| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
|
|
| import copy |
|
|
| from .general_dit import GeneralDIT |
| from .lazy_config_init import LazyCall as L |
| from .lazy_config_init import LazyDict |
|
|
| FADITV2Config: LazyDict = L(GeneralDIT)( |
| max_img_h=240, |
| max_img_w=240, |
| max_frames=128, |
| in_channels=16, |
| out_channels=16, |
| patch_spatial=2, |
| patch_temporal=1, |
| model_channels=4096, |
| block_config="FA-CA-MLP", |
| num_blocks=28, |
| num_heads=32, |
| concat_padding_mask=True, |
| pos_emb_cls="rope3d", |
| pos_emb_learnable=False, |
| pos_emb_interpolation="crop", |
| block_x_format="THWBD", |
| affline_emb_norm=True, |
| use_adaln_lora=True, |
| adaln_lora_dim=256, |
| ) |
|
|
|
|
| FADITV2_14B_Config = copy.deepcopy(FADITV2Config) |
| FADITV2_14B_Config.model_channels = 5120 |
| FADITV2_14B_Config.num_heads = 40 |
| FADITV2_14B_Config.num_blocks = 36 |
|
|