Skip to content

dreamx_world

DreamX-World-5B-Cam FastVideo model configuration helpers.

Classes

fastvideo.configs.pipelines.dreamx_world.DreamXWorld5BARPipelineConfig dataclass 

DreamXWorld5BARPipelineConfig(model_path: str = '', pipeline_config_path: str | None = None, embedded_cfg_scale: float = 6.0, flow_shift: float | None = 5.0, flow_shift_sr: float | None = None, disable_autocast: bool = False, scheduler_step_in_fp32: bool = False, is_causal: bool = True, dit_config: DiTConfig = make_dreamx_world_5b_ar_dit_config(), dit_precision: str = 'bf16', upsampler_config: UpsamplerConfig = UpsamplerConfig(), upsampler_precision: str = 'fp32', vae_config: VAEConfig = make_dreamx_world_5b_cam_vae_config(), vae_precision: str = 'fp32', vae_decode_precision: str | None = 'bf16', vae_tiling: bool = False, vae_sp: bool = False, image_encoder_config: EncoderConfig = EncoderConfig(), image_encoder_precision: str = 'fp32', text_encoder_configs: tuple[EncoderConfig, ...] = (lambda: (make_dreamx_world_5b_cam_text_encoder_config(),))(), text_encoder_precisions: tuple[str, ...] = (lambda: ('bf16',))(), preprocess_text_funcs: tuple[Callable[[str], str], ...] = (lambda: (preprocess_text,))(), postprocess_text_funcs: tuple[Callable[[BaseEncoderOutput], Tensor], ...] = (lambda: (t5_postprocess_text,))(), dmd_denoising_steps: tuple[int, ...] = (1000, 750, 500, 250), ti2v_task: bool = True, lucy_edit_task: bool = False, boundary_ratio: float | None = None, expand_timesteps: bool = True, warp_denoising_step: bool = True, context_noise: float = 0.1, num_frames_per_block: int = 3, color_correction_strength: float = 1.0)

Bases: DreamXWorld5BCamPipelineConfig

Pipeline config for DreamX-World-5B autoregressive forcing.

fastvideo.configs.pipelines.dreamx_world.DreamXWorld5BCamPipelineConfig dataclass 

DreamXWorld5BCamPipelineConfig(model_path: str = '', pipeline_config_path: str | None = None, embedded_cfg_scale: float = 6.0, flow_shift: float | None = 3.0, flow_shift_sr: float | None = None, disable_autocast: bool = False, scheduler_step_in_fp32: bool = False, is_causal: bool = False, dit_config: DiTConfig = make_dreamx_world_5b_cam_dit_config(), dit_precision: str = 'bf16', upsampler_config: UpsamplerConfig = UpsamplerConfig(), upsampler_precision: str = 'fp32', vae_config: VAEConfig = make_dreamx_world_5b_cam_vae_config(), vae_precision: str = 'fp32', vae_decode_precision: str | None = 'bf16', vae_tiling: bool = False, vae_sp: bool = False, image_encoder_config: EncoderConfig = EncoderConfig(), image_encoder_precision: str = 'fp32', text_encoder_configs: tuple[EncoderConfig, ...] = (lambda: (make_dreamx_world_5b_cam_text_encoder_config(),))(), text_encoder_precisions: tuple[str, ...] = (lambda: ('bf16',))(), preprocess_text_funcs: tuple[Callable[[str], str], ...] = (lambda: (preprocess_text,))(), postprocess_text_funcs: tuple[Callable[[BaseEncoderOutput], Tensor], ...] = (lambda: (t5_postprocess_text,))(), dmd_denoising_steps: list[int] | None = None, ti2v_task: bool = True, lucy_edit_task: bool = False, boundary_ratio: float | None = None, expand_timesteps: bool = True)

Bases: PipelineConfig

Pipeline config for the first-scope DreamX-World-5B-Cam mode.

Functions:

fastvideo.configs.pipelines.dreamx_world.make_dreamx_world_5b_ar_dit_config 

make_dreamx_world_5b_ar_dit_config() -> DreamXWorldARConfig

Return the DreamX-World-5B autoregressive causal DiT config.

Source code in fastvideo/configs/pipelines/dreamx_world.py 
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
def make_dreamx_world_5b_ar_dit_config() -> DreamXWorldARConfig:
    """Return the DreamX-World-5B autoregressive causal DiT config."""
    return DreamXWorldARConfig(arch_config=DreamXWorldARArchConfig(
        model_type="ti2v",
        num_attention_heads=24,
        attention_head_dim=128,
        in_channels=48,
        out_channels=48,
        ffn_dim=14336,
        num_layers=30,
        cross_attn_norm=True,
        qk_norm=True,
        add_control_adapter=True,
        cam_method="prope",
        attn_compress=4,
        cam_self_attn_layers=tuple(range(30)),
        local_attn_size=12,
        sink_size=3,
        num_frames_per_block=3,
    ))

fastvideo.configs.pipelines.dreamx_world.make_dreamx_world_5b_cam_dit_config 

make_dreamx_world_5b_cam_dit_config() -> DreamXWorldConfig

Return the DreamX-World DiT config matching DreamX-World-5B-Cam.

Source code in fastvideo/configs/pipelines/dreamx_world.py 
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
def make_dreamx_world_5b_cam_dit_config() -> DreamXWorldConfig:
    """Return the DreamX-World DiT config matching DreamX-World-5B-Cam."""
    return DreamXWorldConfig(arch_config=DreamXWorldArchConfig(
        num_attention_heads=24,
        attention_head_dim=128,
        in_channels=48,
        out_channels=48,
        ffn_dim=14336,
        num_layers=30,
        cross_attn_norm=True,
        qk_norm="rms_norm_across_heads",
        add_control_adapter=True,
        cam_method="prope",
        attn_compress=1,
        cam_self_attn_layers=None,
    ))

fastvideo.configs.pipelines.dreamx_world.make_dreamx_world_5b_cam_text_encoder_config 

make_dreamx_world_5b_cam_text_encoder_config() -> T5Config

Return the UMT5-XXL text encoder config used by DreamX-World-5B-Cam.

Source code in fastvideo/configs/pipelines/dreamx_world.py 
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
def make_dreamx_world_5b_cam_text_encoder_config() -> T5Config:
    """Return the UMT5-XXL text encoder config used by DreamX-World-5B-Cam."""
    return T5Config(
        arch_config=T5ArchConfig(
            vocab_size=256384,
            d_model=4096,
            d_kv=64,
            d_ff=10240,
            num_layers=24,
            num_decoder_layers=None,
            num_heads=64,
            relative_attention_num_buckets=32,
            dropout_rate=0.0,
            text_len=512,
            feed_forward_proj="gelu",
            is_encoder_decoder=False,
        ),
        prefix="umt5",
    )

fastvideo.configs.pipelines.dreamx_world.make_dreamx_world_5b_cam_vae_config 

make_dreamx_world_5b_cam_vae_config() -> WanVAEConfig

Return the Wan2.2 48-channel VAE config used by DreamX-World-5B-Cam.

Source code in fastvideo/configs/pipelines/dreamx_world.py 
59
60
61
def make_dreamx_world_5b_cam_vae_config() -> WanVAEConfig:
    """Return the Wan2.2 48-channel VAE config used by DreamX-World-5B-Cam."""
    return LucyEditDevConfig().vae_config