Hugging Face's logo

ThomasWalker1
/
paarbench-checkpoints

Model card Files
xet
 Community
paarbench-checkpoints
File size: 2,102 Bytes
e407564
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
329f421
e407564
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
ckpt_base_path: <path>
training:
  seed: 0
  epochs: 5
  batch_size: 64
  save_every_x_epoch: 1
  save_every_x_iterations: -1
  straighten: cos1e-3
  decoder_start_epoch: 5
  reconstruct_every_x_batch: 1000
  num_reconstruct_samples: 5
  encoder_lr: 1.0e-05
  decoder_lr: 0.0003
  predictor_lr: 0.0005
  action_encoder_lr: 0.0005
  stop_grad: true
  vcreg: false
  vcreg_apply_to: enc
  vcreg_std_coeff: 0
  vcreg_cov_coeff: 0
  mixed_precision: bf16
img_size: 224
frameskip: 5
concat_dim: 1
normalize_action: true
action_emb_dim: 10
num_action_repeat: 1
proprio_emb_dim: 10
num_proprio_repeat: 1
num_hist: 3
num_pred: 1
has_predictor: true
has_decoder: true
model:
  _target_: models.visual_world_model.VWorldModel
  image_size: 224
  num_hist: 3
  num_pred: 1
  train_encoder: true
  train_predictor: true
  train_decoder: false
plan_settings:
  plan_cfg_path: null
  planner:
  - gd
  - cem
  goal_source:
  - dset
  - random_state
  goal_H:
  - 5
  alpha:
  - 0.1
  - 1
debug: false
env:
  name: pushobj
  args: []
  kwargs:
    with_velocity: true
    with_target: true
  dataset:
    _target_: datasets.pusht_dset.load_pusht_slice_train_val
    with_velocity: true
    n_rollout: null
    normalize_action: true
    data_path: data/pushobj_multishape
    split_ratio: 0.9
    transform:
      _target_: datasets.img_transforms.default_transform
      img_size: 224
  decoder_path: null
  num_workers: 16
encoder:
  _target_: models.encoder.resnet.SmallResNetGeM
  dim: 384
  gem_p: 3.0
action_encoder:
  _target_: models.proprio.ProprioceptiveEmbedding
  num_frames: 1
  tubelet_size: 1
  use_3d_pos: false
  use_layernorm: true
proprio_encoder:
  _target_: models.proprio.ProprioceptiveEmbedding
  num_frames: 1
  tubelet_size: 1
  use_3d_pos: false
  use_layernorm: true
decoder:
  _target_: models.vqvae.VQVAE
  channel: 384
  n_embed: 2048
  n_res_block: 4
  n_res_channel: 128
  quantize: false
predictor:
  _target_: models.vit.ViTPredictor
  depth: 6
  heads: 16
  mlp_dim: 2048
  dropout: 0.1
  emb_dropout: 0
  pool: mean
saved_folder: <path>
effective_batch_size: 64
gpu_batch_size: 64