Hugging Face's logo

AlayaLab
/
FloodDiffusion2

PyTorch
English
motion-generation
text-to-motion
diffusion
streaming
humanml3d
Model card Files
xet
 Community
FloodDiffusion2
File size: 2,581 Bytes
2402033
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
dirs:
  deps: ./deps
  raw_data: ./data
  outputs: ./outputs
  checkpoints: ./checkpoints
exp_name: vae_263
seed: 1234
debug: false
train: true
save_dir: ${dirs.outputs}
resume_ckpt: null
test_ckpt: ${dirs.checkpoints}/vae_263/model.ckpt
representation: humanml3d263
test_setting:
  render: true
  HumanML3D:
    compare_folders:
    - ${dirs.raw_data}/HumanML3D/HumanML3D263/animations
    compare_names:
    - Ground Truth
val_repeat: 1
logger:
  wandb:
    wandb_key: ${oc.env:WANDB_API_KEY,null}
    project: FloodDiffusion2
    entity: ${oc.env:WANDB_ENTITY,null}
trainer:
  max_steps: 3000000
  accelerator: gpu
  devices: 1
  log_every_n_steps: 50
  num_nodes: 1
  precision: 32-true
validation:
  validation_steps: 1000
  test_steps: 10000
  save_every_n_steps: 10000
  save_top_k: 100
metrics:
  mr:
    target: metrics.mr.MRMetrics
  t2m:
    target: metrics.HumanML3D263.t2m.T2MMetrics
    params:
      evaluate_text: false
      metric_mean_path: ${dirs.deps}/t2m/meta/mean.npy
      metric_std_path: ${dirs.deps}/t2m/meta/std.npy
      moveencoder:
        target: metrics.HumanML3D263.t2m_evaluator.MovementConvEncoder
        ckpt: ${dirs.deps}/t2m/humanml3d/movement_encoder.pt
        params:
          input_size: 259
          hidden_size: 512
          output_size: 512
      motionencoder:
        target: metrics.HumanML3D263.t2m_evaluator.MotionEncoderBiGRUCo
        ckpt: ${dirs.deps}/t2m/humanml3d/motion_encoder.pt
        params:
          input_size: 512
          hidden_size: 1024
          output_size: 512
data:
  target: datasets.humanml3d.HumanML3DDataset
  collate_fn: datasets.humanml3d.collate_fn
  train_bs: 256
  val_bs: 64
  test_bs: 16
  num_workers: 8
  train_meta_paths:
  - path: ${dirs.raw_data}/HumanML3D/HumanML3D263/train.txt
    name: HumanML3D
  val_meta_paths:
  - path: ${dirs.raw_data}/HumanML3D/HumanML3D263/test.txt
    name: HumanML3D
  test_meta_paths:
  - path: ${dirs.raw_data}/HumanML3D/HumanML3D263/test_min.txt
    name: HumanML3D
  feature_path: new_joint_vecs
  token_path: null
  text_path: texts
  min_length: 40
  max_length: 200
  window_length: 189
model:
  target: models.vae_wan.VAEWanModel
  ema_decay: 0.99
  params:
    input_dim: 263
    z_dim: 4
    mean_path: ${dirs.checkpoints}/vae_263/assets/Mean.npy
    std_path: ${dirs.checkpoints}/vae_263/assets/Std.npy
optimizer:
  target: AdamW
  params:
    lr: 0.0002
    betas:
    - 0.9
    - 0.99
    weight_decay: 0.0
    eps: 1.0e-08
lr_scheduler:
  target: diffusers.optimization.get_constant_schedule_with_warmup
  params:
    num_warmup_steps: 1000