py @@ -1,5 +1,6 @@ import torch import os +import collections from collections import namedtuple from modules import shared, devices, script_callbacks from modules.paths import models_path @@ -30,6 +31,7 @@ base_vae = None loaded_vae_file = None checkpoint_info = None +checkpoints_loaded = collections.OrderedDict() def get_base_vae(model): if base_vae is not None and checkpoint_info == model.sd_checkpoint_info and model: @@ -149,13 +151,30 @@ def load_vae(model, vae_file=None): global first_load, vae_dict, vae_list, loaded_vae_file # save_settings = False + cache_enabled = shared.opts.sd_vae_checkpoint_cache > 0 + if vae_file: - assert os.path.isfile(vae_file), f"VAE file doesn't exist: {vae_file}" - print(f"Loading VAE weights from: {vae_file}") - store_base_vae(model) - vae_ckpt = torch.load(vae_file, map_location=shared.weight_load_location) - vae_dict_1 = {k: v for k, v in vae_ckpt["state_dict"].items() if k[0:4] != "loss" and k not in vae_ignore_keys} - _load_vae_dict(model, vae_dict_1) + if cache_enabled and vae_file in checkpoints_loaded: + # use vae checkpoint cache + print(f"Loading VAE weights [{get_filename(vae_file)}] from cache") + store_base_vae(model) + _load_vae_dict(model, checkpoints_loaded[vae_file]) + else: + assert os.path.isfile(vae_file), f"VAE file doesn't exist: {vae_file}" + print(f"Loading VAE weights from: {vae_file}") + store_base_vae(model) + vae_ckpt = torch.load(vae_file, map_location=shared.weight_load_location) + vae_dict_1 = {k: v for k, v in vae_ckpt["state_dict"].items() if k[0:4] != "loss" and k not in vae_ignore_keys} + _load_vae_dict(model, vae_dict_1) + + if cache_enabled: + # cache newly loaded vae + checkpoints_loaded[vae_file] = vae_dict_1.copy() + + # clean up cache if limit is reached + if cache_enabled: + while len(checkpoints_loaded) > shared.opts.sd_vae_checkpoint_cache + 1: # we need to count the current model + checkpoints_loaded.popitem(last=False) # LRU # If vae used is not in dict, update it # It will be removed on refresh though diff --git a/modules/shared.py b/modules/shared.py index d4ddeea0..7f430b93 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -23,7 +23,7 @@ demo = None sd_model_file = os.path.join(script_path, 'model.ckpt')
default_sd_model_file = sd_model_file
parser = argparse.ArgumentParser()
-parser.add_argument("--config", type=str, default=os.path.join(script_path, "v1-inference.yaml"), help="path to config which constructs model",)
+parser.add_argument("--config", type=str, default=os.path.join(script_path, "configs/v1-inference.yaml"), help="path to config which constructs model",)
parser.add_argument("--ckpt", type=str, default=sd_model_file, help="path to checkpoint of stable diffusion model; if specified, this checkpoint will be added to the list of checkpoints and loaded",)
parser.add_argument("--ckpt-dir", type=str, default=None, help="Path to directory with stable diffusion checkpoints")
parser.add_argument("--gfpgan-dir", type=str, help="GFPGAN directory", default=('./src/gfpgan' if os.path.exists('./src/gfpgan') else './GFPGAN'))
@@ -168,7 +168,7 @@ class State: def dict(self):
obj = {
"skipped": self.skipped,
- "interrupted": self.skipped,
+ "interrupted": self.interrupted,
"job": self.job,
"job_count": self.job_count,
"job_no": self.job_no,
@@ -356,6 +356,7 @@ options_templates.update(options_section(('training', "Training"), { options_templates.update(options_section(('sd', "Stable Diffusion"), {
"sd_model_checkpoint": OptionInfo(None, "Stable Diffusion checkpoint", gr.Dropdown, lambda: {"choices": list_checkpoint_tiles()}, refresh=refresh_checkpoints),
"sd_checkpoint_cache": OptionInfo(0, "Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}),
+ "sd_vae_checkpoint_cache": OptionInfo(0, "VAE Checkpoints to cache in RAM", gr.Slider, {"minimum": 0, "maximum": 10, "step": 1}),
"sd_vae": OptionInfo("auto", "SD VAE", gr.Dropdown, lambda: {"choices": sd_vae.vae_list}, refresh=sd_vae.refresh_vae_list),
"sd_vae_as_default": OptionInfo(False, "Ignore selected VAE for stable diffusion checkpoints that have their own .vae.pt next to them"),
"sd_hypernetwork": OptionInfo("None", "Hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}, refresh=reload_hypernetworks),
@@ -367,13 +368,17 @@ options_templates.update(options_section(('sd', "Stable Diffusion"), { "img2img_background_color": OptionInfo("#ffffff", "With img2img, fill image's transparent parts with this color.", gr.ColorPicker, {}),
"enable_quantization": OptionInfo(False, "Enable quantization in K samplers for sharper and cleaner results. This may change existing seeds. Requires restart to apply."),
"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"),
"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 }),
'CLIP_stop_at_last_layers': OptionInfo(1, "Clip skip", gr.Slider, {"minimum": 1, "maximum": 12, "step": 1}),
"random_artist_categories": OptionInfo([], "Allowed categories for random artists selection when using the Roll button", gr.CheckboxGroup, {"choices": artist_db.categories()}),
}))
+options_templates.update(options_section(('compatibility', "Compatibility"), {
+ "use_old_emphasis_implementation": OptionInfo(False, "Use old emphasis implementation. Can be useful to reproduce old seeds."),
+ "use_old_karras_scheduler_sigmas": OptionInfo(False, "Use old karras scheduler sigmas (0.1 to 10)."),
+}))
+
options_templates.update(options_section(('interrogate', "Interrogate Options"), {
"interrogate_keep_models_in_memory": OptionInfo(False, "Interrogate: keep models in VRAM"),
"interrogate_use_builtin_artists": OptionInfo(True, "Interrogate: use artists from artists.csv"),
@@ -405,6 +410,7 @@ options_templates.update(options_section(('ui', "User interface"), { "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"),
"show_progress_in_title": OptionInfo(True, "Show generation progress in window title."),
+ "samplers_in_dropdown": OptionInfo(True, "Use dropdown for sampler selection instead of radio group"),
'quicksettings': OptionInfo("sd_model_checkpoint", "Quicksettings list"),
'localization': OptionInfo("None", "Localization (requires restart)", gr.Dropdown, lambda: {"choices": ["None"] + list(localization.localizations.keys())}, refresh=lambda: localization.list_localizations(cmd_opts.localizations_dir)),
}))
diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index f6112578..66f40367 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -23,6 +23,8 @@ class Embedding: self.vec = vec
self.name = name
self.step = step
+ self.shape = None
+ self.vectors = 0
self.cached_checksum = None
self.sd_checkpoint = None
self.sd_checkpoint_name = None
@@ -57,8 +59,10 @@ class EmbeddingDatabase: def __init__(self, embeddings_dir):
self.ids_lookup = {}
self.word_embeddings = {}
+ self.skipped_embeddings = []
self.dir_mtime = None
self.embeddings_dir = embeddings_dir
+ self.expected_shape = -1
def register_embedding(self, embedding, model):
@@ -75,20 +79,24 @@ class EmbeddingDatabase: return embedding
- def load_textual_inversion_embeddings(self):
+ def get_expected_shape(self):
+ vec = shared.sd_model.cond_stage_model.encode_embedding_init_text(",", 1)
+ return vec.shape[1]
+
+ def load_textual_inversion_embeddings(self, force_reload = False):
mt = os.path.getmtime(self.embeddings_dir)
- if self.dir_mtime is not None and mt <= self.dir_mtime:
+ if not force_reload and self.dir_mtime is not None and mt <= self.dir_mtime:
return
self.dir_mtime = mt
self.ids_lookup.clear()
self.word_embeddings.clear()
+ self.skipped_embeddings = []
+ self.expected_shape = self.get_expected_shape()
def process_file(path, filename):
name = os.path.splitext(filename)[0]
- data = []
-
if os.path.splitext(filename.upper())[-1] in ['.PNG', '.WEBP', '.JXL', '.AVIF']:
embed_image = Image.open(path)
if hasattr(embed_image, 'text') and 'sd-ti-embedding' in embed_image.text:
@@ -122,7 +130,13 @@ class EmbeddingDatabase: embedding.step = data.get('step', None)
embedding.sd_checkpoint = data.get('sd_checkpoint', None)
embedding.sd_checkpoint_name = data.get('sd_checkpoint_name', None)
- self.register_embedding(embedding, shared.sd_model)
+ embedding.vectors = vec.shape[0]
+ embedding.shape = vec.shape[-1]
+
+ if self.expected_shape == -1 or self.expected_shape == embedding.shape:
+ self.register_embedding(embedding, shared.sd_model)
+ else:
+ self.skipped_embeddings.append(name)
for fn in os.listdir(self.embeddings_dir):
try:
@@ -137,8 +151,9 @@ class EmbeddingDatabase: print(traceback.format_exc(), file=sys.stderr)
continue
- print(f"Loaded a total of {len(self.word_embeddings)} textual inversion embeddings.")
- print("Embeddings:", ', '.join(self.word_embeddings.keys()))
+ print(f"Textual inversion embeddings loaded({len(self.word_embeddings)}): {', '.join(self.word_embeddings.keys())}")
+ if len(self.skipped_embeddings) > 0:
+ print(f"Textual inversion embeddings skipped({len(self.skipped_embeddings)}): {', '.join(self.skipped_embeddings)}")
def find_embedding_at_position(self, tokens, offset):
token = tokens[offset]
diff --git a/modules/txt2img.py b/modules/txt2img.py index c8f81176..7f61e19a 100644 --- a/modules/txt2img.py +++ b/modules/txt2img.py @@ -59,4 +59,4 @@ def txt2img(prompt: str, negative_prompt: str, prompt_style: str, prompt_style2: if opts.do_not_show_images:
processed.images = []
- return processed.images, generation_info_js, plaintext_to_html(processed.info)
+ return processed.images, generation_info_js, plaintext_to_html(processed.info), plaintext_to_html(processed.comments)
diff --git a/modules/ui.py b/modules/ui.py index 57ee0465..27da2c2c 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -19,7 +19,7 @@ import numpy as np from PIL import Image, PngImagePlugin
from modules.call_queue import wrap_gradio_gpu_call, wrap_queued_call, wrap_gradio_call
-from modules import sd_hijack, sd_models, localization, script_callbacks, ui_extensions, deepbooru
+from modules import sd_hijack, sd_models, localization, script_callbacks, ui_extensions, deepbooru, ui_components
from modules.paths import script_path
from modules.shared import opts, cmd_opts, restricted_opts
@@ -80,7 +80,6 @@ css_hide_progressbar = """ # Important that they exactly match script.js for tooltip to work.
random_symbol = '\U0001f3b2\ufe0f' # 🎲️
reuse_symbol = '\u267b\ufe0f' # ♻️
-art_symbol = '\U0001f3a8' # 🎨
paste_symbol = '\u2199\ufe0f' # ↙
folder_symbol = '\U0001f4c2' # 📂
refresh_symbol = '\U0001f504' # 🔄
@@ -159,7 +158,7 @@ def save_files(js_data, images, do_make_zip, index): zip_file.writestr(filenames[i], f.read())
fullfns.insert(0, zip_filepath)
- return gr.File.update(value=fullfns, visible=True), '', '', plaintext_to_html(f"Saved: {filenames[0]}")
+ return gr.File.update(value=fullfns, visible=True), plaintext_to_html(f"Saved: {filenames[0]}")
@@ -234,13 +233,6 @@ def check_progress_call_initial(id_part): return check_progress_call(id_part)
-def roll_artist(prompt):
- allowed_cats = set([x for x in shared.artist_db.categories() if len(opts.random_artist_categories)==0 or x in opts.random_artist_categories])
- artist = random.choice([x for x in shared.artist_db.artists if x.category in allowed_cats])
-
- return prompt + ", " + artist.name if prompt != '' else artist.name
-
-
def visit(x, func, path=""):
if hasattr(x, 'children'):
for c in x.children:
@@ -403,7 +395,6 @@ def create_toprow(is_img2img): )
with gr.Column(scale=1, elem_id="roll_col"):
- roll = gr.Button(value=art_symbol, elem_id="roll", visible=len(shared.artist_db.artists) > 0)
paste = gr.Button(value=paste_symbol, elem_id="paste")
save_style = gr.Button(value=save_style_symbol, elem_id="style_create")
prompt_style_apply = gr.Button(value=apply_style_symbol, elem_id="style_apply")
@@ -452,7 +443,7 @@ def create_toprow(is_img2img): 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())))
prompt_style2.save_to_config = True
- return prompt, roll, prompt_style, negative_prompt, prompt_style2, submit, button_interrogate, button_deepbooru, prompt_style_apply, save_style, paste, token_counter, token_button
+ return prompt, prompt_style, negative_prompt, prompt_style2, submit, button_interrogate, button_deepbooru, prompt_style_apply, save_style, paste, token_counter, token_button
def setup_progressbar(progressbar, preview, id_part, textinfo=None):
@@ -532,7 +523,7 @@ def create_refresh_button(refresh_component, refresh_method, refreshed_args, ele return gr.update(**(args or {}))
- refresh_button = gr.Button(value=refresh_symbol, elem_id=elem_id)
+ refresh_button = ui_components.ToolButton(value=refresh_symbol, elem_id=elem_id)
refresh_button.click(
fn=refresh,
inputs=[],
@@ -570,13 +561,14 @@ Requested path was: {f} generation_info = None
with gr.Column():
- with gr.Row():
+ with gr.Row(elem_id=f"image_buttons_{tabname}"):
+ open_folder_button = gr.Button(folder_symbol, elem_id="hidden_element" if shared.cmd_opts.hide_ui_dir_config else 'open_folder')
+
if tabname != "extras":
save = gr.Button('Save', elem_id=f'save_{tabname}')
+ save_zip = gr.Button('Zip', elem_id=f'save_zip_{tabname}')
buttons = parameters_copypaste.create_buttons(["img2img", "inpaint", "extras"])
- button_id = "hidden_element" if shared.cmd_opts.hide_ui_dir_config else 'open_folder'
- open_folder_button = gr.Button(folder_symbol, elem_id=button_id)
open_folder_button.click(
fn=lambda: open_folder(opts.outdir_samples or outdir),
@@ -586,13 +578,12 @@ Requested path was: {f} if tabname != "extras":
with gr.Row():
- do_make_zip = gr.Checkbox(label="Make Zip when Save?", value=False)
-
- with gr.Row():
download_files = gr.File(None, file_count="multiple", interactive=False, show_label=False, visible=False)
with gr.Group():
html_info = gr.HTML()
+ html_log = gr.HTML()
+
generation_info = gr.Textbox(visible=False)
if tabname == 'txt2img' or tabname == 'img2img':
generation_info_button = gr.Button(visible=False, elem_id=f"{tabname}_generation_info_button")
@@ -606,25 +597,54 @@ Requested path was: {f} save.click(
fn=wrap_gradio_call(save_files),
- _js="(x, y, z, w) => [x, y, z, selected_gallery_index()]",
+ _js="(x, y, z, w) => [x, y, false, selected_gallery_index()]",
inputs=[
generation_info,
result_gallery,
- do_make_zip,
+ html_info,
html_info,
],
outputs=[
download_files,
+ html_log,
+ ]
+ )
+
+ save_zip.click(
+ fn=wrap_gradio_call(save_files),
+ _js="(x, y, z, w) => [x, y, true, selected_gallery_index()]",
+ inputs=[
+ generation_info,
+ result_gallery,
html_info,
html_info,
- html_info,
+ ],
+ outputs=[
+ download_files,
+ html_log,
]
)
+
else:
html_info_x = gr.HTML()
html_info = gr.HTML()
+ html_log = gr.HTML()
+
parameters_copypaste.bind_buttons(buttons, result_gallery, "txt2img" if tabname == "txt2img" else None)
- return result_gallery, generation_info if tabname != "extras" else html_info_x, html_info
+ return result_gallery, generation_info if tabname != "extras" else html_info_x, html_info, html_log
+
+
+def create_sampler_and_steps_selection(choices, tabname):
+ if opts.samplers_in_dropdown:
+ with gr.Row(elem_id=f"sampler_selection_{tabname}"):
+ sampler_index = gr.Dropdown(label='Sampling method', elem_id=f"{tabname}_sampling", choices=[x.name for x in choices], value=choices[0].name, type="index")
+ steps = gr.Slider(minimum=1, maximum=150, step=1, elem_id=f"{tabname}_steps", label="Sampling Steps", value=20)
+ else:
+ with gr.Group(elem_id=f"sampler_selection_{tabname}"):
+ steps = gr.Slider(minimum=1, maximum=150, step=1, elem_id=f"{tabname}_steps", label="Sampling Steps", value=20)
+ sampler_index = gr.Radio(label='Sampling method', elem_id=f"{tabname}_sampling", choices=[x.name for x in choices], value=choices[0].name, type="index")
+
+ return steps, sampler_index
def create_ui():
@@ -639,14 +659,11 @@ def create_ui(): modules.scripts.scripts_txt2img.initialize_scripts(is_img2img=False)
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)
+ txt2img_prompt, 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)
dummy_component = gr.Label(visible=False)
txt_prompt_img = gr.File(label="", elem_id="txt2img_prompt_image", file_count="single", type="bytes", visible=False)
-
-
-
with gr.Row(elem_id='txt2img_progress_row'):
with gr.Column(scale=1):
pass
@@ -658,8 +675,7 @@ def create_ui(): with gr.Row().style(equal_height=False):
with gr.Column(variant='panel', elem_id="txt2img_settings"):
- steps = gr.Slider(minimum=1, maximum=150, step=1, label="Sampling Steps", value=20)
- sampler_index = gr.Radio(label='Sampling method', elem_id="txt2img_sampling", choices=[x.name for x in samplers], value=samplers[0].name, type="index")
+ steps, sampler_index = create_sampler_and_steps_selection(samplers, "txt2img")
with gr.Group():
width = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512)
@@ -686,14 +702,14 @@ def create_ui(): with gr.Group():
custom_inputs = modules.scripts.scripts_txt2img.setup_ui()
- txt2img_gallery, generation_info, html_info = create_output_panel("txt2img", opts.outdir_txt2img_samples)
+ txt2img_gallery, generation_info, html_info, html_log = create_output_panel("txt2img", opts.outdir_txt2img_samples)
parameters_copypaste.bind_buttons({"txt2img": txt2img_paste}, None, txt2img_prompt)
connect_reuse_seed(seed, reuse_seed, generation_info, dummy_component, is_subseed=False)
connect_reuse_seed(subseed, reuse_subseed, generation_info, dummy_component, is_subseed=True)
txt2img_args = dict(
- fn=wrap_gradio_gpu_call(modules.txt2img.txt2img),
+ fn=wrap_gradio_gpu_call(modules.txt2img.txt2img, extra_outputs=[None, '', '']),
_js="submit",
inputs=[
txt2img_prompt,
@@ -720,7 +736,8 @@ def create_ui(): outputs=[
txt2img_gallery,
generation_info,
- html_info
+ html_info,
+ html_log,
],
show_progress=False,
)
@@ -745,16 +762,6 @@ def create_ui(): outputs=[hr_options],
)
- roll.click(
- fn=roll_artist,
- _js="update_txt2img_tokens",
- inputs=[
- txt2img_prompt,
- ],
- outputs=[
- txt2img_prompt,
- ]
- )
txt2img_paste_fields = [
(txt2img_prompt, "Prompt"),
@@ -797,8 +804,7 @@ def create_ui(): modules.scripts.scripts_img2img.initialize_scripts(is_img2img=True)
with gr.Blocks(analytics_enabled=False) as img2img_interface:
- img2img_prompt, roll, img2img_prompt_style, img2img_negative_prompt, img2img_prompt_style2, submit, img2img_interrogate, img2img_deepbooru, img2img_prompt_style_apply, img2img_save_style, img2img_paste,token_counter, token_button = create_toprow(is_img2img=True)
-
+ img2img_prompt, img2img_prompt_style, img2img_negative_prompt, img2img_prompt_style2, submit, img2img_interrogate, img2img_deepbooru, img2img_prompt_style_apply, img2img_save_style, img2img_paste,token_counter, token_button = create_toprow(is_img2img=True)
with gr.Row(elem_id='img2img_progress_row'):
img2img_prompt_img = gr.File(label="", elem_id="img2img_prompt_image", file_count="single", type="bytes", visible=False)
@@ -859,8 +865,7 @@ def create_ui(): with gr.Row():
resize_mode = gr.Radio(label="Resize mode", elem_id="resize_mode", show_label=False, choices=["Just resize", "Crop and resize", "Resize and fill", "Just resize (latent upscale)"], type="index", value="Just resize")
- steps = gr.Slider(minimum=1, maximum=150, step=1, label="Sampling Steps", value=20)
- sampler_index = gr.Radio(label='Sampling method', choices=[x.name for x in samplers_for_img2img], value=samplers_for_img2img[0].name, type="index")
+ steps, sampler_index = create_sampler_and_steps_selection(samplers_for_img2img, "img2img")
with gr.Group():
width = gr.Slider(minimum=64, maximum=2048, step=8, label="Width", value=512, elem_id="img2img_width")
@@ -883,7 +888,7 @@ def create_ui(): with gr.Group():
custom_inputs = modules.scripts.scripts_img2img.setup_ui()
- img2img_gallery, generation_info, html_info = create_output_panel("img2img", opts.outdir_img2img_samples)
+ img2img_gallery, generation_info, html_info, html_log = create_output_panel("img2img", opts.outdir_img2img_samples)
parameters_copypaste.bind_buttons({"img2img": img2img_paste}, None, img2img_prompt)
connect_reuse_seed(seed, reuse_seed, generation_info, dummy_component, is_subseed=False)
@@ -915,7 +920,7 @@ def create_ui(): )
img2img_args = dict(
- fn=wrap_gradio_gpu_call(modules.img2img.img2img),
+ fn=wrap_gradio_gpu_call(modules.img2img.img2img, extra_outputs=[None, '', '']),
_js="submit_img2img",
inputs=[
dummy_component,
@@ -954,7 +959,8 @@ def create_ui(): outputs=[
img2img_gallery,
generation_info,
- html_info
+ html_info,
+ html_log,
],
show_progress=False,
)
@@ -974,18 +980,6 @@ def create_ui(): outputs=[img2img_prompt],
)
-
- roll.click(
- fn=roll_artist,
- _js="update_img2img_tokens",
- inputs=[
- img2img_prompt,
- ],
- outputs=[
- img2img_prompt,
- ]
- )
-
prompts = [(txt2img_prompt, txt2img_negative_prompt), (img2img_prompt, img2img_negative_prompt)]
style_dropdowns = [(txt2img_prompt_style, txt2img_prompt_style2), (img2img_prompt_style, img2img_prompt_style2)]
style_js_funcs = ["update_txt2img_tokens", "update_img2img_tokens"]
@@ -1078,10 +1072,10 @@ def create_ui(): with gr.Group():
upscale_before_face_fix = gr.Checkbox(label='Upscale Before Restoring Faces', value=False)
- result_images, html_info_x, html_info = create_output_panel("extras", opts.outdir_extras_samples)
+ result_images, html_info_x, html_info, html_log = create_output_panel("extras", opts.outdir_extras_samples)
submit.click(
- fn=wrap_gradio_gpu_call(modules.extras.run_extras),
+ fn=wrap_gradio_gpu_call(modules.extras.run_extras, extra_outputs=[None, '']),
_js="get_extras_tab_index",
inputs=[
dummy_component,
@@ -1142,8 +1136,14 @@ def create_ui(): with gr.Row():
primary_model_name = gr.Dropdown(modules.sd_models.checkpoint_tiles(), elem_id="modelmerger_primary_model_name", label="Primary model (A)")
+ create_refresh_button(primary_model_name, modules.sd_models.list_models, lambda: {"choices": modules.sd_models.checkpoint_tiles()}, "refresh_checkpoint_A")
+
secondary_model_name = gr.Dropdown(modules.sd_models.checkpoint_tiles(), elem_id="modelmerger_secondary_model_name", label="Secondary model (B)")
+ create_refresh_button(secondary_model_name, modules.sd_models.list_models, lambda: {"choices": modules.sd_models.checkpoint_tiles()}, "refresh_checkpoint_B")
+
tertiary_model_name = gr.Dropdown(modules.sd_models.checkpoint_tiles(), elem_id="modelmerger_tertiary_model_name", label="Tertiary model (C)")
+ create_refresh_button(tertiary_model_name, modules.sd_models.list_models, lambda: {"choices": modules.sd_models.checkpoint_tiles()}, "refresh_checkpoint_C")
+
custom_name = gr.Textbox(label="Custom Name (Optional)")
interp_amount = gr.Slider(minimum=0.0, maximum=1.0, step=0.05, label='Multiplier (M) - set to 0 to get model A', value=0.3)
interp_method = gr.Radio(choices=["Weighted sum", "Add difference"], value="Weighted sum", label="Interpolation Method")
@@ -1157,8 +1157,6 @@ def create_ui(): with gr.Column(variant='panel'):
submit_result = gr.Textbox(elem_id="modelmerger_result", show_label=False)
- sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings()
-
with gr.Blocks(analytics_enabled=False) as train_interface:
with gr.Row().style(equal_height=False):
gr.HTML(value="<p style='margin-bottom: 0.7em'>See <b><a href=\"https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Textual-Inversion\">wiki</a></b> for detailed explanation.</p>")
@@ -1447,7 +1445,7 @@ def create_ui(): res = comp(label=info.label, value=fun(), elem_id=elem_id, **(args or {}))
create_refresh_button(res, info.refresh, info.component_args, "refresh_" + key)
else:
- with gr.Row(variant="compact"):
+ with ui_components.FormRow():
res = comp(label=info.label, value=fun(), elem_id=elem_id, **(args or {}))
create_refresh_button(res, info.refresh, info.component_args, "refresh_" + key)
else:
diff --git a/modules/ui_components.py b/modules/ui_components.py new file mode 100644 index 00000000..d0519d2d --- /dev/null +++ b/modules/ui_components.py @@ -0,0 +1,18 @@ +import gradio as gr
+
+
+class ToolButton(gr.Button, gr.components.FormComponent):
+ """Small button with single emoji as text, fits inside gradio forms"""
+
+ def __init__(self, **kwargs):
+ super().__init__(variant="tool", **kwargs)
+
+ def get_block_name(self):
+ return "button"
+
+
+class FormRow(gr.Row, gr.components.FormComponent):
+ """Same as gr.Row but fits inside gradio forms"""
+
+ def get_block_name(self):
+ return "row"
diff --git a/modules/ui_tempdir.py b/modules/ui_tempdir.py index 07210d14..363d449d 100644 --- a/modules/ui_tempdir.py +++ b/modules/ui_tempdir.py @@ -15,7 +15,8 @@ Savedfile = namedtuple("Savedfile", ["name"]) def save_pil_to_file(pil_image, dir=None):
already_saved_as = getattr(pil_image, 'already_saved_as', None)
if already_saved_as and os.path.isfile(already_saved_as):
- shared.demo.temp_dirs = shared.demo.temp_dirs | {os.path.abspath(os.path.dirname(already_saved_as))}
+ shared.demo.temp_file_sets[0] = shared.demo.temp_file_sets[0] | {os.path.abspath(already_saved_as)}
+
file_obj = Savedfile(already_saved_as)
return file_obj
@@ -44,7 +45,7 @@ def on_tmpdir_changed(): os.makedirs(shared.opts.temp_dir, exist_ok=True)
- shared.demo.temp_dirs = shared.demo.temp_dirs | {os.path.abspath(shared.opts.temp_dir)}
+ shared.demo.temp_file_sets[0] = shared.demo.temp_file_sets[0] | {os.path.abspath(shared.opts.temp_dir)}
def cleanup_tmpdr():
diff --git a/modules/xlmr.py b/modules/xlmr.py new file mode 100644 index 00000000..beab3fdf --- /dev/null +++ b/modules/xlmr.py @@ -0,0 +1,137 @@ +from transformers import BertPreTrainedModel,BertModel,BertConfig +import torch.nn as nn +import torch +from transformers.models.xlm_roberta.configuration_xlm_roberta import XLMRobertaConfig +from transformers import XLMRobertaModel,XLMRobertaTokenizer +from typing import Optional + +class BertSeriesConfig(BertConfig): + def __init__(self, vocab_size=30522, hidden_size=768, num_hidden_layers=12, num_attention_heads=12, intermediate_size=3072, hidden_act="gelu", hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, type_vocab_size=2, initializer_range=0.02, layer_norm_eps=1e-12, pad_token_id=0, position_embedding_type="absolute", use_cache=True, classifier_dropout=None,project_dim=512, pooler_fn="average",learn_encoder=False,model_type='bert',**kwargs): + + super().__init__(vocab_size, hidden_size, num_hidden_layers, num_attention_heads, intermediate_size, hidden_act, hidden_dropout_prob, attention_probs_dropout_prob, max_position_embeddings, type_vocab_size, initializer_range, layer_norm_eps, pad_token_id, position_embedding_type, use_cache, classifier_dropout, **kwargs) + self.project_dim = project_dim + self.pooler_fn = pooler_fn + self.learn_encoder = learn_encoder + +class RobertaSeriesConfig(XLMRobertaConfig): + def __init__(self, pad_token_id=1, bos_token_id=0, eos_token_id=2,project_dim=512,pooler_fn='cls',learn_encoder=False, **kwargs): + super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs) + self.project_dim = project_dim + self.pooler_fn = pooler_fn + self.learn_encoder = learn_encoder + + +class BertSeriesModelWithTransformation(BertPreTrainedModel): + + _keys_to_ignore_on_load_unexpected = [r"pooler"] + _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"] + config_class = BertSeriesConfig + + def __init__(self, config=None, **kargs): + # modify initialization for autoloading + if config is None: + config = XLMRobertaConfig() + config.attention_probs_dropout_prob= 0.1 + config.bos_token_id=0 + config.eos_token_id=2 + config.hidden_act='gelu' + config.hidden_dropout_prob=0.1 + config.hidden_size=1024 + config.initializer_range=0.02 + config.intermediate_size=4096 + config.layer_norm_eps=1e-05 + config.max_position_embeddings=514 + + config.num_attention_heads=16 + config.num_hidden_layers=24 + config.output_past=True + config.pad_token_id=1 + config.position_embedding_type= "absolute" + + config.type_vocab_size= 1 + config.use_cache=True + config.vocab_size= 250002 + config.project_dim = 768 + config.learn_encoder = False + super().__init__(config) + self.roberta = XLMRobertaModel(config) + self.transformation = nn.Linear(config.hidden_size,config.project_dim) + self.pre_LN=nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) + self.tokenizer = XLMRobertaTokenizer.from_pretrained('xlm-roberta-large') + self.pooler = lambda x: x[:,0] + self.post_init() + + def encode(self,c): + device = next(self.parameters()).device + text = self.tokenizer(c, + truncation=True, + max_length=77, + return_length=False, + return_overflowing_tokens=False, + padding="max_length", + return_tensors="pt") + text["input_ids"] = torch.tensor(text["input_ids"]).to(device) + text["attention_mask"] = torch.tensor( + text['attention_mask']).to(device) + features = self(**text) + return features['projection_state'] + + def forward( + self, + input_ids: Optional[torch.Tensor] = None, + attention_mask: Optional[torch.Tensor] = None, + token_type_ids: Optional[torch.Tensor] = None, + position_ids: Optional[torch.Tensor] = None, + head_mask: Optional[torch.Tensor] = None, + inputs_embeds: Optional[torch.Tensor] = None, + encoder_hidden_states: Optional[torch.Tensor] = None, + encoder_attention_mask: Optional[torch.Tensor] = None, + output_attentions: Optional[bool] = None, + return_dict: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + ) : + r""" + """ + + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + + outputs = self.roberta( + input_ids=input_ids, + attention_mask=attention_mask, + token_type_ids=token_type_ids, + position_ids=position_ids, + head_mask=head_mask, + inputs_embeds=inputs_embeds, + encoder_hidden_states=encoder_hidden_states, + encoder_attention_mask=encoder_attention_mask, + output_attentions=output_attentions, + output_hidden_states=True, + return_dict=return_dict, + ) + + # last module outputs + sequence_output = outputs[0] + + + # project every module + sequence_output_ln = self.pre_LN(sequence_output) + + # pooler + pooler_output = self.pooler(sequence_output_ln) + pooler_output = self.transformation(pooler_output) + projection_state = self.transformation(outputs.last_hidden_state) + + return { + 'pooler_output':pooler_output, + 'last_hidden_state':outputs.last_hidden_state, + 'hidden_states':outputs.hidden_states, + 'attentions':outputs.attentions, + 'projection_state':projection_state, + 'sequence_out': sequence_output + } + + +class RobertaSeriesModelWithTransformation(BertSeriesModelWithTransformation): + base_model_prefix = 'roberta' + config_class= RobertaSeriesConfig
\ No newline at end of file diff --git a/requirements.txt b/requirements.txt index 5bed694e..e2c3876b 100644 --- a/requirements.txt +++ b/requirements.txt @@ -5,7 +5,7 @@ fairscale==0.4.4 fonts
font-roboto
gfpgan
-gradio==3.9
+gradio==3.15.0
invisible-watermark
numpy
omegaconf
diff --git a/requirements_versions.txt b/requirements_versions.txt index c126c8c4..836523ba 100644 --- a/requirements_versions.txt +++ b/requirements_versions.txt @@ -3,7 +3,7 @@ transformers==4.19.2 accelerate==0.12.0
basicsr==1.4.2
gfpgan==1.3.8
-gradio==3.9
+gradio==3.15.0
numpy==1.23.3
Pillow==9.2.0
realesrgan==0.3.0
@@ -245,11 +245,6 @@ input[type="range"]{ margin: 0.5em 0 -0.3em 0;
}
-#txt2img_sampling label{
- padding-left: 0.6em;
- padding-right: 0.6em;
-}
-
#mask_bug_info {
text-align: center;
display: block;
@@ -501,13 +496,6 @@ input[type="range"]{ padding: 0;
}
-#refresh_sd_model_checkpoint, #refresh_sd_vae, #refresh_sd_hypernetwork, #refresh_train_hypernetwork_name, #refresh_train_embedding_name, #refresh_localization{
- max-width: 2.5em;
- min-width: 2.5em;
- height: 2.4em;
-}
-
-
canvas[key="mask"] {
z-index: 12 !important;
filter: invert();
@@ -568,6 +556,33 @@ img2maskimg, #img2maskimg > .h-60, #img2maskimg > .h-60 > div, #img2maskimg > .h font-size: 95%;
}
+#image_buttons_txt2img button, #image_buttons_img2img button, #image_buttons_extras button{
+ min-width: auto;
+ padding-left: 0.5em;
+ padding-right: 0.5em;
+}
+
+.gr-form{
+ background-color: white;
+}
+
+.dark .gr-form{
+ background-color: rgb(31 41 55 / var(--tw-bg-opacity));
+}
+
+.gr-button-tool{
+ max-width: 2.5em;
+ min-width: 2.5em !important;
+ height: 2.4em;
+ margin: 0.55em 0;
+}
+
+#quicksettings .gr-button-tool{
+ margin: 0;
+}
+
+
+
/* The following handles localization for right-to-left (RTL) languages like Arabic.
The rtl media type will only be activated by the logic in javascript/localization.js.
If you change anything above, you need to make sure it is RTL compliant by just running
diff --git a/v2-inference-v.yaml b/v2-inference-v.yaml new file mode 100644 index 00000000..513cd635 --- /dev/null +++ b/v2-inference-v.yaml @@ -0,0 +1,68 @@ +model: + base_learning_rate: 1.0e-4 + target: ldm.models.diffusion.ddpm.LatentDiffusion + params: + parameterization: "v" + linear_start: 0.00085 + linear_end: 0.0120 + num_timesteps_cond: 1 + log_every_t: 200 + timesteps: 1000 + first_stage_key: "jpg" + cond_stage_key: "txt" + image_size: 64 + channels: 4 + cond_stage_trainable: false + conditioning_key: crossattn + monitor: val/loss_simple_ema + scale_factor: 0.18215 + use_ema: False # we set this to false because this is an inference only config + + unet_config: + target: ldm.modules.diffusionmodules.openaimodel.UNetModel + params: + use_checkpoint: True + use_fp16: True + image_size: 32 # unused + in_channels: 4 + out_channels: 4 + model_channels: 320 + attention_resolutions: [ 4, 2, 1 ] + num_res_blocks: 2 + channel_mult: [ 1, 2, 4, 4 ] + num_head_channels: 64 # need to fix for flash-attn + use_spatial_transformer: True + use_linear_in_transformer: True + transformer_depth: 1 + context_dim: 1024 + legacy: False + + first_stage_config: + target: ldm.models.autoencoder.AutoencoderKL + params: + embed_dim: 4 + monitor: val/rec_loss + ddconfig: + #attn_type: "vanilla-xformers" + double_z: true + z_channels: 4 + resolution: 256 + in_channels: 3 + out_ch: 3 + ch: 128 + ch_mult: + - 1 + - 2 + - 4 + - 4 + num_res_blocks: 2 + attn_resolutions: [] + dropout: 0.0 + lossconfig: + target: torch.nn.Identity + + cond_stage_config: + target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder + params: + freeze: True + layer: "penultimate"
\ No newline at end of file |