LAM/config.yaml at main · Nemo-1024/LAM · GitHub

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
138
139
# lightning.pytorch==2.4.0
seed_everything: 42
trainer:
  accelerator: gpu
  strategy: auto
  devices: '1'
  num_nodes: 1
  precision: bf16-mixed
  logger: false
  callbacks:
  - class_path: lightning.pytorch.callbacks.ModelCheckpoint
    init_args:
      dirpath: null
      filename: '{epoch}'
      monitor: null
      verbose: true
      save_last: true
      save_top_k: -1
      save_weights_only: false
      mode: min
      auto_insert_metric_name: true
      every_n_train_steps: null
      train_time_interval: null
      every_n_epochs: 1
      save_on_train_epoch_end: true
      enable_version_counter: true
  - class_path: latent_action_model.training.SaveConfigToCheckpointCallback
    init_args:
      config_path: ''
      filename: ''
  fast_dev_run: false
  max_epochs: 1
  min_epochs: null
  max_steps: 1
  min_steps: null
  max_time: null
  limit_train_batches: null
  limit_val_batches: 0
  limit_test_batches: null
  limit_predict_batches: null
  overfit_batches: 0.0
  val_check_interval: 1.0
  check_val_every_n_epoch: 1
  num_sanity_val_steps: 0
  log_every_n_steps: 1000
  enable_checkpointing: null
  enable_progress_bar: null
  enable_model_summary: null
  accumulate_grad_batches: 1
  gradient_clip_val: 1.0
  gradient_clip_algorithm: null
  deterministic: null
  benchmark: null
  inference_mode: true
  use_distributed_sampler: true
  profiler: null
  detect_anomaly: false
  barebones: false
  plugins: null
  sync_batchnorm: false
  reload_dataloaders_every_n_epochs: 0
  default_root_dir: null
model:
  dim: 768
  num_heads: 12
  ffn_expansion_factor: 3
  enc_layers: 12
  codebook_size: 32
  code_dim: 32
  max_state_dim: 32
  num_frames: 2
  num_queries: 1
  vq_kwargs:
    layer_norm: true
  dec_layers: 6
  dropout: 0.0
  lambda_aux: 1.0
  loss_type: smooth_l1
  state_loss_type: l2
  project: lam
  task_name: dino_base_ae
  wandb_offline: true
  optimizer:
    class_path: torch.optim.AdamW
    init_args:
      lr: 0.0001
      betas:
      - 0.9
      - 0.999
      eps: 1.0e-08
      weight_decay: 0.01
      amsgrad: false
      maximize: false
      foreach: null
      capturable: false
      differentiable: false
      fused: null
  weight_decay: 0.01
  exclude_bias_norm_from_wd: true
  make_data_pair: false
  output_dir: output_pairs
  vision_model_id: /mnt/project_rlinf/jlchen/weights/dinov3-vitb16-pretrain-lvd1689m
  warmup_steps: 2000
  lambda_diversity: 0.0
  norm_latents: true
  norm_latents_type: ln
  disable_vq: false
  vq_type: ae
  enc_add_state: false
  enc_modal_mask: true
  latent_layer_to_use: -2
  multi_input: false
  num_embodiments: 32
  image_hw:
  - 256
  - 256
  patch_size: 16
  image_aug: true
  dual_view_aug: true
  decoder_last_ln: true
data:
  data_root_dir: /mnt/project_rlinf/jlchen/datasets
  data_mix: libero
  num_frames: 2
  video_backend: pyav
  preferred_video_key: null
  state_keys: null
  frame_dt_sec: 1.6
  human_frame_dt_sec: 0.4
  batch_size: 1
  num_workers: 0
  prefetch_factor: null
  in_order: false
  pin_memory: false
  persistent_workers: false
  debug_repeat_batch: false
  val_tail_ratio: 0.001
  max_state_dim: 32
ckpt_path: null