Implementation:AUTOMATIC1111 Stable diffusion webui WebuiSdModel Type

Knowledge Sources	AUTOMATIC1111_Stable_diffusion_webui
Domains	Model Loading, Type System
Last Updated	2025-05-15 00:00 GMT

Overview

Defines the WebuiSdModel type stub class that documents the runtime fields added to Stable Diffusion model instances by the web UI.

Description

The WebuiSdModel class extends LatentDiffusion from the underlying ldm library but is never directly instantiated. Instead, it serves as a type annotation stub whose fields are dynamically populated by the web UI at runtime. It declares typed annotations for model metadata such as hash, checkpoint path, architecture flags (is_sdxl, is_ssd, is_sd2, is_sd1, is_sd3), optimization state (lowvram), and latent channel count. This allows other modules to use WebuiSdModel as a type hint for shared.sd_model, enabling IDE autocompletion and static type checking across the codebase.

Usage

Use this type when you need to annotate a variable that holds the currently loaded Stable Diffusion model instance or when you need to check model architecture flags during conditional processing.

Code Reference

Source Location

• Repository: AUTOMATIC1111_Stable_diffusion_webui
• File: modules/sd_models_types.py
• Lines: 1-40

Signature

class WebuiSdModel(LatentDiffusion):
    lowvram: bool
    sd_model_hash: str
    sd_model_checkpoint: str
    sd_checkpoint_info: 'CheckpointInfo'
    is_sdxl: bool
    is_ssd: bool
    is_sd2: bool
    is_sd1: bool
    is_sd3: bool
    latent_channels: int

Import

from modules.sd_models_types import WebuiSdModel

I/O Contract

Inputs

Name	Type	Required	Description
(none)	N/A	N/A	This class is not instantiated; fields are set dynamically at runtime by the model loading pipeline.

Outputs

Name	Type	Description
lowvram	bool	True if lowvram/medvram optimizations are enabled.
sd_model_hash	str	Short hash (first 10 chars of SHA1) of the model file; may be None if --no-hashing is used.
sd_model_checkpoint	str	Path to the file on disk that model weights were obtained from.
sd_checkpoint_info	CheckpointInfo	Structure with additional information about the model weights file.
is_sdxl	bool	True if the model architecture is SDXL or SSD.
is_ssd	bool	True if the model is SSD.
is_sd2	bool	True if the model architecture is SD 2.x.
is_sd1	bool	True if the model architecture is SD 1.x.
is_sd3	bool	True if the model architecture is SD 3.
latent_channels	int	Number of channels in the latent image representation (16 for SD3, 4 for others).

Usage Examples

from modules import shared
from modules.sd_models_types import WebuiSdModel

model: WebuiSdModel = shared.sd_model

if model.is_sdxl:
    print("Using SDXL architecture")

print(f"Model hash: {model.sd_model_hash}")
print(f"Latent channels: {model.latent_channels}")

Related Pages

• Principle:AUTOMATIC1111_Stable_diffusion_webui_Model_Loading