--- title: lingbot_video ---

Qwen3-VL text-only encoder configuration used by LingBot-Video.

fastvideo.configs.models.encoders.lingbot_video.TextEncoderArchConfig

TextEncoderArchConfig(stacked_params_mapping: list[tuple[str, str, str]] = list(), architectures: list[str] = (lambda: [])(), _supported_attention_backends: tuple[AttentionBackendEnum, ...] = (AttentionBackendEnum.FLASH_ATTN, AttentionBackendEnum.TORCH_SDPA), output_hidden_states: bool = False, use_return_dict: bool = True, vocab_size: int = 0, hidden_size: int = 0, num_hidden_layers: int = 0, num_attention_heads: int = 0, pad_token_id: int = 0, eos_token_id: int = 0, text_len: int = 0, hidden_state_skip_layer: int = 0, decoder_start_token_id: int = 0, output_past: bool = True, scalable_attention: bool = True, tie_word_embeddings: bool = False, tokenizer_kwargs: dict[str, Any] = dict(), _fsdp_shard_conditions: list = (lambda: [])(), require_processor: bool = False)

fastvideo.configs.models.encoders.lingbot_video.Qwen3TextArchConfig

Qwen3TextArchConfig(stacked_params_mapping: list[tuple[str, str, str | int]] = (lambda: [('.qkv_proj', '.q_proj', 'q'), ('.qkv_proj', '.k_proj', 'k'), ('.qkv_proj', '.v_proj', 'v'), ('.gate_up_proj', '.gate_proj', 0), ('.gate_up_proj', '.up_proj', 1)])(), architectures: list[str] = (lambda: [])(), _supported_attention_backends: tuple[AttentionBackendEnum, ...] = (AttentionBackendEnum.FLASH_ATTN, AttentionBackendEnum.TORCH_SDPA), output_hidden_states: bool = True, use_return_dict: bool = True, vocab_size: int = 151936, hidden_size: int = 2560, num_hidden_layers: int = 36, num_attention_heads: int = 32, pad_token_id: int = 151643, eos_token_id: int = 151645, text_len: int = 512, hidden_state_skip_layer: int = 0, decoder_start_token_id: int = 0, output_past: bool = True, scalable_attention: bool = True, tie_word_embeddings: bool = True, tokenizer_kwargs: dict[str, Any] = dict(), _fsdp_shard_conditions: list = (lambda: [_is_transformer_layer, _is_embeddings, _is_final_norm])(), require_processor: bool = False, intermediate_size: int = 9728, num_key_value_heads: int = 8, hidden_act: str = 'silu', max_position_embeddings: int = 40960, initializer_range: float = 0.02, rms_norm_eps: float = 1e-06, use_cache: bool = True, bos_token_id: int = 151643, rope_theta: float = 1000000.0, rope_scaling: dict | None = None, attention_bias: bool = False, attention_dropout: float = 0.0, mlp_bias: bool = False, head_dim: int = 128)

Architecture config for Qwen3 text encoder.

Qwen3 is similar to LLaMA but with QK-Norm (RMSNorm on Q and K before attention). Used by Flux2 Klein.

fastvideo.configs.models.encoders.lingbot_video.Qwen3TextConfig

Qwen3TextConfig(arch_config: TextEncoderArchConfig = Qwen3TextArchConfig(), prefix: str = 'qwen3', quant_config: QuantizationConfig | None = None, lora_config: Any | None = None, is_chat_model: bool = True, treat_empty_as_dot: bool = False, *, _resolved_attention_backend: AttentionBackendEnum | None = None, chat_template_enable_thinking: bool = False)

Top-level config for Qwen3 text encoder.

fastvideo.configs.models.encoders.lingbot_video.LingBotVideoQwen3VLTextArchConfig

LingBotVideoQwen3VLTextArchConfig(stacked_params_mapping: list[tuple[str, str, str | int]] = (lambda: [('.qkv_proj', '.q_proj', 'q'), ('.qkv_proj', '.k_proj', 'k'), ('.qkv_proj', '.v_proj', 'v'), ('.gate_up_proj', '.gate_proj', 0), ('.gate_up_proj', '.up_proj', 1)])(), architectures: list[str] = (lambda: ['LingBotVideoQwen3VLTextModel'])(), _supported_attention_backends: tuple[AttentionBackendEnum, ...] = (AttentionBackendEnum.FLASH_ATTN, AttentionBackendEnum.TORCH_SDPA), output_hidden_states: bool = True, use_return_dict: bool = True, vocab_size: int = 151936, hidden_size: int = 2560, num_hidden_layers: int = 36, num_attention_heads: int = 32, pad_token_id: int = 151643, eos_token_id: int = 151645, text_len: int = 37698, hidden_state_skip_layer: int = 0, decoder_start_token_id: int = 0, output_past: bool = True, scalable_attention: bool = True, tie_word_embeddings: bool = True, tokenizer_kwargs: dict[str, Any] = dict(), _fsdp_shard_conditions: list = (lambda: [_is_transformer_layer, _is_embeddings, _is_final_norm])(), require_processor: bool = True, intermediate_size: int = 9728, num_key_value_heads: int = 8, hidden_act: str = 'silu', max_position_embeddings: int = 262144, initializer_range: float = 0.02, rms_norm_eps: float = 1e-06, use_cache: bool = True, bos_token_id: int = 151643, rope_theta: float = 5000000.0, rope_scaling: dict | None = None, attention_bias: bool = False, attention_dropout: float = 0.0, mlp_bias: bool = False, head_dim: int = 128, mrope_interleaved: bool = True, mrope_section: tuple[int, int, int] = (24, 20, 20))

Exact Qwen3-VL language-model architecture released with LingBot-Video.

Source
@dataclass
class LingBotVideoQwen3VLTextArchConfig(Qwen3TextArchConfig):
    """Exact Qwen3-VL language-model architecture released with LingBot-Video."""

    architectures: list[str] = field(default_factory=lambda: ["LingBotVideoQwen3VLTextModel"])
    vocab_size: int = 151936
    hidden_size: int = 2560
    intermediate_size: int = 9728
    num_hidden_layers: int = 36
    num_attention_heads: int = 32
    num_key_value_heads: int = 8
    max_position_embeddings: int = 262144
    rms_norm_eps: float = 1e-6
    rope_theta: float = 5000000.0
    rope_scaling: dict | None = None
    mrope_interleaved: bool = True
    mrope_section: tuple[int, int, int] = (24, 20, 20)
    bos_token_id: int = 151643
    eos_token_id: int = 151645
    pad_token_id: int = 151643
    text_len: int = 37698
    output_hidden_states: bool = True
    require_processor: bool = True

    def __post_init__(self) -> None:
        """Match the official processor call used by LingBotVideoPipeline."""
        self.tokenizer_kwargs = {
            "truncation": True,
            "max_length": self.text_len,
            "padding": "longest",
            "return_tensors": "pt",
        }

fastvideo.configs.models.encoders.lingbot_video.LingBotVideoQwen3VLTextConfig

LingBotVideoQwen3VLTextConfig(arch_config: TextEncoderArchConfig = LingBotVideoQwen3VLTextArchConfig(), prefix: str = 'language_model', 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 loader config for the LingBot-Video text-only Qwen3-VL path.

Source
@dataclass
class LingBotVideoQwen3VLTextConfig(Qwen3TextConfig):
    """FastVideo loader config for the LingBot-Video text-only Qwen3-VL path."""

    arch_config: TextEncoderArchConfig = field(default_factory=LingBotVideoQwen3VLTextArchConfig)
    prefix: str = "language_model"
    is_chat_model: bool = False