stable-diffusion-webui/webui.py

from __future__ import annotations

import os
import sys
import time
import importlib
import signal
import re
import warnings
import json
from threading import Thread
from typing import Iterable

from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from fastapi.middleware.gzip import GZipMiddleware
from packaging import version

import logging

# We can't use cmd_opts for this because it will not have been initialized at this point.
log_level = os.environ.get("SD_WEBUI_LOG_LEVEL")
if log_level:
    log_level = getattr(logging, log_level.upper(), None) or logging.INFO
    logging.basicConfig(
        level=log_level,
        format='%(asctime)s %(levelname)s [%(name)s] %(message)s',
        datefmt='%Y-%m-%d %H:%M:%S',
    )

logging.getLogger("torch.distributed.nn").setLevel(logging.ERROR)  # sshh...
logging.getLogger("xformers").addFilter(lambda record: 'A matching Triton is not available' not in record.getMessage())

from modules import timer
startup_timer = timer.startup_timer
startup_timer.record("launcher")

import torch
import pytorch_lightning   # noqa: F401 # pytorch_lightning should be imported after torch, but it re-enables warnings on import so import once to disable them
warnings.filterwarnings(action="ignore", category=DeprecationWarning, module="pytorch_lightning")
warnings.filterwarnings(action="ignore", category=UserWarning, module="torchvision")
startup_timer.record("import torch")

import gradio  # noqa: F401
startup_timer.record("import gradio")

from modules import paths, timer, import_hook, errors, devices  # noqa: F401
startup_timer.record("setup paths")

import ldm.modules.encoders.modules  # noqa: F401
startup_timer.record("import ldm")

from modules import extra_networks
from modules.call_queue import wrap_gradio_gpu_call, wrap_queued_call, queue_lock  # noqa: F401

# Truncate version number of nightly/local build of PyTorch to not cause exceptions with CodeFormer or Safetensors
if ".dev" in torch.__version__ or "+git" in torch.__version__:
    torch.__long_version__ = torch.__version__
    torch.__version__ = re.search(r'[\d.]+[\d]', torch.__version__).group(0)

from modules import shared, sd_samplers, upscaler, extensions, localization, ui_tempdir, ui_extra_networks, config_states
import modules.codeformer_model as codeformer
import modules.face_restoration
import modules.gfpgan_model as gfpgan
import modules.img2img

import modules.lowvram
import modules.scripts
import modules.sd_hijack
import modules.sd_hijack_optimizations
import modules.sd_models
import modules.sd_vae
import modules.sd_unet
import modules.txt2img
import modules.script_callbacks
import modules.textual_inversion.textual_inversion
import modules.progress

import modules.ui
from modules import modelloader
from modules.shared import cmd_opts
import modules.hypernetworks.hypernetwork

startup_timer.record("other imports")


if cmd_opts.server_name:
    server_name = cmd_opts.server_name
else:
    server_name = "0.0.0.0" if cmd_opts.listen else None


def fix_asyncio_event_loop_policy():
    """
        The default `asyncio` event loop policy only automatically creates
        event loops in the main threads. Other threads must create event
        loops explicitly or `asyncio.get_event_loop` (and therefore
        `.IOLoop.current`) will fail. Installing this policy allows event
        loops to be created automatically on any thread, matching the
        behavior of Tornado versions prior to 5.0 (or 5.0 on Python 2).
    """

    import asyncio

    if sys.platform == "win32" and hasattr(asyncio, "WindowsSelectorEventLoopPolicy"):
        # "Any thread" and "selector" should be orthogonal, but there's not a clean
        # interface for composing policies so pick the right base.
        _BasePolicy = asyncio.WindowsSelectorEventLoopPolicy  # type: ignore
    else:
        _BasePolicy = asyncio.DefaultEventLoopPolicy

    class AnyThreadEventLoopPolicy(_BasePolicy):  # type: ignore
        """Event loop policy that allows loop creation on any thread.
        Usage::

            asyncio.set_event_loop_policy(AnyThreadEventLoopPolicy())
        """

        def get_event_loop(self) -> asyncio.AbstractEventLoop:
            try:
                return super().get_event_loop()
            except (RuntimeError, AssertionError):
                # This was an AssertionError in python 3.4.2 (which ships with debian jessie)
                # and changed to a RuntimeError in 3.4.3.
                # "There is no current event loop in thread %r"
                loop = self.new_event_loop()
                self.set_event_loop(loop)
                return loop

    asyncio.set_event_loop_policy(AnyThreadEventLoopPolicy())


def check_versions():
    if shared.cmd_opts.skip_version_check:
        return

    expected_torch_version = "2.0.0"

    if version.parse(torch.__version__) < version.parse(expected_torch_version):
        errors.print_error_explanation(f"""
You are running torch {torch.__version__}.
The program is tested to work with torch {expected_torch_version}.
To reinstall the desired version, run with commandline flag --reinstall-torch.
Beware that this will cause a lot of large files to be downloaded, as well as
there are reports of issues with training tab on the latest version.

Use --skip-version-check commandline argument to disable this check.
        """.strip())

    expected_xformers_version = "0.0.20"
    if shared.xformers_available:
        import xformers

        if version.parse(xformers.__version__) < version.parse(expected_xformers_version):
            errors.print_error_explanation(f"""
You are running xformers {xformers.__version__}.
The program is tested to work with xformers {expected_xformers_version}.
To reinstall the desired version, run with commandline flag --reinstall-xformers.

Use --skip-version-check commandline argument to disable this check.
            """.strip())


def restore_config_state_file():
    config_state_file = shared.opts.restore_config_state_file
    if config_state_file == "":
        return

    shared.opts.restore_config_state_file = ""
    shared.opts.save(shared.config_filename)

    if os.path.isfile(config_state_file):
        print(f"*** About to restore extension state from file: {config_state_file}")
        with open(config_state_file, "r", encoding="utf-8") as f:
            config_state = json.load(f)
            config_states.restore_extension_config(config_state)
        startup_timer.record("restore extension config")
    elif config_state_file:
        print(f"!!! Config state backup not found: {config_state_file}")


def validate_tls_options():
    if not (cmd_opts.tls_keyfile and cmd_opts.tls_certfile):
        return

    try:
        if not os.path.exists(cmd_opts.tls_keyfile):
            print("Invalid path to TLS keyfile given")
        if not os.path.exists(cmd_opts.tls_certfile):
            print(f"Invalid path to TLS certfile: '{cmd_opts.tls_certfile}'")
    except TypeError:
        cmd_opts.tls_keyfile = cmd_opts.tls_certfile = None
        print("TLS setup invalid, running webui without TLS")
    else:
        print("Running with TLS")
    startup_timer.record("TLS")


def get_gradio_auth_creds() -> Iterable[tuple[str, ...]]:
    """
    Convert the gradio_auth and gradio_auth_path commandline arguments into
    an iterable of (username, password) tuples.
    """
    def process_credential_line(s) -> tuple[str, ...] | None:
        s = s.strip()
        if not s:
            return None
        return tuple(s.split(':', 1))

    if cmd_opts.gradio_auth:
        for cred in cmd_opts.gradio_auth.split(','):
            cred = process_credential_line(cred)
            if cred:
                yield cred

    if cmd_opts.gradio_auth_path:
        with open(cmd_opts.gradio_auth_path, 'r', encoding="utf8") as file:
            for line in file.readlines():
                for cred in line.strip().split(','):
                    cred = process_credential_line(cred)
                    if cred:
                        yield cred


def configure_sigint_handler():
    # 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}')
        os._exit(0)

    if not os.environ.get("COVERAGE_RUN"):
        # Don't install the immediate-quit handler when running under coverage,
        # as then the coverage report won't be generated.
        signal.signal(signal.SIGINT, sigint_handler)


def configure_opts_onchange():
    shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights()), call=False)
    shared.opts.onchange("sd_vae", wrap_queued_call(lambda: modules.sd_vae.reload_vae_weights()), call=False)
    shared.opts.onchange("sd_vae_as_default", wrap_queued_call(lambda: modules.sd_vae.reload_vae_weights()), call=False)
    shared.opts.onchange("temp_dir", ui_tempdir.on_tmpdir_changed)
    shared.opts.onchange("gradio_theme", shared.reload_gradio_theme)
    shared.opts.onchange("cross_attention_optimization", wrap_queued_call(lambda: modules.sd_hijack.model_hijack.redo_hijack(shared.sd_model)), call=False)
    startup_timer.record("opts onchange")


def initialize():
    fix_asyncio_event_loop_policy()
    validate_tls_options()
    configure_sigint_handler()
    check_versions()
    modelloader.cleanup_models()
    configure_opts_onchange()

    modules.sd_models.setup_model()
    startup_timer.record("setup SD model")

    codeformer.setup_model(cmd_opts.codeformer_models_path)
    startup_timer.record("setup codeformer")

    gfpgan.setup_model(cmd_opts.gfpgan_models_path)
    startup_timer.record("setup gfpgan")

    initialize_rest(reload_script_modules=False)


def initialize_rest(*, reload_script_modules=False):
    """
    Called both from initialize() and when reloading the webui.
    """
    sd_samplers.set_samplers()
    extensions.list_extensions()
    startup_timer.record("list extensions")

    restore_config_state_file()

    if cmd_opts.ui_debug_mode:
        shared.sd_upscalers = upscaler.UpscalerLanczos().scalers
        modules.scripts.load_scripts()
        return

    modules.sd_models.list_models()
    startup_timer.record("list SD models")

    localization.list_localizations(cmd_opts.localizations_dir)

    with startup_timer.subcategory("load scripts"):
        modules.scripts.load_scripts()

    if reload_script_modules:
        for module in [module for name, module in sys.modules.items() if name.startswith("modules.ui")]:
            importlib.reload(module)
        startup_timer.record("reload script modules")

    modelloader.load_upscalers()
    startup_timer.record("load upscalers")

    modules.sd_vae.refresh_vae_list()
    startup_timer.record("refresh VAE")
    modules.textual_inversion.textual_inversion.list_textual_inversion_templates()
    startup_timer.record("refresh textual inversion templates")

    modules.script_callbacks.on_list_optimizers(modules.sd_hijack_optimizations.list_optimizers)
    modules.sd_hijack.list_optimizers()
    startup_timer.record("scripts list_optimizers")

    modules.sd_unet.list_unets()
    startup_timer.record("scripts list_unets")

    def load_model():
        """
        Accesses shared.sd_model property to load model.
        After it's available, if it has been loaded before this access by some extension,
        its optimization may be None because the list of optimizaers has neet been filled
        by that time, so we apply optimization again.
        """

        shared.sd_model  # noqa: B018

        if modules.sd_hijack.current_optimizer is None:
            modules.sd_hijack.apply_optimizations()

        devices.first_time_calculation()

    Thread(target=load_model).start()

    shared.reload_hypernetworks()
    startup_timer.record("reload hypernetworks")

    ui_extra_networks.initialize()
    ui_extra_networks.register_default_pages()

    extra_networks.initialize()
    extra_networks.register_default_extra_networks()
    startup_timer.record("initialize extra networks")


def setup_middleware(app):
    app.middleware_stack = None  # reset current middleware to allow modifying user provided list
    app.add_middleware(GZipMiddleware, minimum_size=1000)
    configure_cors_middleware(app)
    app.build_middleware_stack()  # rebuild middleware stack on-the-fly


def configure_cors_middleware(app):
    cors_options = {
        "allow_methods": ["*"],
        "allow_headers": ["*"],
        "allow_credentials": True,
    }
    if cmd_opts.cors_allow_origins:
        cors_options["allow_origins"] = cmd_opts.cors_allow_origins.split(',')
    if cmd_opts.cors_allow_origins_regex:
        cors_options["allow_origin_regex"] = cmd_opts.cors_allow_origins_regex
    app.add_middleware(CORSMiddleware, **cors_options)


def create_api(app):
    from modules.api.api import Api
    api = Api(app, queue_lock)
    return api


def api_only():
    initialize()

    app = FastAPI()
    setup_middleware(app)
    api = create_api(app)

    modules.script_callbacks.app_started_callback(None, app)

    print(f"Startup time: {startup_timer.summary()}.")
    api.launch(
        server_name="0.0.0.0" if cmd_opts.listen else "127.0.0.1",
        port=cmd_opts.port if cmd_opts.port else 7861,
        root_path=f"/{cmd_opts.subpath}" if cmd_opts.subpath else ""
    )


def webui():
    launch_api = cmd_opts.api
    initialize()

    while 1:
        if shared.opts.clean_temp_dir_at_start:
            ui_tempdir.cleanup_tmpdr()
            startup_timer.record("cleanup temp dir")

        modules.script_callbacks.before_ui_callback()
        startup_timer.record("scripts before_ui_callback")

        shared.demo = modules.ui.create_ui()
        startup_timer.record("create ui")

        if not cmd_opts.no_gradio_queue:
            shared.demo.queue(64)

        gradio_auth_creds = list(get_gradio_auth_creds()) or None

        app, local_url, share_url = shared.demo.launch(
            share=cmd_opts.share,
            server_name=server_name,
            server_port=cmd_opts.port,
            ssl_keyfile=cmd_opts.tls_keyfile,
            ssl_certfile=cmd_opts.tls_certfile,
            ssl_verify=cmd_opts.disable_tls_verify,
            debug=cmd_opts.gradio_debug,
            auth=gradio_auth_creds,
            inbrowser=cmd_opts.autolaunch and os.getenv('SD_WEBUI_RESTARTING') != '1',
            prevent_thread_lock=True,
            allowed_paths=cmd_opts.gradio_allowed_path,
            app_kwargs={
                "docs_url": "/docs",
                "redoc_url": "/redoc",
            },
            root_path=f"/{cmd_opts.subpath}" if cmd_opts.subpath else "",
        )

        # after initial launch, disable --autolaunch for subsequent restarts
        cmd_opts.autolaunch = False

        startup_timer.record("gradio launch")

        # gradio uses a very open CORS policy via app.user_middleware, which makes it possible for
        # an attacker to trick the user into opening a malicious HTML page, which makes a request to the
        # running web ui and do whatever the attacker wants, including installing an extension and
        # running its code. We disable this here. Suggested by RyotaK.
        app.user_middleware = [x for x in app.user_middleware if x.cls.__name__ != 'CORSMiddleware']

        setup_middleware(app)

        modules.progress.setup_progress_api(app)
        modules.ui.setup_ui_api(app)

        if launch_api:
            create_api(app)

        ui_extra_networks.add_pages_to_demo(app)

        startup_timer.record("add APIs")

        with startup_timer.subcategory("app_started_callback"):
            modules.script_callbacks.app_started_callback(shared.demo, app)

        timer.startup_record = startup_timer.dump()
        print(f"Startup time: {startup_timer.summary()}.")

        try:
            while True:
                server_command = shared.state.wait_for_server_command(timeout=5)
                if server_command:
                    if server_command in ("stop", "restart"):
                        break
                    else:
                        print(f"Unknown server command: {server_command}")
        except KeyboardInterrupt:
            print('Caught KeyboardInterrupt, stopping...')
            server_command = "stop"

        if server_command == "stop":
            print("Stopping server...")
            # If we catch a keyboard interrupt, we want to stop the server and exit.
            shared.demo.close()
            break

        print('Restarting UI...')
        shared.demo.close()
        time.sleep(0.5)
        startup_timer.reset()
        modules.script_callbacks.app_reload_callback()
        startup_timer.record("app reload callback")
        modules.script_callbacks.script_unloaded_callback()
        startup_timer.record("scripts unloaded callback")
        initialize_rest(reload_script_modules=True)


if __name__ == "__main__":
    if cmd_opts.nowebui:
        api_only()
    else:
        webui()