fastvideo.configs.models.dits.wanvideo

Contents

`fastvideo.configs.models.dits.wanvideo`#

Module Contents#

Classes#

`WanVideoArchConfig`
`WanVideoConfig`

Functions#

API#

class fastvideo.configs.models.dits.wanvideo.WanVideoArchConfig[source]#

Bases: fastvideo.configs.models.dits.base.DiTArchConfig

added_kv_proj_dim: int | None[source]#: None

attention_head_dim: int[source]#: 128

boundary_ratio: float | None[source]#: None

cross_attn_norm: bool[source]#: True

eps: float[source]#: 1e-06

exclude_lora_layers: list[str][source]#: ‘field(…)’

ffn_dim: int[source]#: 13824

freq_dim: int[source]#: 256

image_dim: int | None[source]#: None

in_channels: int[source]#: 16

local_attn_size: int[source]#: None

lora_param_names_mapping: dict[source]#: ‘field(…)’

num_attention_heads: int[source]#: 40

num_frames_per_block: int[source]#: 3

num_layers: int[source]#: 40

out_channels: int[source]#: 16

param_names_mapping: dict[source]#: ‘field(…)’

patch_size: tuple[int, int, int][source]#: (1, 2, 2)

pos_embed_seq_len: int | None[source]#: None

qk_norm: str[source]#: ‘rms_norm_across_heads’

reverse_param_names_mapping: dict[source]#: ‘field(…)’

rope_max_seq_len: int[source]#: 1024

sink_size: int[source]#: 0

sliding_window_num_frames: int[source]#: 21

text_dim: int[source]#: 4096

text_len[source]#: 512

class fastvideo.configs.models.dits.wanvideo.WanVideoConfig[source]#

Bases: fastvideo.configs.models.dits.base.DiTConfig

arch_config: fastvideo.configs.models.dits.base.DiTArchConfig[source]#: ‘field(…)’

prefix: str#: ‘Wan’

fastvideo.configs.models.dits.wanvideo.is_blocks(n: str, m) → bool[source]#