lingbotworld2

fastvideo.models.dits.lingbotworld2.LingBotWorld2CausalFastTransformer3DModel

LingBotWorld2CausalFastTransformer3DModel(config: LingBotWorld2CausalFastVideoConfig, hf_config: dict[str, Any])

Released LingBot World 2 14B causal-fast model with native FastVideo loading.

fastvideo.models.dits.lingbotworld2.LingBotWorld2CausalFastTransformer3DModel.forward

forward(hidden_states: torch.Tensor | list[torch.Tensor] | None = None, encoder_hidden_states: torch.Tensor | list[torch.Tensor] | None = None, timestep: torch.Tensor | None = None, encoder_hidden_states_image: torch.Tensor | list[torch.Tensor] | None = None, guidance = None, *, x: list[torch.Tensor] | None = None, t: torch.Tensor | None = None, context: list[torch.Tensor] | torch.Tensor | None = None, seq_len: int | None = None, y: list[torch.Tensor] | None = None, dit_cond_dict: dict[str, Any] | None = None, kv_cache: list[dict] | None = None, crossattn_cache: list[dict] | None = None, current_start: int = 0, max_attention_size: int = 1000000, frame_seqlen: int | None = None, cross_attn_first_call: bool | None = None, **kwargs) -> list[torch.Tensor]

Run one cached causal-fast DiT forward using the released LingBot World 2 ABI.

fastvideo.models.dits.lingbotworld2.LingBotWorld2CausalFastTransformer3DModel.unpatchify

unpatchify(x: torch.Tensor, grid_sizes: torch.Tensor) -> list[torch.Tensor]

Reconstruct latent videos from flattened patch tokens.

fastvideo.models.dits.lingbotworld2.LingBotWorld2CausalFastTransformer3DModel.init_weights

init_weights() -> None

Initialize modules for non-meta construction; checkpoint load overwrites them.