From 873efeed49bb5197a42da18272115b326c5d68f3 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 11 Oct 2022 15:51:22 +0300 Subject: rename hypernetwork dir to hypernetworks to prevent clash with an old filename that people who use zip instead of git clone will have --- modules/hypernetworks/ui.py | 43 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) create mode 100644 modules/hypernetworks/ui.py (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py new file mode 100644 index 00000000..811bc31e --- /dev/null +++ b/modules/hypernetworks/ui.py @@ -0,0 +1,43 @@ +import html +import os + +import gradio as gr + +import modules.textual_inversion.textual_inversion +import modules.textual_inversion.preprocess +from modules import sd_hijack, shared +from modules.hypernetworks import hypernetwork + + +def create_hypernetwork(name): + 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.hypernetwork.hypernetwork.Hypernetwork(name=name) + hypernet.save(fn) + + shared.reload_hypernetworks() + + return gr.Dropdown.update(choices=sorted([x for x in shared.hypernetworks.keys()])), f"Created: {fn}", "" + + +def train_hypernetwork(*args): + + initial_hypernetwork = shared.loaded_hypernetwork + + try: + sd_hijack.undo_optimizations() + + hypernetwork, filename = modules.hypernetwork.hypernetwork.train_hypernetwork(*args) + + res = f""" +Training {'interrupted' if shared.state.interrupted else 'finished'} at {hypernetwork.step} steps. +Hypernetwork saved to {html.escape(filename)} +""" + return res, "" + except Exception: + raise + finally: + shared.loaded_hypernetwork = initial_hypernetwork + sd_hijack.apply_optimizations() + -- cgit v1.2.3 From b0583be0884cd17dafb408fd79b52b2a0a972563 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 11 Oct 2022 15:54:34 +0300 Subject: more renames --- modules/hypernetworks/ui.py | 4 ++-- modules/ui.py | 4 ++-- webui.py | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 811bc31e..e7540f41 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -13,7 +13,7 @@ def create_hypernetwork(name): 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.hypernetwork.hypernetwork.Hypernetwork(name=name) + hypernet = modules.hypernetworks.hypernetwork.Hypernetwork(name=name) hypernet.save(fn) shared.reload_hypernetworks() @@ -28,7 +28,7 @@ def train_hypernetwork(*args): try: sd_hijack.undo_optimizations() - hypernetwork, filename = modules.hypernetwork.hypernetwork.train_hypernetwork(*args) + hypernetwork, filename = modules.hypernetworks.hypernetwork.train_hypernetwork(*args) res = f""" Training {'interrupted' if shared.state.interrupted else 'finished'} at {hypernetwork.step} steps. diff --git a/modules/ui.py b/modules/ui.py index 42e5d866..ee333c3b 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1111,7 +1111,7 @@ def create_ui(wrap_gradio_gpu_call): ) create_hypernetwork.click( - fn=modules.hypernetwork.ui.create_hypernetwork, + fn=modules.hypernetworks.ui.create_hypernetwork, inputs=[ new_hypernetwork_name, ], @@ -1164,7 +1164,7 @@ def create_ui(wrap_gradio_gpu_call): ) train_hypernetwork.click( - fn=wrap_gradio_gpu_call(modules.hypernetwork.ui.train_hypernetwork, extra_outputs=[gr.update()]), + fn=wrap_gradio_gpu_call(modules.hypernetworks.ui.train_hypernetwork, extra_outputs=[gr.update()]), _js="start_training_textual_inversion", inputs=[ train_hypernetwork_name, diff --git a/webui.py b/webui.py index faa38a0d..338f58e1 100644 --- a/webui.py +++ b/webui.py @@ -83,7 +83,7 @@ modules.scripts.load_scripts(os.path.join(script_path, "scripts")) shared.sd_model = modules.sd_models.load_model() shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights(shared.sd_model))) -shared.opts.onchange("sd_hypernetwork", wrap_queued_call(lambda: modules.hypernetwork.hypernetwork.load_hypernetwork(shared.opts.sd_hypernetwork))) +shared.opts.onchange("sd_hypernetwork", wrap_queued_call(lambda: modules.hypernetworks.hypernetwork.load_hypernetwork(shared.opts.sd_hypernetwork))) def webui(): -- cgit v1.2.3 From d682444ecc99319fbd2b142a12727501e2884ba7 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 11 Oct 2022 18:04:47 +0300 Subject: add option to select hypernetwork modules when creating --- modules/hypernetworks/hypernetwork.py | 4 ++-- modules/hypernetworks/ui.py | 4 ++-- modules/ui.py | 2 ++ 3 files changed, 6 insertions(+), 4 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index aa701bda..b081f14e 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -42,7 +42,7 @@ class Hypernetwork: filename = None name = None - def __init__(self, name=None): + def __init__(self, name=None, enable_sizes=None): self.filename = None self.name = name self.layers = {} @@ -50,7 +50,7 @@ class Hypernetwork: self.sd_checkpoint = None self.sd_checkpoint_name = None - for size in [320, 640, 768, 1280]: + for size in enable_sizes or [320, 640, 768, 1280]: self.layers[size] = (HypernetworkModule(size), HypernetworkModule(size)) def weights(self): diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index e7540f41..cdddcce1 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -9,11 +9,11 @@ from modules import sd_hijack, shared from modules.hypernetworks import hypernetwork -def create_hypernetwork(name): +def create_hypernetwork(name, enable_sizes): 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) + hypernet = modules.hypernetworks.hypernetwork.Hypernetwork(name=name, enable_sizes=[int(x) for x in enable_sizes]) hypernet.save(fn) shared.reload_hypernetworks() diff --git a/modules/ui.py b/modules/ui.py index f2d16b12..14b87b92 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1037,6 +1037,7 @@ def create_ui(wrap_gradio_gpu_call): gr.HTML(value="

Create a new 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"]) with gr.Row(): with gr.Column(scale=3): @@ -1114,6 +1115,7 @@ def create_ui(wrap_gradio_gpu_call): fn=modules.hypernetworks.ui.create_hypernetwork, inputs=[ new_hypernetwork_name, + new_hypernetwork_sizes, ], outputs=[ train_hypernetwork_name, -- cgit v1.2.3 From 6d09b8d1df3a96e1380bb1650f5961781630af96 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 11 Oct 2022 18:33:57 +0300 Subject: produce error when training with medvram/lowvram enabled --- modules/hypernetworks/ui.py | 2 ++ modules/textual_inversion/ui.py | 3 +++ 2 files changed, 5 insertions(+) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index cdddcce1..3541a388 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -25,6 +25,8 @@ def train_hypernetwork(*args): initial_hypernetwork = shared.loaded_hypernetwork + assert not shared.cmd_opts.lowvram and not shared.cmd_opts.medvram, 'Training models with lowvram or medvram is not possible' + try: sd_hijack.undo_optimizations() diff --git a/modules/textual_inversion/ui.py b/modules/textual_inversion/ui.py index c57de1f9..70f47343 100644 --- a/modules/textual_inversion/ui.py +++ b/modules/textual_inversion/ui.py @@ -22,6 +22,9 @@ def preprocess(*args): def train_embedding(*args): + + assert not shared.cmd_opts.lowvram and not shared.cmd_opts.medvram, 'Training models with lowvram or medvram is not possible' + try: sd_hijack.undo_optimizations() -- cgit v1.2.3 From d4ea5f4d8631f778d11efcde397e4a5b8801d43b Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 11 Oct 2022 19:03:08 +0300 Subject: add an option to unload models during hypernetwork training to save VRAM --- modules/hypernetworks/hypernetwork.py | 25 +++++++++++++++------- modules/hypernetworks/ui.py | 4 +++- modules/shared.py | 4 ++++ modules/textual_inversion/dataset.py | 29 ++++++++++++++++++-------- modules/textual_inversion/textual_inversion.py | 2 +- 5 files changed, 46 insertions(+), 18 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index b081f14e..4700e1ec 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -175,6 +175,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, filename = os.path.join(shared.cmd_opts.hypernetwork_dir, f'{hypernetwork_name}.pt') log_directory = os.path.join(log_directory, datetime.datetime.now().strftime("%Y-%m-%d"), hypernetwork_name) + unload = shared.opts.unload_models_when_training if save_hypernetwork_every > 0: hypernetwork_dir = os.path.join(log_directory, "hypernetworks") @@ -188,11 +189,13 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, else: images_dir = None - cond_model = shared.sd_model.cond_stage_model - 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=1, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file) + ds = modules.textual_inversion.dataset.PersonalizedBase(data_root=data_root, width=512, height=512, repeats=1, placeholder_token=hypernetwork_name, model=shared.sd_model, device=devices.device, template_file=template_file, include_cond=True) + + if unload: + shared.sd_model.cond_stage_model.to(devices.cpu) + shared.sd_model.first_stage_model.to(devices.cpu) hypernetwork = shared.loaded_hypernetwork weights = hypernetwork.weights() @@ -211,7 +214,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, return hypernetwork, filename pbar = tqdm.tqdm(enumerate(ds), total=steps - ititial_step) - for i, (x, text) in pbar: + for i, (x, text, cond) in pbar: hypernetwork.step = i + ititial_step if hypernetwork.step > steps: @@ -221,11 +224,11 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, break with torch.autocast("cuda"): - c = cond_model([text]) - + cond = cond.to(devices.device) x = x.to(devices.device) - loss = shared.sd_model(x.unsqueeze(0), c)[0] + loss = shared.sd_model(x.unsqueeze(0), cond)[0] del x + del cond losses[hypernetwork.step % losses.shape[0]] = loss.item() @@ -244,6 +247,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, preview_text = text if preview_image_prompt == "" else preview_image_prompt + optimizer.zero_grad() + shared.sd_model.cond_stage_model.to(devices.device) + shared.sd_model.first_stage_model.to(devices.device) + p = processing.StableDiffusionProcessingTxt2Img( sd_model=shared.sd_model, prompt=preview_text, @@ -255,6 +262,10 @@ def train_hypernetwork(hypernetwork_name, learn_rate, data_root, log_directory, processed = processing.process_images(p) image = processed.images[0] + if unload: + shared.sd_model.cond_stage_model.to(devices.cpu) + shared.sd_model.first_stage_model.to(devices.cpu) + shared.state.current_image = image image.save(last_saved_image) diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 3541a388..c67facbb 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -5,7 +5,7 @@ import gradio as gr import modules.textual_inversion.textual_inversion import modules.textual_inversion.preprocess -from modules import sd_hijack, shared +from modules import sd_hijack, shared, devices from modules.hypernetworks import hypernetwork @@ -41,5 +41,7 @@ Hypernetwork saved to {html.escape(filename)} raise finally: shared.loaded_hypernetwork = initial_hypernetwork + shared.sd_model.cond_stage_model.to(devices.device) + shared.sd_model.first_stage_model.to(devices.device) sd_hijack.apply_optimizations() diff --git a/modules/shared.py b/modules/shared.py index 20b45f23..c1092ff7 100644 --- a/modules/shared.py +++ b/modules/shared.py @@ -228,6 +228,10 @@ options_templates.update(options_section(('system', "System"), { "multiple_tqdm": OptionInfo(True, "Add a second progress bar to the console that shows progress for an entire job."), })) +options_templates.update(options_section(('training', "Training"), { + "unload_models_when_training": OptionInfo(False, "Unload VAE and CLIP form VRAM when training"), +})) + 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()}, show_on_main_page=True), "sd_hypernetwork": OptionInfo("None", "Stable Diffusion finetune hypernetwork", gr.Dropdown, lambda: {"choices": ["None"] + [x for x in hypernetworks.keys()]}), diff --git a/modules/textual_inversion/dataset.py b/modules/textual_inversion/dataset.py index 4d006366..f61f40d3 100644 --- a/modules/textual_inversion/dataset.py +++ b/modules/textual_inversion/dataset.py @@ -8,14 +8,14 @@ from torchvision import transforms import random import tqdm -from modules import devices +from modules import devices, shared import re re_tag = re.compile(r"[a-zA-Z][_\w\d()]+") class PersonalizedBase(Dataset): - def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None): + def __init__(self, data_root, width, height, repeats, flip_p=0.5, placeholder_token="*", model=None, device=None, template_file=None, include_cond=False): self.placeholder_token = placeholder_token @@ -32,6 +32,8 @@ class PersonalizedBase(Dataset): assert data_root, 'dataset directory not specified' + cond_model = shared.sd_model.cond_stage_model + self.image_paths = [os.path.join(data_root, file_path) for file_path in os.listdir(data_root)] print("Preparing dataset...") for path in tqdm.tqdm(self.image_paths): @@ -53,7 +55,13 @@ class PersonalizedBase(Dataset): init_latent = model.get_first_stage_encoding(model.encode_first_stage(torchdata.unsqueeze(dim=0))).squeeze() init_latent = init_latent.to(devices.cpu) - self.dataset.append((init_latent, filename_tokens)) + if include_cond: + text = self.create_text(filename_tokens) + cond = cond_model([text]).to(devices.cpu) + else: + cond = None + + self.dataset.append((init_latent, filename_tokens, cond)) self.length = len(self.dataset) * repeats @@ -64,6 +72,12 @@ class PersonalizedBase(Dataset): def shuffle(self): self.indexes = self.initial_indexes[torch.randperm(self.initial_indexes.shape[0])] + def create_text(self, filename_tokens): + text = random.choice(self.lines) + text = text.replace("[name]", self.placeholder_token) + text = text.replace("[filewords]", ' '.join(filename_tokens)) + return text + def __len__(self): return self.length @@ -72,10 +86,7 @@ class PersonalizedBase(Dataset): self.shuffle() index = self.indexes[i % len(self.indexes)] - x, filename_tokens = self.dataset[index] - - text = random.choice(self.lines) - text = text.replace("[name]", self.placeholder_token) - text = text.replace("[filewords]", ' '.join(filename_tokens)) + x, filename_tokens, cond = self.dataset[index] - return x, text + text = self.create_text(filename_tokens) + return x, text, cond diff --git a/modules/textual_inversion/textual_inversion.py b/modules/textual_inversion/textual_inversion.py index bb05cdc6..35f4bd9e 100644 --- a/modules/textual_inversion/textual_inversion.py +++ b/modules/textual_inversion/textual_inversion.py @@ -201,7 +201,7 @@ def train_embedding(embedding_name, learn_rate, data_root, log_directory, traini return embedding, filename pbar = tqdm.tqdm(enumerate(ds), total=steps-ititial_step) - for i, (x, text) in pbar: + for i, (x, text, _) in pbar: embedding.step = i + ititial_step if embedding.step > steps: -- cgit v1.2.3 From 6be32b31d181e42c639dad3451229aa7b9cfd1cf Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Tue, 11 Oct 2022 23:07:09 +0300 Subject: reports that training with medvram is possible. --- modules/hypernetworks/ui.py | 2 +- modules/textual_inversion/ui.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index c67facbb..dfa599af 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -25,7 +25,7 @@ def train_hypernetwork(*args): initial_hypernetwork = shared.loaded_hypernetwork - assert not shared.cmd_opts.lowvram and not shared.cmd_opts.medvram, 'Training models with lowvram or medvram is not possible' + assert not shared.cmd_opts.lowvram, 'Training models with lowvram is not possible' try: sd_hijack.undo_optimizations() diff --git a/modules/textual_inversion/ui.py b/modules/textual_inversion/ui.py index 70f47343..36881e7a 100644 --- a/modules/textual_inversion/ui.py +++ b/modules/textual_inversion/ui.py @@ -23,7 +23,7 @@ def preprocess(*args): def train_embedding(*args): - assert not shared.cmd_opts.lowvram and not shared.cmd_opts.medvram, 'Training models with lowvram or medvram is not possible' + assert not shared.cmd_opts.lowvram, 'Training models with lowvram not possible' try: sd_hijack.undo_optimizations() -- 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 +++-- webui.py | 8 ++--- 5 files changed, 48 insertions(+), 43 deletions(-) (limited to 'modules/hypernetworks/ui.py') 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, diff --git a/webui.py b/webui.py index c7260c7a..177bef74 100644 --- a/webui.py +++ b/webui.py @@ -85,9 +85,7 @@ def initialize(): shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights(shared.sd_model))) shared.opts.onchange("sd_hypernetwork", wrap_queued_call(lambda: modules.hypernetworks.hypernetwork.load_hypernetwork(shared.opts.sd_hypernetwork))) shared.opts.onchange("sd_hypernetwork_strength", modules.hypernetworks.hypernetwork.apply_strength) - shared.opts.onchange("sd_hypernetwork_layer_structure", modules.hypernetworks.hypernetwork.apply_layer_structure) - shared.opts.onchange("sd_hypernetwork_add_layer_norm", modules.hypernetworks.hypernetwork.apply_layer_norm) - + # make the program just exit at ctrl+c without waiting for anything def sigint_handler(sig, frame): print(f'Interrupted with signal {sig} in {frame}') @@ -142,7 +140,7 @@ def webui(launch_api=False): create_api(app) wait_on_server(demo) - + sd_samplers.set_samplers() print('Reloading Custom Scripts') @@ -160,4 +158,4 @@ if __name__ == "__main__": if cmd_opts.nowebui: api_only() else: - webui(cmd_opts.api) \ No newline at end of file + webui(cmd_opts.api) -- 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/hypernetworks/ui.py') 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 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/hypernetworks/ui.py') 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 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/hypernetworks/ui.py') 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 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/hypernetworks/ui.py') 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 51e3dc9ccad157d7161b697a246e26c868d46a7c Mon Sep 17 00:00:00 2001 From: timntorres Date: Fri, 21 Oct 2022 02:11:12 -0700 Subject: Sanitize hypernet name input. --- modules/hypernetworks/ui.py | 3 +++ 1 file changed, 3 insertions(+) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 266f04f6..e6f50a1f 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -11,6 +11,9 @@ from modules.hypernetworks import hypernetwork def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, add_layer_norm=False, activation_func=None): + # Remove illegal characters from name. + name = "".join( x for x in name if (x.isalnum() or x in "._- ")) + fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") if not overwrite_old: assert not os.path.exists(fn), f"file {fn} already exists" -- cgit v1.2.3 From 0e8ca8e7af05be22d7d2c07a47c3c7febe0f0ab6 Mon Sep 17 00:00:00 2001 From: discus0434 Date: Sat, 22 Oct 2022 11:07:00 +0000 Subject: add dropout --- modules/hypernetworks/hypernetwork.py | 68 +++++++++++++++++++++-------------- modules/hypernetworks/ui.py | 10 +++--- modules/ui.py | 43 +++++++++++----------- 3 files changed, 70 insertions(+), 51 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index 905cbeef..e493f366 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -1,47 +1,60 @@ +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 class HypernetworkModule(torch.nn.Module): multiplier = 1.0 - activation_dict = {"relu": torch.nn.ReLU, "leakyrelu": torch.nn.LeakyReLU, "elu": torch.nn.ELU, - "swish": torch.nn.Hardswish} - - def __init__(self, dim, state_dict=None, layer_structure=None, add_layer_norm=False, activation_func=None): + activation_dict = { + "relu": torch.nn.ReLU, + "leakyrelu": torch.nn.LeakyReLU, + "elu": torch.nn.ELU, + "swish": torch.nn.Hardswish, + } + + def __init__(self, dim, state_dict=None, layer_structure=None, activation_func=None, add_layer_norm=False, use_dropout=False): super().__init__() 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!" - + assert activation_func not in self.activation_dict.keys() + "linear", f"Valid activation funcs: 'linear', 'relu', 'leakyrelu', 'elu', 'swish'" + linears = [] for i in range(len(layer_structure) - 1): + + # Add a fully-connected layer linears.append(torch.nn.Linear(int(dim * layer_structure[i]), int(dim * layer_structure[i+1]))) - # if skip_first_layer because first parameters potentially contain negative values - # if i < 1: continue - if activation_func in HypernetworkModule.activation_dict: - linears.append(HypernetworkModule.activation_dict[activation_func]()) + + # Add an activation func + if activation_func == "linear": + pass + elif activation_func in self.activation_dict: + linears.append(self.activation_dict[activation_func]()) else: - print("Invalid key {} encountered as activation function!".format(activation_func)) - # if use_dropout: - # linears.append(torch.nn.Dropout(p=0.3)) + raise NotImplementedError( + "Valid activation funcs: 'linear', 'relu', 'leakyrelu', 'elu', 'swish'" + ) + + # Add dropout + if use_dropout: + linears.append(torch.nn.Dropout(p=0.3)) + + # Add layer normalization if add_layer_norm: linears.append(torch.nn.LayerNorm(int(dim * layer_structure[i+1]))) @@ -93,7 +106,7 @@ class Hypernetwork: filename = None name = None - def __init__(self, name=None, enable_sizes=None, layer_structure=None, add_layer_norm=False, activation_func=None): + def __init__(self, name=None, enable_sizes=None, layer_structure=None, activation_func=None, add_layer_norm=False, use_dropout=False): self.filename = None self.name = name self.layers = {} @@ -101,13 +114,14 @@ class Hypernetwork: self.sd_checkpoint = None self.sd_checkpoint_name = None self.layer_structure = layer_structure - self.add_layer_norm = add_layer_norm self.activation_func = activation_func + self.add_layer_norm = add_layer_norm + self.use_dropout = use_dropout for size in enable_sizes or []: self.layers[size] = ( - 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), + HypernetworkModule(size, None, self.layer_structure, self.activation_func, self.add_layer_norm, self.use_dropout), + HypernetworkModule(size, None, self.layer_structure, self.activation_func, self.add_layer_norm, self.use_dropout), ) def weights(self): @@ -129,8 +143,9 @@ 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['activation_func'] = self.activation_func + state_dict['is_layer_norm'] = self.add_layer_norm + state_dict['use_dropout'] = self.use_dropout state_dict['sd_checkpoint'] = self.sd_checkpoint state_dict['sd_checkpoint_name'] = self.sd_checkpoint_name @@ -144,8 +159,9 @@ 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) self.activation_func = state_dict.get('activation_func', None) + self.add_layer_norm = state_dict.get('is_layer_norm', False) + self.use_dropout = state_dict.get('use_dropout', False) for size, sd in state_dict.items(): if type(size) == int: diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 1a5a27d8..5f6f17b6 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -3,14 +3,13 @@ import os import re import gradio as gr - -import modules.textual_inversion.textual_inversion import modules.textual_inversion.preprocess -from modules import sd_hijack, shared, devices +import modules.textual_inversion.textual_inversion +from modules import devices, sd_hijack, shared from modules.hypernetworks import hypernetwork -def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm=False, activation_func=None): +def create_hypernetwork(name, enable_sizes, layer_structure=None, activation_func=None, add_layer_norm=False, use_dropout=False): fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") assert not os.path.exists(fn), f"file {fn} already exists" @@ -21,8 +20,9 @@ def create_hypernetwork(name, enable_sizes, layer_structure=None, add_layer_norm name=name, enable_sizes=[int(x) for x in enable_sizes], layer_structure=layer_structure, - add_layer_norm=add_layer_norm, activation_func=activation_func, + add_layer_norm=add_layer_norm, + use_dropout=use_dropout, ) hypernet.save(fn) diff --git a/modules/ui.py b/modules/ui.py index 716f14b8..d4b32c05 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() @@ -1223,8 +1224,9 @@ def create_ui(wrap_gradio_gpu_call): 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.Textbox("1, 2, 1", label="Enter hypernetwork layer structure", placeholder="1st and last digit must be 1. ex:'1, 2, 1'") + new_hypernetwork_activation_func = gr.Dropdown(value="relu", label="Select activation function of hypernetwork", choices=["linear", "relu", "leakyrelu", "elu", "swish"]) 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=["linear", "relu", "leakyrelu"]) + new_hypernetwork_use_dropout = gr.Checkbox(label="Use dropout") with gr.Row(): with gr.Column(scale=3): @@ -1308,8 +1310,9 @@ def create_ui(wrap_gradio_gpu_call): new_hypernetwork_name, new_hypernetwork_sizes, new_hypernetwork_layer_structure, - new_hypernetwork_add_layer_norm, new_hypernetwork_activation_func, + new_hypernetwork_add_layer_norm, + new_hypernetwork_use_dropout ], outputs=[ train_hypernetwork_name, -- cgit v1.2.3 From de096d0ce752c96e45508dcc7b9e84f7dbe10cca Mon Sep 17 00:00:00 2001 From: AngelBottomless <35677394+aria1th@users.noreply.github.com> Date: Tue, 25 Oct 2022 14:48:49 +0900 Subject: Weight initialization and More activation func add weight init add weight init option in create_hypernetwork fstringify hypernet info save weight initialization info for further debugging fill bias with zero for He/Xavier initialize LayerNorm with Normal fix loading weight_init --- modules/hypernetworks/hypernetwork.py | 47 ++++++++++++++++++++++++++++------- modules/hypernetworks/ui.py | 4 ++- modules/ui.py | 4 ++- 3 files changed, 44 insertions(+), 11 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index d647ea55..afbcdff8 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -5,6 +5,7 @@ import html import os import sys import traceback +import inspect import modules.textual_inversion.dataset import torch @@ -15,10 +16,12 @@ 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 +from torch.nn.init import normal_, xavier_normal_, xavier_uniform_, kaiming_normal_, kaiming_uniform_, zeros_ from collections import defaultdict, deque from statistics import stdev, mean + class HypernetworkModule(torch.nn.Module): multiplier = 1.0 activation_dict = { @@ -26,9 +29,12 @@ class HypernetworkModule(torch.nn.Module): "leakyrelu": torch.nn.LeakyReLU, "elu": torch.nn.ELU, "swish": torch.nn.Hardswish, + "tanh": torch.nn.Tanh, + "sigmoid": torch.nn.Sigmoid, } + activation_dict.update({cls_name: cls_obj for cls_name, cls_obj in inspect.getmembers(torch.nn.modules.activation) if inspect.isclass(cls_obj) and cls_obj.__module__ == 'torch.nn.modules.activation'}) - def __init__(self, dim, state_dict=None, layer_structure=None, activation_func=None, add_layer_norm=False, use_dropout=False): + def __init__(self, dim, state_dict=None, layer_structure=None, activation_func=None, weight_init='Normal', add_layer_norm=False, use_dropout=False): super().__init__() assert layer_structure is not None, "layer_structure must not be None" @@ -65,9 +71,24 @@ class HypernetworkModule(torch.nn.Module): else: for layer in self.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_() - + w, b = layer.weight.data, layer.bias.data + if weight_init == "Normal" or type(layer) == torch.nn.LayerNorm: + normal_(w, mean=0.0, std=0.01) + normal_(b, mean=0.0, std=0.005) + elif weight_init == 'XavierUniform': + xavier_uniform_(w) + zeros_(b) + elif weight_init == 'XavierNormal': + xavier_normal_(w) + zeros_(b) + elif weight_init == 'KaimingUniform': + kaiming_uniform_(w, nonlinearity='leaky_relu' if 'leakyrelu' == activation_func else 'relu') + zeros_(b) + elif weight_init == 'KaimingNormal': + kaiming_normal_(w, nonlinearity='leaky_relu' if 'leakyrelu' == activation_func else 'relu') + zeros_(b) + else: + raise KeyError(f"Key {weight_init} is not defined as initialization!") self.to(devices.device) def fix_old_state_dict(self, state_dict): @@ -105,7 +126,7 @@ class Hypernetwork: filename = None name = None - def __init__(self, name=None, enable_sizes=None, layer_structure=None, activation_func=None, add_layer_norm=False, use_dropout=False): + def __init__(self, name=None, enable_sizes=None, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): self.filename = None self.name = name self.layers = {} @@ -114,13 +135,14 @@ class Hypernetwork: self.sd_checkpoint_name = None self.layer_structure = layer_structure self.activation_func = activation_func + self.weight_init = weight_init self.add_layer_norm = add_layer_norm self.use_dropout = use_dropout for size in enable_sizes or []: self.layers[size] = ( - HypernetworkModule(size, None, self.layer_structure, self.activation_func, self.add_layer_norm, self.use_dropout), - HypernetworkModule(size, None, self.layer_structure, self.activation_func, self.add_layer_norm, self.use_dropout), + HypernetworkModule(size, None, self.layer_structure, self.activation_func, self.weight_init, self.add_layer_norm, self.use_dropout), + HypernetworkModule(size, None, self.layer_structure, self.activation_func, self.weight_init, self.add_layer_norm, self.use_dropout), ) def weights(self): @@ -144,6 +166,7 @@ class Hypernetwork: state_dict['layer_structure'] = self.layer_structure state_dict['activation_func'] = self.activation_func state_dict['is_layer_norm'] = self.add_layer_norm + state_dict['weight_initialization'] = self.weight_init state_dict['use_dropout'] = self.use_dropout state_dict['sd_checkpoint'] = self.sd_checkpoint state_dict['sd_checkpoint_name'] = self.sd_checkpoint_name @@ -158,15 +181,21 @@ class Hypernetwork: state_dict = torch.load(filename, map_location='cpu') self.layer_structure = state_dict.get('layer_structure', [1, 2, 1]) + print(self.layer_structure) self.activation_func = state_dict.get('activation_func', None) + print(f"Activation function is {self.activation_func}") + self.weight_init = state_dict.get('weight_initialization', 'Normal') + print(f"Weight initialization is {self.weight_init}") self.add_layer_norm = state_dict.get('is_layer_norm', False) + print(f"Layer norm is set to {self.add_layer_norm}") self.use_dropout = state_dict.get('use_dropout', False) + print(f"Dropout usage is set to {self.use_dropout}" ) for size, sd in state_dict.items(): if type(size) == int: self.layers[size] = ( - HypernetworkModule(size, sd[0], self.layer_structure, self.activation_func, self.add_layer_norm, self.use_dropout), - HypernetworkModule(size, sd[1], self.layer_structure, self.activation_func, self.add_layer_norm, self.use_dropout), + HypernetworkModule(size, sd[0], self.layer_structure, self.activation_func, self.weight_init, self.add_layer_norm, self.use_dropout), + HypernetworkModule(size, sd[1], self.layer_structure, self.activation_func, self.weight_init, self.add_layer_norm, self.use_dropout), ) self.name = state_dict.get('name', self.name) diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 2b472d87..2c6c0470 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -8,8 +8,9 @@ import modules.textual_inversion.textual_inversion from modules import devices, sd_hijack, shared from modules.hypernetworks import hypernetwork +keys = list(hypernetwork.HypernetworkModule.activation_dict.keys()) -def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, add_layer_norm=False, use_dropout=False): +def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): # Remove illegal characters from name. name = "".join( x for x in name if (x.isalnum() or x in "._- ")) @@ -25,6 +26,7 @@ def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, enable_sizes=[int(x) for x in enable_sizes], layer_structure=layer_structure, activation_func=activation_func, + weight_init=weight_init, add_layer_norm=add_layer_norm, use_dropout=use_dropout, ) diff --git a/modules/ui.py b/modules/ui.py index 03528968..8e343258 100644 --- a/modules/ui.py +++ b/modules/ui.py @@ -1238,7 +1238,8 @@ def create_ui(wrap_gradio_gpu_call): 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.Textbox("1, 2, 1", label="Enter hypernetwork layer structure", placeholder="1st and last digit must be 1. ex:'1, 2, 1'") - new_hypernetwork_activation_func = gr.Dropdown(value="relu", label="Select activation function of hypernetwork", choices=["linear", "relu", "leakyrelu", "elu", "swish"]) + new_hypernetwork_activation_func = gr.Dropdown(value="relu", label="Select activation function of hypernetwork", choices=modules.hypernetworks.ui.keys) + new_hypernetwork_initialization_option = gr.Dropdown(value = "Normal", label="Select Layer weights initialization. relu-like - Kaiming, sigmoid-like - Xavier is recommended", choices=["Normal", "KaimingUniform", "KaimingNormal", "XavierUniform", "XavierNormal"]) new_hypernetwork_add_layer_norm = gr.Checkbox(label="Add layer normalization") new_hypernetwork_use_dropout = gr.Checkbox(label="Use dropout") overwrite_old_hypernetwork = gr.Checkbox(value=False, label="Overwrite Old Hypernetwork") @@ -1342,6 +1343,7 @@ def create_ui(wrap_gradio_gpu_call): overwrite_old_hypernetwork, new_hypernetwork_layer_structure, new_hypernetwork_activation_func, + new_hypernetwork_initialization_option, new_hypernetwork_add_layer_norm, new_hypernetwork_use_dropout ], -- cgit v1.2.3 From 462e6ba6675bd14c0f82e465423a0eedfff82372 Mon Sep 17 00:00:00 2001 From: AngelBottomless <35677394+aria1th@users.noreply.github.com> Date: Thu, 27 Oct 2022 15:40:24 +0900 Subject: Disable unavailable or duplicate options --- modules/hypernetworks/ui.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index 2c6c0470..c2d4b51c 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -8,7 +8,8 @@ import modules.textual_inversion.textual_inversion from modules import devices, sd_hijack, shared from modules.hypernetworks import hypernetwork -keys = list(hypernetwork.HypernetworkModule.activation_dict.keys()) +not_available = ["hardswish", "multiheadattention"] +keys = list(x for x in hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): # Remove illegal characters from name. -- cgit v1.2.3 From f361e804ebaa5af4a10711ece2522869fb64a4c6 Mon Sep 17 00:00:00 2001 From: AngelBottomless <35677394+aria1th@users.noreply.github.com> Date: Sat, 29 Oct 2022 08:36:50 +0900 Subject: Re enable linear --- modules/hypernetworks/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index c2d4b51c..aad09ffc 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -9,7 +9,7 @@ from modules import devices, sd_hijack, shared from modules.hypernetworks import hypernetwork not_available = ["hardswish", "multiheadattention"] -keys = list(x for x in hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) +keys = ["linear"] + list(x for x in hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): # Remove illegal characters from name. -- cgit v1.2.3 From 20194fd9752a280306fb66b57b258609b0918c46 Mon Sep 17 00:00:00 2001 From: AngelBottomless <35677394+aria1th@users.noreply.github.com> Date: Sat, 29 Oct 2022 16:56:42 +0900 Subject: We have duplicate linear now --- modules/hypernetworks/ui.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index aad09ffc..c2d4b51c 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -9,7 +9,7 @@ from modules import devices, sd_hijack, shared from modules.hypernetworks import hypernetwork not_available = ["hardswish", "multiheadattention"] -keys = ["linear"] + list(x for x in hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) +keys = list(x for x in hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): # Remove illegal characters from name. -- cgit v1.2.3 From 5f1dfbbc959855fd90ba80c0c76301d2063772fa Mon Sep 17 00:00:00 2001 From: Vladimir Mandic Date: Sat, 24 Dec 2022 18:02:22 -0500 Subject: implement train api --- modules/api/api.py | 94 ++++++++++++++++++++++++++++++++++- modules/api/models.py | 9 ++++ modules/hypernetworks/hypernetwork.py | 26 ++++++++++ modules/hypernetworks/ui.py | 31 ++---------- 4 files changed, 132 insertions(+), 28 deletions(-) (limited to 'modules/hypernetworks/ui.py') diff --git a/modules/api/api.py b/modules/api/api.py index b43dd16b..1ceba75d 100644 --- a/modules/api/api.py +++ b/modules/api/api.py @@ -10,13 +10,17 @@ from fastapi.security import HTTPBasic, HTTPBasicCredentials from secrets import compare_digest import modules.shared as shared -from modules import sd_samplers, deepbooru +from modules import sd_samplers, deepbooru, sd_hijack from modules.api.models import * from modules.processing import StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img, process_images from modules.extras import run_extras, run_pnginfo +from modules.textual_inversion.textual_inversion import create_embedding, train_embedding +from modules.textual_inversion.preprocess import preprocess +from modules.hypernetworks.hypernetwork import create_hypernetwork, train_hypernetwork from PIL import PngImagePlugin,Image from modules.sd_models import checkpoints_list from modules.realesrgan_model import get_realesrgan_models +from modules import devices from typing import List def upscaler_to_index(name: str): @@ -97,6 +101,11 @@ class Api: self.add_api_route("/sdapi/v1/artist-categories", self.get_artists_categories, methods=["GET"], response_model=List[str]) self.add_api_route("/sdapi/v1/artists", self.get_artists, methods=["GET"], response_model=List[ArtistItem]) self.add_api_route("/sdapi/v1/refresh-checkpoints", self.refresh_checkpoints, methods=["POST"]) + self.add_api_route("/sdapi/v1/create/embedding", self.create_embedding, methods=["POST"], response_model=CreateResponse) + self.add_api_route("/sdapi/v1/create/hypernetwork", self.create_hypernetwork, methods=["POST"], response_model=CreateResponse) + self.add_api_route("/sdapi/v1/preprocess", self.preprocess, methods=["POST"], response_model=PreprocessResponse) + self.add_api_route("/sdapi/v1/train/embedding", self.train_embedding, methods=["POST"], response_model=TrainResponse) + self.add_api_route("/sdapi/v1/train/hypernetwork", self.train_hypernetwork, methods=["POST"], response_model=TrainResponse) def add_api_route(self, path: str, endpoint, **kwargs): if shared.cmd_opts.api_auth: @@ -326,6 +335,89 @@ class Api: def refresh_checkpoints(self): shared.refresh_checkpoints() + def create_embedding(self, args: dict): + try: + shared.state.begin() + filename = create_embedding(**args) # create empty embedding + sd_hijack.model_hijack.embedding_db.load_textual_inversion_embeddings() # reload embeddings so new one can be immediately used + shared.state.end() + return CreateResponse(info = "create embedding filename: {filename}".format(filename = filename)) + except AssertionError as e: + shared.state.end() + return TrainResponse(info = "create embedding error: {error}".format(error = e)) + + def create_hypernetwork(self, args: dict): + try: + shared.state.begin() + filename = create_hypernetwork(**args) # create empty embedding + shared.state.end() + return CreateResponse(info = "create hypernetwork filename: {filename}".format(filename = filename)) + except AssertionError as e: + shared.state.end() + return TrainResponse(info = "create hypernetwork error: {error}".format(error = e)) + + def preprocess(self, args: dict): + try: + shared.state.begin() + preprocess(**args) # quick operation unless blip/booru interrogation is enabled + shared.state.end() + return PreprocessResponse(info = 'preprocess complete') + except KeyError as e: + shared.state.end() + return PreprocessResponse(info = "preprocess error: invalid token: {error}".format(error = e)) + except AssertionError as e: + shared.state.end() + return PreprocessResponse(info = "preprocess error: {error}".format(error = e)) + except FileNotFoundError as e: + shared.state.end() + return PreprocessResponse(info = 'preprocess error: {error}'.format(error = e)) + + def train_embedding(self, args: dict): + try: + shared.state.begin() + apply_optimizations = shared.opts.training_xattention_optimizations + error = None + filename = '' + if not apply_optimizations: + sd_hijack.undo_optimizations() + try: + embedding, filename = train_embedding(**args) # can take a long time to complete + except Exception as e: + error = e + finally: + if not apply_optimizations: + sd_hijack.apply_optimizations() + shared.state.end() + return TrainResponse(info = "train embedding complete: filename: {filename} error: {error}".format(filename = filename, error = error)) + except AssertionError as msg: + shared.state.end() + return TrainResponse(info = "train embedding error: {msg}".format(msg = msg)) + + def train_hypernetwork(self, args: dict): + try: + shared.state.begin() + initial_hypernetwork = shared.loaded_hypernetwork + apply_optimizations = shared.opts.training_xattention_optimizations + error = None + filename = '' + if not apply_optimizations: + sd_hijack.undo_optimizations() + try: + hypernetwork, filename = train_hypernetwork(*args) + except Exception as e: + error = e + finally: + shared.loaded_hypernetwork = initial_hypernetwork + shared.sd_model.cond_stage_model.to(devices.device) + shared.sd_model.first_stage_model.to(devices.device) + if not apply_optimizations: + sd_hijack.apply_optimizations() + shared.state.end() + return TrainResponse(info = "train embedding complete: filename: {filename} error: {error}".format(filename = filename, error = error)) + except AssertionError as msg: + shared.state.end() + return TrainResponse(info = "train embedding error: {error}".format(error = error)) + def launch(self, server_name, port): self.app.include_router(self.router) uvicorn.run(self.app, host=server_name, port=port) diff --git a/modules/api/models.py b/modules/api/models.py index a22bc6b3..c446ce7a 100644 --- a/modules/api/models.py +++ b/modules/api/models.py @@ -175,6 +175,15 @@ class InterrogateRequest(BaseModel): class InterrogateResponse(BaseModel): caption: str = Field(default=None, title="Caption", description="The generated caption for the image.") +class TrainResponse(BaseModel): + info: str = Field(title="Train info", description="Response string from train embedding or hypernetwork task.") + +class CreateResponse(BaseModel): + info: str = Field(title="Create info", description="Response string from create embedding or hypernetwork task.") + +class PreprocessResponse(BaseModel): + info: str = Field(title="Preprocess info", description="Response string from preprocessing task.") + fields = {} for key, metadata in opts.data_labels.items(): value = opts.data.get(key) diff --git a/modules/hypernetworks/hypernetwork.py b/modules/hypernetworks/hypernetwork.py index c406ffb3..3182ff03 100644 --- a/modules/hypernetworks/hypernetwork.py +++ b/modules/hypernetworks/hypernetwork.py @@ -378,6 +378,32 @@ def report_statistics(loss_info:dict): print(e) +def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): + # Remove illegal characters from name. + name = "".join( x for x in name if (x.isalnum() or x in "._- ")) + + fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") + if not overwrite_old: + assert not os.path.exists(fn), f"file {fn} already exists" + + if type(layer_structure) == str: + layer_structure = [float(x.strip()) for x in layer_structure.split(",")] + + hypernet = modules.hypernetworks.hypernetwork.Hypernetwork( + name=name, + enable_sizes=[int(x) for x in enable_sizes], + layer_structure=layer_structure, + activation_func=activation_func, + weight_init=weight_init, + add_layer_norm=add_layer_norm, + use_dropout=use_dropout, + ) + hypernet.save(fn) + + shared.reload_hypernetworks() + + return fn + def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, gradient_step, data_root, log_directory, training_width, training_height, steps, shuffle_tags, tag_drop_out, latent_sampling_method, 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 allows training previews to have infotext. Importing it at the top causes a circular import problem. diff --git a/modules/hypernetworks/ui.py b/modules/hypernetworks/ui.py index c2d4b51c..e7f9e593 100644 --- a/modules/hypernetworks/ui.py +++ b/modules/hypernetworks/ui.py @@ -3,39 +3,16 @@ import os import re import gradio as gr -import modules.textual_inversion.preprocess -import modules.textual_inversion.textual_inversion +import modules.hypernetworks.hypernetwork from modules import devices, sd_hijack, shared -from modules.hypernetworks import hypernetwork not_available = ["hardswish", "multiheadattention"] -keys = list(x for x in hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) +keys = list(x for x in modules.hypernetworks.hypernetwork.HypernetworkModule.activation_dict.keys() if x not in not_available) def create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure=None, activation_func=None, weight_init=None, add_layer_norm=False, use_dropout=False): - # Remove illegal characters from name. - name = "".join( x for x in name if (x.isalnum() or x in "._- ")) + filename = modules.hypernetworks.hypernetwork.create_hypernetwork(name, enable_sizes, overwrite_old, layer_structure, activation_func, weight_init, add_layer_norm, use_dropout) - fn = os.path.join(shared.cmd_opts.hypernetwork_dir, f"{name}.pt") - if not overwrite_old: - assert not os.path.exists(fn), f"file {fn} already exists" - - if type(layer_structure) == str: - layer_structure = [float(x.strip()) for x in layer_structure.split(",")] - - hypernet = modules.hypernetworks.hypernetwork.Hypernetwork( - name=name, - enable_sizes=[int(x) for x in enable_sizes], - layer_structure=layer_structure, - activation_func=activation_func, - weight_init=weight_init, - add_layer_norm=add_layer_norm, - use_dropout=use_dropout, - ) - hypernet.save(fn) - - shared.reload_hypernetworks() - - return gr.Dropdown.update(choices=sorted([x for x in shared.hypernetworks.keys()])), f"Created: {fn}", "" + return gr.Dropdown.update(choices=sorted([x for x in shared.hypernetworks.keys()])), f"Created: {filename}", "" def train_hypernetwork(*args): -- cgit v1.2.3