fastvideo.v1.models.loader.component_loader

Module Contents

Classes

ComponentLoader	Base class for loading a specific type of model component.
GenericComponentLoader	Generic loader for components that don’t have a specific loader.
ImageEncoderLoader
ImageProcessorLoader	Loader for image processor.
PipelineComponentLoader	Utility class for loading pipeline components. This replaces the chain of if-else statements in load_pipeline_module.
SchedulerLoader	Loader for scheduler.
TextEncoderLoader	Loader for text encoders.
TokenizerLoader	Loader for tokenizers.
TransformerLoader	Loader for transformer.
VAELoader	Loader for VAE.

Data

logger

API

class fastvideo.v1.models.loader.component_loader.ComponentLoader(device=None)

Bases: abc.ABC

Base class for loading a specific type of model component.

Initialization

classmethod for_module_type(module_type: str, transformers_or_diffusers: str) → fastvideo.v1.models.loader.component_loader.ComponentLoader

Factory method to create a component loader for a specific module type.

Parameters:

• module_type – Type of module (e.g., “vae”, “text_encoder”, “transformer”, “scheduler”)

• transformers_or_diffusers – Whether the module is from transformers or diffusers

Returns:

A component loader for the specified module type

abstract load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the component based on the model path, architecture, and inference args.

Parameters:

• model_path – Path to the component model

• architecture – Architecture of the component model

• fastvideo_args – Inference arguments

Returns:

The loaded component

class fastvideo.v1.models.loader.component_loader.GenericComponentLoader(library='transformers')

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Generic loader for components that don’t have a specific loader.

Initialization

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load a generic component based on the model path, architecture, and inference args.

class fastvideo.v1.models.loader.component_loader.ImageEncoderLoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.TextEncoderLoader

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the text encoders based on the model path, architecture, and inference args.

class fastvideo.v1.models.loader.component_loader.ImageProcessorLoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Loader for image processor.

Initialization

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the image processor based on the model path, architecture, and inference args.

class fastvideo.v1.models.loader.component_loader.PipelineComponentLoader

Utility class for loading pipeline components. This replaces the chain of if-else statements in load_pipeline_module.

static load_module(module_name: str, component_model_path: str, transformers_or_diffusers: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load a pipeline module.

Parameters:

• module_name – Name of the module (e.g., “vae”, “text_encoder”, “transformer”, “scheduler”)

• component_model_path – Path to the component model

• transformers_or_diffusers – Whether the module is from transformers or diffusers

• architecture – Architecture of the component model

• fastvideo_args – Inference arguments

Returns:

The loaded module

class fastvideo.v1.models.loader.component_loader.SchedulerLoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Loader for scheduler.

Initialization

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the scheduler based on the model path, architecture, and inference args.

class fastvideo.v1.models.loader.component_loader.TextEncoderLoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Loader for text encoders.

Initialization

class Source

A source for weights.

allow_patterns_overrides: Optional[list[str]]

None

If defined, weights will load exclusively using these patterns.

fall_back_to_pt: bool

True

Whether .pt weights can be used.

model_or_path: str

None

The model ID or path.

prefix: str = <Multiline-String>

A prefix to prepend to all weights.

counter_after_loading_weights: float

0.0

counter_before_loading_weights: float

0.0

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the text encoders based on the model path, architecture, and inference args.

load_model(model_path: str, model_config, target_device: torch.device, dtype: str = 'fp16')

class fastvideo.v1.models.loader.component_loader.TokenizerLoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Loader for tokenizers.

Initialization

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the tokenizer based on the model path, architecture, and inference args.

class fastvideo.v1.models.loader.component_loader.TransformerLoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Loader for transformer.

Initialization

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the transformer based on the model path, architecture, and inference args.

class fastvideo.v1.models.loader.component_loader.VAELoader(device=None)

Bases: fastvideo.v1.models.loader.component_loader.ComponentLoader

Loader for VAE.

Initialization

load(model_path: str, architecture: str, fastvideo_args: fastvideo.v1.fastvideo_args.FastVideoArgs)

Load the VAE based on the model path, architecture, and inference args.

fastvideo.v1.models.loader.component_loader.logger

‘init_logger(…)’