--- title: base ---

fastvideo.models.encoders.base.BaseEncoderOutput

BaseEncoderOutput(last_hidden_state: torch.FloatTensor | None = None, pooler_output: torch.FloatTensor | None = None, hidden_states: tuple[torch.FloatTensor, ...] | None = None, attentions: tuple[torch.FloatTensor, ...] | None = None, attention_mask: torch.Tensor | None = None)

fastvideo.models.encoders.base.ImageEncoderConfig

ImageEncoderConfig(arch_config: ArchConfig = ImageEncoderArchConfig(), prefix: str = '', quant_config: QuantizationConfig | None = None, lora_config: Any | None = None, *, _resolved_attention_backend: AttentionBackendEnum | None = None)

fastvideo.models.encoders.base.TextEncoderConfig

TextEncoderConfig(arch_config: ArchConfig = TextEncoderArchConfig(), prefix: str = '', quant_config: QuantizationConfig | None = None, lora_config: Any | None = None, is_chat_model: bool = False, treat_empty_as_dot: bool = False, *, _resolved_attention_backend: AttentionBackendEnum | None = None, chat_template_enable_thinking: bool = False)

fastvideo.models.encoders.base.AttentionBackendEnum

AttentionBackendEnum()

fastvideo.models.encoders.base.TextEncoder

TextEncoder(config: TextEncoderConfig)

Base for native encoders with a model-specific forward output contract.

Source
class TextEncoder(nn.Module, ABC, Generic[TextEncoderOutputT]):
    """Base for native encoders with a model-specific forward output contract."""

    _fsdp_shard_conditions: list = field(default_factory=lambda: [])
    _stacked_params_mapping: list[tuple[str, str, str]] = field(default_factory=list)
    _supported_attention_backends: tuple[AttentionBackendEnum, ...] = TextEncoderConfig()._supported_attention_backends
    supported_checkpoint_quantization_methods: frozenset[str] = frozenset()

    def __init__(self, config: TextEncoderConfig) -> None:
        super().__init__()
        self.config = config
        self._fsdp_shard_conditions = config._fsdp_shard_conditions
        self._stacked_params_mapping = config.arch_config.stacked_params_mapping
        if not self.supported_attention_backends:
            raise ValueError(f"Subclass {self.__class__.__name__} must define _supported_attention_backends")

    @abstractmethod
    def forward(self, *args: Any, **kwargs: Any) -> TextEncoderOutputT:
        pass

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

fastvideo.models.encoders.base.TextEncoder.forward

forward(*args: Any, **kwargs: Any) -> TextEncoderOutputT
Source
@abstractmethod
def forward(self, *args: Any, **kwargs: Any) -> TextEncoderOutputT:
    pass

fastvideo.models.encoders.base.ImageEncoder

ImageEncoder(config: ImageEncoderConfig)
Source
class ImageEncoder(nn.Module, ABC):
    _supported_attention_backends: tuple[AttentionBackendEnum, ...] = ImageEncoderConfig()._supported_attention_backends

    def __init__(self, config: ImageEncoderConfig) -> None:
        super().__init__()
        self.config = config
        if not self.supported_attention_backends:
            raise ValueError(f"Subclass {self.__class__.__name__} must define _supported_attention_backends")

    @abstractmethod
    def forward(self, pixel_values: torch.Tensor, **kwargs) -> BaseEncoderOutput:
        pass

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

fastvideo.models.encoders.base.ImageEncoder.forward

forward(pixel_values: torch.Tensor, **kwargs) -> BaseEncoderOutput
Source
@abstractmethod
def forward(self, pixel_values: torch.Tensor, **kwargs) -> BaseEncoderOutput:
    pass