Implementation:AUTOMATIC1111 Stable diffusion webui Startup Performance Timer

Knowledge Sources	AUTOMATIC1111_Stable_diffusion_webui
Domains	Performance Monitoring, Startup Profiling
Last Updated	2025-05-15 00:00 GMT

Overview

Provides a lightweight timing utility with the Timer and TimerSubcategory classes for measuring and reporting elapsed time across categorized code sections, primarily used for startup profiling.

Description

The Timer class records elapsed time between consecutive record() calls, accumulating results into named categories stored in a dictionary. Each call to record(category) captures the time since the last call and adds it to the named category. The elapsed() method returns and resets the internal stopwatch. The summary() method produces a human-readable string showing total time and a breakdown of categories that took at least 0.1 seconds. The dump() method serializes timing data as a dictionary, and reset() reinitializes the timer.

The TimerSubcategory class implements a context manager that creates hierarchical timing categories. When entering a subcategory, it prepends the category name as a prefix to subsequent records, enabling nested timing hierarchies with "/" as a separator. It tracks its own elapsed time and records the subcategory duration on exit.

The module also creates a global startup_timer instance that is used throughout the application to profile startup phases. An optional --log-startup command-line flag enables real-time printing of timing records during startup. The startup_record variable stores the final dumped startup timing data.

Usage

Use this module to measure and report execution time for application startup phases, model loading, UI creation, and any other operations where timing visibility is needed.

Code Reference

Source Location

• Repository: AUTOMATIC1111_Stable_diffusion_webui
• File: modules/timer.py
• Lines: 1-91

Signature

class TimerSubcategory:
    def __init__(self, timer: Timer, category: str)
    def __enter__(self) -> None
    def __exit__(self, exc_type, exc_val, exc_tb) -> None

class Timer:
    def __init__(self, print_log: bool = False)
    def elapsed(self) -> float
    def add_time_to_record(self, category: str, amount: float) -> None
    def record(self, category: str, extra_time: float = 0, disable_log: bool = False) -> None
    def subcategory(self, name: str) -> TimerSubcategory
    def summary(self) -> str
    def dump(self) -> dict
    def reset(self) -> None

startup_timer: Timer
startup_record: dict

Import

from modules import timer

# Access the global startup timer
startup_timer = timer.startup_timer

I/O Contract

Inputs

Name	Type	Required	Description
print_log	bool	No	Whether to print timing records to stdout; defaults to False.
category	str	Yes	Name for the timing category in record() and subcategory().
extra_time	float	No	Additional time to add to a record; defaults to 0.
name	str	Yes	Subcategory name for subcategory().

Outputs

Name	Type	Description
elapsed	float	Seconds elapsed since the last elapsed() or record() call.
summary	str	Human-readable string showing total time and category breakdown.
dump	dict	Dictionary with "total" (float) and "records" (dict of category to seconds).

Usage Examples

from modules.timer import Timer

t = Timer(print_log=True)

# Time a sequence of operations
import time
time.sleep(0.5)
t.record("load model")

time.sleep(0.2)
t.record("create ui")

# Use subcategories
with t.subcategory("extensions"):
    time.sleep(0.1)
    t.record("builtin")
    time.sleep(0.1)
    t.record("custom")

print(t.summary())  # e.g., "0.9s (load model: 0.5s, create ui: 0.2s)"
print(t.dump())

Related Pages

• Principle:AUTOMATIC1111_Stable_diffusion_webui_Startup_Performance_Profiling