Merge branch 'master' into fix/encode-pnginfo

Author: evshiron

modules/api/api.py

@@ -231,6 +231,10 @@ def get_config(self):
         return options
 
     def set_config(self, req: OptionsModel):
+        # currently req has all options fields even if you send a dict like { "send_seed": false }, which means it will
+        # overwrite all options with default values.
+        raise RuntimeError('Setting options via API is not supported')
+
         reqDict = vars(req)
         for o in reqDict:
             setattr(shared.opts, o, reqDict[o])

modules/api/models.py

@@ -1,6 +1,6 @@
 import inspect
 from pydantic import BaseModel, Field, create_model
-from typing import Any, Optional, Union
+from typing import Any, Optional
 from typing_extensions import Literal
 from inflection import underscore
 from modules.processing import StableDiffusionProcessingTxt2Img, StableDiffusionProcessingImg2Img
@@ -185,22 +185,22 @@ class ProgressResponse(BaseModel):
 for key in _options:
     if(_options[key].dest != 'help'):
         flag = _options[key]
-        _type = str 
-        if(_options[key].default != None): _type = type(_options[key].default) 
+        _type = str
+        if _options[key].default is not None: _type = type(_options[key].default)
         flags.update({flag.dest: (_type,Field(default=flag.default, description=flag.help))})
 
 FlagsModel = create_model("Flags", **flags)
 
 class SamplerItem(BaseModel):
     name: str = Field(title="Name")
-    aliases: list[str]  = Field(title="Aliases")
+    aliases: list[str] = Field(title="Aliases")
     options: dict[str, str] = Field(title="Options")
 
 class UpscalerItem(BaseModel):
     name: str = Field(title="Name")
-    model_name: str | None = Field(title="Model Name")
-    model_path: str | None = Field(title="Path")
-    model_url: str | None = Field(title="URL")
+    model_name: Optional[str] = Field(title="Model Name")
+    model_path: Optional[str] = Field(title="Path")
+    model_url: Optional[str] = Field(title="URL")
 
 class SDModelItem(BaseModel):
     title: str = Field(title="Title")
@@ -211,21 +211,21 @@ class SDModelItem(BaseModel):
 
 class HypernetworkItem(BaseModel):
     name: str = Field(title="Name")
-    path: str | None = Field(title="Path")
+    path: Optional[str] = Field(title="Path")
 
 class FaceRestorerItem(BaseModel):
     name: str = Field(title="Name")
-    cmd_dir: str | None = Field(title="Path")
+    cmd_dir: Optional[str] = Field(title="Path")
 
 class RealesrganItem(BaseModel):
     name: str = Field(title="Name")
-    path: str | None = Field(title="Path")
-    scale: int | None = Field(title="Scale")
+    path: Optional[str] = Field(title="Path")
+    scale: Optional[int] = Field(title="Scale")
 
 class PromptStyleItem(BaseModel):
     name: str = Field(title="Name")
-    prompt: str | None = Field(title="Prompt")
-    negative_prompt: str | None = Field(title="Negative Prompt")
+    prompt: Optional[str] = Field(title="Prompt")
+    negative_prompt: Optional[str] = Field(title="Negative Prompt")
 
 class ArtistItem(BaseModel):
     name: str = Field(title="Name")

modules/extensions.py

@@ -34,8 +34,11 @@ def __init__(self, name, path, enabled=True):
         if repo is None or repo.bare:
             self.remote = None
         else:
-            self.remote = next(repo.remote().urls, None)
-            self.status = 'unknown'
+            try:
+                self.remote = next(repo.remote().urls, None)
+                self.status = 'unknown'
+            except Exception:
+                self.remote = None
 
     def list_files(self, subdir, extension):
         from modules import scripts

modules/hypernetworks/hypernetwork.py

@@ -22,6 +22,8 @@
 from statistics import stdev, mean
 
 
+optimizer_dict = {optim_name : cls_obj for optim_name, cls_obj in inspect.getmembers(torch.optim, inspect.isclass) if optim_name != "Optimizer"}
+
 class HypernetworkModule(torch.nn.Module):
     multiplier = 1.0
     activation_dict = {
@@ -142,6 +144,8 @@ def __init__(self, name=None, enable_sizes=None, layer_structure=None, activatio
         self.use_dropout = use_dropout
         self.activate_output = activate_output
         self.last_layer_dropout = kwargs['last_layer_dropout'] if 'last_layer_dropout' in kwargs else True
+        self.optimizer_name = None
+        self.optimizer_state_dict = None
 
         for size in enable_sizes or []:
             self.layers[size] = (
@@ -163,6 +167,7 @@ def weights(self):
 
     def save(self, filename):
         state_dict = {}
+        optimizer_saved_dict = {}
 
         for k, v in self.layers.items():
             state_dict[k] = (v[0].state_dict(), v[1].state_dict())
@@ -178,8 +183,15 @@ def save(self, filename):
         state_dict['sd_checkpoint_name'] = self.sd_checkpoint_name
         state_dict['activate_output'] = self.activate_output
         state_dict['last_layer_dropout'] = self.last_layer_dropout
-        
+
+        if self.optimizer_name is not None:
+            optimizer_saved_dict['optimizer_name'] = self.optimizer_name
+
         torch.save(state_dict, filename)
+        if shared.opts.save_optimizer_state and self.optimizer_state_dict:
+            optimizer_saved_dict['hash'] = sd_models.model_hash(filename)
+            optimizer_saved_dict['optimizer_state_dict'] = self.optimizer_state_dict
+            torch.save(optimizer_saved_dict, filename + '.optim')
 
     def load(self, filename):
         self.filename = filename
@@ -202,6 +214,18 @@ def load(self, filename):
         print(f"Activate last layer is set to {self.activate_output}")
         self.last_layer_dropout = state_dict.get('last_layer_dropout', False)
 
+        optimizer_saved_dict = torch.load(self.filename + '.optim', map_location = 'cpu') if os.path.exists(self.filename + '.optim') else {}
+        self.optimizer_name = optimizer_saved_dict.get('optimizer_name', 'AdamW')
+        print(f"Optimizer name is {self.optimizer_name}")
+        if sd_models.model_hash(filename) == optimizer_saved_dict.get('hash', None):
+            self.optimizer_state_dict = optimizer_saved_dict.get('optimizer_state_dict', None)
+        else:
+            self.optimizer_state_dict = None
+        if self.optimizer_state_dict:
+            print("Loaded existing optimizer from checkpoint")
+        else:
+            print("No saved optimizer exists in checkpoint")
+
         for size, sd in state_dict.items():
             if type(size) == int:
                 self.layers[size] = (
@@ -219,11 +243,11 @@ def load(self, filename):
 
 def list_hypernetworks(path):
     res = {}
-    for filename in glob.iglob(os.path.join(path, '**/*.pt'), recursive=True):
+    for filename in sorted(glob.iglob(os.path.join(path, '**/*.pt'), recursive=True)):
         name = os.path.splitext(os.path.basename(filename))[0]
         # Prevent a hypothetical "None.pt" from being listed.
         if name != "None":
-            res[name] = filename
+            res[name + f"({sd_models.model_hash(filename)})"] = filename
     return res
 
 
@@ -358,6 +382,7 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
     shared.state.textinfo = "Initializing hypernetwork training..."
     shared.state.job_count = steps
 
+    hypernetwork_name = hypernetwork_name.rsplit('(', 1)[0]
     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)
@@ -404,8 +429,22 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
     weights = hypernetwork.weights()
     for weight in weights:
         weight.requires_grad = True
-    # if optimizer == "AdamW": or else Adam / AdamW / SGD, etc...
-    optimizer = torch.optim.AdamW(weights, lr=scheduler.learn_rate)
+
+    # Here we use optimizer from saved HN, or we can specify as UI option.
+    if hypernetwork.optimizer_name in optimizer_dict:
+        optimizer = optimizer_dict[hypernetwork.optimizer_name](params=weights, lr=scheduler.learn_rate)
+        optimizer_name = hypernetwork.optimizer_name
+    else:
+        print(f"Optimizer type {hypernetwork.optimizer_name} is not defined!")
+        optimizer = torch.optim.AdamW(params=weights, lr=scheduler.learn_rate)
+        optimizer_name = 'AdamW'
+
+    if hypernetwork.optimizer_state_dict:  # This line must be changed if Optimizer type can be different from saved optimizer.
+        try:
+            optimizer.load_state_dict(hypernetwork.optimizer_state_dict)
+        except RuntimeError as e:
+            print("Cannot resume from saved optimizer!")
+            print(e)
 
     steps_without_grad = 0
 
@@ -467,7 +506,11 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
             # Before saving, change name to match current checkpoint.
             hypernetwork_name_every = f'{hypernetwork_name}-{steps_done}'
             last_saved_file = os.path.join(hypernetwork_dir, f'{hypernetwork_name_every}.pt')
+            hypernetwork.optimizer_name = optimizer_name
+            if shared.opts.save_optimizer_state:
+                hypernetwork.optimizer_state_dict = optimizer.state_dict()
             save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, last_saved_file)
+            hypernetwork.optimizer_state_dict = None  # dereference it after saving, to save memory.
 
         textual_inversion.write_loss(log_directory, "hypernetwork_loss.csv", hypernetwork.step, len(ds), {
             "loss": f"{previous_mean_loss:.7f}",
@@ -530,8 +573,12 @@ def train_hypernetwork(hypernetwork_name, learn_rate, batch_size, data_root, log
     report_statistics(loss_dict)
 
     filename = os.path.join(shared.cmd_opts.hypernetwork_dir, f'{hypernetwork_name}.pt')
+    hypernetwork.optimizer_name = optimizer_name
+    if shared.opts.save_optimizer_state:
+        hypernetwork.optimizer_state_dict = optimizer.state_dict()
     save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, filename)
-
+    del optimizer
+    hypernetwork.optimizer_state_dict = None  # dereference it after saving, to save memory.
     return hypernetwork, filename
 
 def save_hypernetwork(hypernetwork, checkpoint, hypernetwork_name, filename):

modules/hypernetworks/ui.py

@@ -9,7 +9,7 @@
 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.

modules/shared.py

@@ -86,6 +86,10 @@
 parser.add_argument("--ui-debug-mode", action='store_true', help="Don't load model to quickly launch UI")
 parser.add_argument("--device-id", type=str, help="Select the default CUDA device to use (export CUDA_VISIBLE_DEVICES=0,1,etc might be needed before)", default=None)
 parser.add_argument("--administrator", action='store_true', help="Administrator rights", default=False)
+parser.add_argument("--cors-allow-origins", type=str, help="Allowed CORS origins", default=None)
+parser.add_argument("--tls-keyfile", type=str, help="Partially enables TLS, requires --tls-certfile to fully function", default=None)
+parser.add_argument("--tls-certfile", type=str, help="Partially enables TLS, requires --tls-keyfile to fully function", default=None)
+parser.add_argument("--server-name", type=str, help="Sets hostname of server", default=None)
 
 cmd_opts = parser.parse_args()
 restricted_opts = {
@@ -147,9 +151,9 @@ def interrupt(self):
         self.interrupted = True
 
     def nextjob(self):
-        if opts.show_progress_every_n_steps == -1: 
+        if opts.show_progress_every_n_steps == -1:
             self.do_set_current_image()
-            
+
         self.job_no += 1
         self.sampling_step = 0
         self.current_image_sampling_step = 0
@@ -198,7 +202,7 @@ def do_set_current_image(self):
             return
         if self.current_latent is None:
             return
-            
+
         if opts.show_progress_grid:
             self.current_image = sd_samplers.samples_to_image_grid(self.current_latent)
         else:
@@ -317,6 +321,7 @@ def options_section(section_identifier, options_dict):
 
 options_templates.update(options_section(('training', "Training"), {
     "unload_models_when_training": OptionInfo(False, "Move VAE and CLIP to RAM when training if possible. Saves VRAM."),
+    "save_optimizer_state": OptionInfo(False, "Saves Optimizer state as separate *.optim file. Training can be resumed with HN itself and matching optim file."),
     "dataset_filename_word_regex": OptionInfo("", "Filename word regex"),
     "dataset_filename_join_string": OptionInfo(" ", "Filename join string"),
     "training_image_repeats_per_epoch": OptionInfo(1, "Number of repeats for a single input image per epoch; used only for displaying epoch number", gr.Number, {"precision": 0}),
@@ -406,7 +411,8 @@ def __setattr__(self, key, value):
             if key in self.data or key in self.data_labels:
                 assert not cmd_opts.freeze_settings, "changing settings is disabled"
 
-                comp_args = opts.data_labels[key].component_args
+                info = opts.data_labels.get(key, None)
+                comp_args = info.component_args if info else None
                 if isinstance(comp_args, dict) and comp_args.get('visible', True) is False:
                     raise RuntimeError(f"not possible to set {key} because it is restricted")
 

modules/ui.py

@@ -1446,17 +1446,19 @@ def run_settings(*args):
                 continue
 
             oldval = opts.data.get(key, None)
-
-            setattr(opts, key, value)
-
+            try:
+                setattr(opts, key, value)
+            except RuntimeError:
+                continue
             if oldval != value:
                 if opts.data_labels[key].onchange is not None:
                     opts.data_labels[key].onchange()
 
                 changed += 1
-
-        opts.save(shared.config_filename)
-
+        try:
+            opts.save(shared.config_filename)
+        except RuntimeError:
+            return opts.dumpjson(), f'{changed} settings changed without save.'
         return opts.dumpjson(), f'{changed} settings changed.'
 
     def run_settings_single(value, key):

modules/ui_extensions.py

@@ -188,7 +188,7 @@ def refresh_available_extensions_from_data():
 
         code += f"""
             <tr>
-                <td><a href="{html.escape(url)}">{html.escape(name)}</a></td>
+                <td><a href="{html.escape(url)}" target="_blank">{html.escape(name)}</a></td>
                 <td>{html.escape(description)}</td>
                 <td>{install_code}</td>
             </tr>

modules/upscaler.py

@@ -57,10 +57,18 @@ def upscale(self, img: PIL.Image, scale: int, selected_model: str = None):
         self.scale = scale
         dest_w = img.width * scale
         dest_h = img.height * scale
+
         for i in range(3):
-            if img.width > dest_w and img.height > dest_h:
-                break
+            shape = (img.width, img.height)
+
             img = self.do_upscale(img, selected_model)
+
+            if shape == (img.width, img.height):
+                break
+
+            if img.width >= dest_w and img.height >= dest_h:
+                break
+
         if img.width != dest_w or img.height != dest_h:
             img = img.resize((int(dest_w), int(dest_h)), resample=LANCZOS)