--- title: base ---

fastvideo.models.dits.base.DiTConfig

DiTConfig(arch_config: DiTArchConfig = DiTArchConfig(), prefix: str = '', quant_config: QuantizationConfig | None = None, *, _resolved_attention_backend: AttentionBackendEnum | None = None)

fastvideo.models.dits.base.DiTConfig.add_cli_args

add_cli_args(parser: Any, prefix: str = 'dit-config') -> Any

Add CLI arguments for DiTConfig fields

fastvideo.models.dits.base.AttentionBackendEnum

AttentionBackendEnum()

fastvideo.models.dits.base.BaseDiT

BaseDiT(config: DiTConfig, hf_config: dict[str, Any], **kwargs)
Source
class BaseDiT(nn.Module, ABC):
    _fsdp_shard_conditions: list = []
    _compile_conditions: list = []
    param_names_mapping: dict
    reverse_param_names_mapping: dict
    hidden_size: int
    num_attention_heads: int
    num_channels_latents: int
    # always supports torch_sdpa
    _supported_attention_backends: tuple[AttentionBackendEnum, ...] = DiTConfig()._supported_attention_backends

    def __init_subclass__(cls) -> None:
        required_class_attrs = ["_fsdp_shard_conditions", "param_names_mapping", "_compile_conditions"]
        super().__init_subclass__()
        for attr in required_class_attrs:
            if not hasattr(cls, attr):
                raise AttributeError(f"Subclasses of BaseDiT must define '{attr}' class variable")

    def __init__(self, config: DiTConfig, hf_config: dict[str, Any], **kwargs) -> None:
        super().__init__()
        self.config = config
        self.hf_config = hf_config
        if not self.supported_attention_backends:
            raise ValueError(f"Subclass {self.__class__.__name__} must define _supported_attention_backends")

    @abstractmethod
    def forward(self,
                hidden_states: torch.Tensor,
                encoder_hidden_states: torch.Tensor | list[torch.Tensor],
                timestep: torch.LongTensor,
                encoder_hidden_states_image: torch.Tensor | list[torch.Tensor]
                | None = None,
                guidance=None,
                **kwargs) -> torch.Tensor:
        pass

    def __post_init__(self) -> None:
        required_attrs = ["hidden_size", "num_attention_heads", "num_channels_latents"]
        for attr in required_attrs:
            if not hasattr(self, attr):
                raise AttributeError(f"Subclasses of BaseDiT must define '{attr}' instance variable")

    @property
    def supported_attention_backends(self) -> tuple[AttentionBackendEnum, ...]:
        return self._supported_attention_backends


    def preprocess_loaded_weight(self, name: str, tensor: torch.Tensor) -> torch.Tensor:
        """Adapt a checkpoint tensor before sharding; default models are unchanged."""
        return tensor

    @property
    def _fastvideo_shard_cache_key_extra(self) -> Any:
        """Extra weight-shard-cache key material for structurally different variants."""
        return None


    def inference_config_overrides(self) -> dict[str, Any]:
        """Extra ``config.json`` keys a deployable checkpoint of this model needs."""
        return {}


    def inference_export_expected_shape(self, output_key: str, base_shape: tuple[int, ...]) -> tuple[int, ...]:
        """The exported shape of *output_key* given the base checkpoint's shape."""
        return tuple(base_shape)

fastvideo.models.dits.base.BaseDiT.forward

forward(hidden_states: torch.Tensor, encoder_hidden_states: torch.Tensor | list[torch.Tensor], timestep: torch.LongTensor, encoder_hidden_states_image: torch.Tensor | list[torch.Tensor] | None = None, guidance = None, **kwargs) -> torch.Tensor
Source
@abstractmethod
def forward(self,
            hidden_states: torch.Tensor,
            encoder_hidden_states: torch.Tensor | list[torch.Tensor],
            timestep: torch.LongTensor,
            encoder_hidden_states_image: torch.Tensor | list[torch.Tensor]
            | None = None,
            guidance=None,
            **kwargs) -> torch.Tensor:
    pass

fastvideo.models.dits.base.BaseDiT.preprocess_loaded_weight

preprocess_loaded_weight(name: str, tensor: torch.Tensor) -> torch.Tensor

Adapt a checkpoint tensor before sharding; default models are unchanged.

Source
def preprocess_loaded_weight(self, name: str, tensor: torch.Tensor) -> torch.Tensor:
    """Adapt a checkpoint tensor before sharding; default models are unchanged."""
    return tensor

fastvideo.models.dits.base.BaseDiT.inference_config_overrides

inference_config_overrides() -> dict[str, Any]

Extra config.json keys a deployable checkpoint of this model needs.

Source
def inference_config_overrides(self) -> dict[str, Any]:
    """Extra ``config.json`` keys a deployable checkpoint of this model needs."""
    return {}

fastvideo.models.dits.base.BaseDiT.inference_export_expected_shape

inference_export_expected_shape(output_key: str, base_shape: tuple[int, ...]) -> tuple[int, ...]

The exported shape of output_key given the base checkpoint's shape.

Source
def inference_export_expected_shape(self, output_key: str, base_shape: tuple[int, ...]) -> tuple[int, ...]:
    """The exported shape of *output_key* given the base checkpoint's shape."""
    return tuple(base_shape)