Hugging Face's logo

OpenWAM
/
Pretrained_OpenWAM_Real_SFT

Model card Files
xet
 Community
Pretrained_OpenWAM_Real_SFT
File size: 3,513 Bytes
72054bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
model:
  video_backbone:
    encoder:
      name: wan_vae
      model_path: /mnt/data/wangyuran/Wan2.2-TI2V-5B
    name: wan22_ti2v_5b
    model_path: /mnt/data/wangyuran/Wan2.2-TI2V-5B
    from_scratch: false
    shift_video: 5.0
    components:
    - attr: dit
      model_class: openwam.model.video_backbone.wan.models.dit.WanModel
      extra_kwargs:
        has_image_input: false
        patch_size:
        - 1
        - 2
        - 2
        in_dim: 48
        dim: 3072
        ffn_dim: 14336
        freq_dim: 256
        text_dim: 4096
        out_dim: 48
        num_heads: 24
        num_layers: 30
        eps: 1.0e-06
        seperated_timestep: true
        require_clip_embedding: false
        require_vae_embedding: false
        fuse_vae_embedding_in_latents: true
    - attr: vae
      model_class: openwam.model.video_backbone.wan.models.vae.WanVideoVAE38
      extra_kwargs: {}
    - attr: text_encoder
      model_class: openwam.model.video_backbone.wan.models.text_encoder.WanTextEncoder
      extra_kwargs: {}
    tokenizer:
      class: openwam.model.video_backbone.wan.models.text_encoder.HuggingfaceTokenizer
      attr: tokenizer
      subdir: tokenizer/google/umt5-xxl
      path_kwarg: name
      kwargs:
        seq_len: 512
        clean: whitespace
  action_backbone:
    dim: 1024
    ffn_dim: 4096
    shift_action: 5.0
  freeze:
  - video_backbone.text_encoder
  - video_backbone.reason1
  - video_backbone.vae
  - video_backbone.image_encoder
  - video_backbone.video_encoder
  architecture:
    framework: dual_system
    variant: joint_self_attn
    action_dim: 80
    use_proprioception: true
    state_dim: 80
    bridge_layers: null
    bridge_interval: 1
    mot_checkpoint_mixed_attn: true
    attention_mask_mode: action_sees_video
    video_attention_mask_mode: first_frame_causal
    detach_bridge: false
    idm_video_cond_noise_prob: 0.5
dataloader:
  type: wuji_real_task
  dataset_dir: /mnt/data/liangjiaqi/wuji_sft_data_eef
  num_frames: 33
  video_stride: 4
  window_stride: 1
  height: 384
  width: 320
  multiview: true
  target_camera: observation.images.stereo
  camera_layout:
  - observation.images.stereo
  - observation.images.cam_left_wrist
  - observation.images.cam_right_wrist
  normalize_mode: quantile
  split: train
  enable_action_supervision: true
  unify_action: true
  unify_action_map:
  - 0-8
  - 10-29
  - 32-40
  - 42-61
  seed: 42
training:
  debug: false
  learning_rate: 0.0001
  adam_betas:
  - 0.9
  - 0.95
  weight_decay: 0.01
  max_grad_norm: 1.0
  num_epochs: null
  max_steps: 100000
  batch_size: 2
  gradient_accumulation_steps: 1
  lr_scheduler: cosine
  warmup_ratio: 0.05
  lr_min_ratio: 0.01
  action_lr: null
  video_lr: null
  mixed_precision: bf16
  zero_stage: 2
  use_gradient_checkpointing: false
  use_gradient_checkpointing_offload: false
  initialize_model_on_cpu: false
  offload_optimizer_device: none
  lambda_video: 1.0
  lambda_action: 1.0
  max_timestep_boundary: 1.0
  min_timestep_boundary: 0.0
  timestep_sampling: default
  timestep_sampling_lead: video
  timestep_sampling_alpha: 1.0
  output_path: /mnt/data/liangjiaqi/openwam_checkpoints
  save_steps: 2000
  keep_last_k_ckpts: 1
  finetune_ckpt_path: /mnt/data/wangyuran/openwam_checkpoints/pretrained_final_dualsystem_jointselfattention
  resume_ckpt_path: null
  dataset_num_workers: 8
project:
  name: open-wam
  output_dir: outputs/${now:%Y-%m-%d_%H-%M-%S}
  seed: 42
  wandb:
    project: open-wam
    run_name: wuji-eef-sft
    entity: null