Spaces:
Running
on
Zero
Running
on
Zero
Update wan/configs/wan_animate_14B.py
Browse files- wan/configs/wan_animate_14B.py +40 -40
wan/configs/wan_animate_14B.py
CHANGED
|
@@ -1,40 +1,40 @@
|
|
| 1 |
-
# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved.
|
| 2 |
-
from easydict import EasyDict
|
| 3 |
-
|
| 4 |
-
from .shared_config import wan_shared_cfg
|
| 5 |
-
|
| 6 |
-
#------------------------ Wan animate 14B ------------------------#
|
| 7 |
-
animate_14B = EasyDict(__name__='Config: Wan animate 14B')
|
| 8 |
-
animate_14B.update(wan_shared_cfg)
|
| 9 |
-
|
| 10 |
-
animate_14B.t5_checkpoint = 'models_t5_umt5-xxl-enc-bf16.pth'
|
| 11 |
-
animate_14B.t5_tokenizer = 'google/umt5-xxl'
|
| 12 |
-
|
| 13 |
-
animate_14B.clip_checkpoint = 'models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth'
|
| 14 |
-
animate_14B.clip_tokenizer = 'xlm-roberta-large'
|
| 15 |
-
animate_14B.lora_checkpoint = 'relighting_lora.ckpt'
|
| 16 |
-
# vae
|
| 17 |
-
animate_14B.vae_checkpoint = 'Wan2.1_VAE.pth'
|
| 18 |
-
animate_14B.vae_stride = (4, 8, 8)
|
| 19 |
-
|
| 20 |
-
# transformer
|
| 21 |
-
animate_14B.patch_size = (1, 2, 2)
|
| 22 |
-
animate_14B.dim = 5120
|
| 23 |
-
animate_14B.ffn_dim = 13824
|
| 24 |
-
animate_14B.freq_dim = 256
|
| 25 |
-
animate_14B.num_heads = 40
|
| 26 |
-
animate_14B.num_layers = 40
|
| 27 |
-
animate_14B.window_size = (-1, -1)
|
| 28 |
-
animate_14B.qk_norm = True
|
| 29 |
-
animate_14B.cross_attn_norm = True
|
| 30 |
-
animate_14B.eps = 1e-6
|
| 31 |
-
animate_14B.use_face_encoder = True
|
| 32 |
-
animate_14B.motion_encoder_dim = 512
|
| 33 |
-
|
| 34 |
-
# inference
|
| 35 |
-
animate_14B.sample_shift = 5.0
|
| 36 |
-
animate_14B.sample_steps =
|
| 37 |
-
animate_14B.sample_guide_scale = 1.0
|
| 38 |
-
animate_14B.frame_num = 77
|
| 39 |
-
animate_14B.sample_fps = 30
|
| 40 |
-
animate_14B.prompt = '视频中的人在做动作'
|
|
|
|
| 1 |
+
# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved.
|
| 2 |
+
from easydict import EasyDict
|
| 3 |
+
|
| 4 |
+
from .shared_config import wan_shared_cfg
|
| 5 |
+
|
| 6 |
+
#------------------------ Wan animate 14B ------------------------#
|
| 7 |
+
animate_14B = EasyDict(__name__='Config: Wan animate 14B')
|
| 8 |
+
animate_14B.update(wan_shared_cfg)
|
| 9 |
+
|
| 10 |
+
animate_14B.t5_checkpoint = 'models_t5_umt5-xxl-enc-bf16.pth'
|
| 11 |
+
animate_14B.t5_tokenizer = 'google/umt5-xxl'
|
| 12 |
+
|
| 13 |
+
animate_14B.clip_checkpoint = 'models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth'
|
| 14 |
+
animate_14B.clip_tokenizer = 'xlm-roberta-large'
|
| 15 |
+
animate_14B.lora_checkpoint = 'relighting_lora.ckpt'
|
| 16 |
+
# vae
|
| 17 |
+
animate_14B.vae_checkpoint = 'Wan2.1_VAE.pth'
|
| 18 |
+
animate_14B.vae_stride = (4, 8, 8)
|
| 19 |
+
|
| 20 |
+
# transformer
|
| 21 |
+
animate_14B.patch_size = (1, 2, 2)
|
| 22 |
+
animate_14B.dim = 5120
|
| 23 |
+
animate_14B.ffn_dim = 13824
|
| 24 |
+
animate_14B.freq_dim = 256
|
| 25 |
+
animate_14B.num_heads = 40
|
| 26 |
+
animate_14B.num_layers = 40
|
| 27 |
+
animate_14B.window_size = (-1, -1)
|
| 28 |
+
animate_14B.qk_norm = True
|
| 29 |
+
animate_14B.cross_attn_norm = True
|
| 30 |
+
animate_14B.eps = 1e-6
|
| 31 |
+
animate_14B.use_face_encoder = True
|
| 32 |
+
animate_14B.motion_encoder_dim = 512
|
| 33 |
+
|
| 34 |
+
# inference
|
| 35 |
+
animate_14B.sample_shift = 5.0
|
| 36 |
+
animate_14B.sample_steps = 5
|
| 37 |
+
animate_14B.sample_guide_scale = 1.0
|
| 38 |
+
animate_14B.frame_num = 77
|
| 39 |
+
animate_14B.sample_fps = 30
|
| 40 |
+
animate_14B.prompt = '视频中的人在做动作'
|