base

Classes

fastvideo.configs.pipelines.base.PipelineConfig dataclass

PipelineConfig(model_path: str = '', pipeline_config_path: str | None = None, embedded_cfg_scale: float = 6.0, flow_shift: float | None = None, disable_autocast: bool = False, dit_config: DiTConfig = DiTConfig(), dit_precision: str = 'bf16', vae_config: VAEConfig = VAEConfig(), vae_precision: str = 'fp32', vae_tiling: bool = True, vae_sp: bool = True, image_encoder_config: EncoderConfig = EncoderConfig(), image_encoder_precision: str = 'fp32', text_encoder_configs: tuple[EncoderConfig, ...] = (lambda: (EncoderConfig(),))(), text_encoder_precisions: tuple[str, ...] = (lambda: ('fp32',))(), preprocess_text_funcs: tuple[Callable[[str], str], ...] = (lambda: (preprocess_text,))(), postprocess_text_funcs: tuple[Callable[[BaseEncoderOutput], tensor], ...] = (lambda: (postprocess_text,))(), pos_magic: str | None = None, neg_magic: str | None = None, timesteps_scale: bool | None = None, mask_strategy_file_path: str | None = None, STA_mode: STA_Mode = STA_INFERENCE, skip_time_steps: int = 15, dmd_denoising_steps: list[int] | None = None, ti2v_task: bool = False, boundary_ratio: float | None = None)

Base configuration for all pipeline architectures.

Functions

fastvideo.configs.pipelines.base.PipelineConfig.from_kwargs classmethod

from_kwargs(kwargs: dict[str, Any], config_cli_prefix: str = '') -> PipelineConfig

Load PipelineConfig from kwargs Dictionary. kwargs: dictionary of kwargs config_cli_prefix: prefix of CLI arguments for this PipelineConfig instance

Source code in fastvideo/configs/pipelines/base.py

@classmethod
def from_kwargs(cls,
                kwargs: dict[str, Any],
                config_cli_prefix: str = "") -> "PipelineConfig":
    """
    Load PipelineConfig from kwargs Dictionary.
    kwargs: dictionary of kwargs
    config_cli_prefix: prefix of CLI arguments for this PipelineConfig instance
    """
    from fastvideo.configs.pipelines.registry import (
        get_pipeline_config_cls_from_name)

    prefix_with_dot = f"{config_cli_prefix}." if (config_cli_prefix.strip()
                                                  != "") else ""
    model_path: str | None = kwargs.get(prefix_with_dot + 'model_path',
                                        None) or kwargs.get('model_path')
    pipeline_config_or_path: str | PipelineConfig | dict[
        str, Any] | None = kwargs.get(prefix_with_dot + 'pipeline_config',
                                      None) or kwargs.get('pipeline_config')
    if model_path is None:
        raise ValueError("model_path is required in kwargs")

    # 1. Get the pipeline config class from the registry
    pipeline_config_cls = get_pipeline_config_cls_from_name(model_path)

    # 2. Instantiate PipelineConfig
    if pipeline_config_cls is None:
        logger.warning(
            "Couldn't find pipeline config for %s. Using the default pipeline config.",
            model_path)
        pipeline_config = cls()
    else:
        pipeline_config = pipeline_config_cls()

    # 3. Load PipelineConfig from a json file or a PipelineConfig object if provided
    if isinstance(pipeline_config_or_path, str):
        pipeline_config.load_from_json(pipeline_config_or_path)
        kwargs[prefix_with_dot +
               'pipeline_config_path'] = pipeline_config_or_path
    elif isinstance(pipeline_config_or_path, PipelineConfig):
        pipeline_config = pipeline_config_or_path
    elif isinstance(pipeline_config_or_path, dict):
        pipeline_config.update_pipeline_config(pipeline_config_or_path)

    # 4. Update PipelineConfig from CLI arguments if provided
    kwargs[prefix_with_dot + 'model_path'] = model_path
    pipeline_config.update_config_from_dict(kwargs, config_cli_prefix)
    return pipeline_config

fastvideo.configs.pipelines.base.PipelineConfig.from_pretrained classmethod

from_pretrained(model_path: str) -> PipelineConfig

use the pipeline class setting from model_path to match the pipeline config

Source code in fastvideo/configs/pipelines/base.py

@classmethod
def from_pretrained(cls, model_path: str) -> "PipelineConfig":
    """
    use the pipeline class setting from model_path to match the pipeline config
    """
    from fastvideo.configs.pipelines.registry import (
        get_pipeline_config_cls_from_name)
    pipeline_config_cls = get_pipeline_config_cls_from_name(model_path)

    return cast(PipelineConfig, pipeline_config_cls(model_path=model_path))

fastvideo.configs.pipelines.base.STA_Mode

Bases: str, Enum

STA (Sliding Tile Attention) modes.

fastvideo.configs.pipelines.base.SlidingTileAttnConfig dataclass

SlidingTileAttnConfig(model_path: str = '', pipeline_config_path: str | None = None, embedded_cfg_scale: float = 6.0, flow_shift: float | None = None, disable_autocast: bool = False, dit_config: DiTConfig = DiTConfig(), dit_precision: str = 'bf16', vae_config: VAEConfig = VAEConfig(), vae_precision: str = 'fp32', vae_tiling: bool = True, vae_sp: bool = True, image_encoder_config: EncoderConfig = EncoderConfig(), image_encoder_precision: str = 'fp32', text_encoder_configs: tuple[EncoderConfig, ...] = (lambda: (EncoderConfig(),))(), text_encoder_precisions: tuple[str, ...] = (lambda: ('fp32',))(), preprocess_text_funcs: tuple[Callable[[str], str], ...] = (lambda: (preprocess_text,))(), postprocess_text_funcs: tuple[Callable[[BaseEncoderOutput], tensor], ...] = (lambda: (postprocess_text,))(), pos_magic: str | None = None, neg_magic: str | None = None, timesteps_scale: bool | None = None, mask_strategy_file_path: str | None = None, STA_mode: STA_Mode = STA_INFERENCE, skip_time_steps: int = 15, dmd_denoising_steps: list[int] | None = None, ti2v_task: bool = False, boundary_ratio: float | None = None, window_size: int = 16, stride: int = 8, height: int = 576, width: int = 1024, pad_to_square: bool = False, use_overlap_optimization: bool = True)

Bases: PipelineConfig

Configuration for sliding tile attention.

Functions

fastvideo.configs.pipelines.base.parse_int_list

parse_int_list(value: str) -> list[int]

Parse a comma-separated string of integers into a list.

Source code in fastvideo/configs/pipelines/base.py

def parse_int_list(value: str) -> list[int]:
    """Parse a comma-separated string of integers into a list."""
    if not value:
        return []
    return [int(x.strip()) for x in value.split(",")]