Upload folder using huggingface_hub
Browse files
pretrain_droid/checkpoints/weights/step_00010000.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a05270518a94800d92aea94684eb08f7b5b5aaf08bc07ad3a7054a3fb3574018
|
| 3 |
+
size 10894838421
|
pretrain_droid/checkpoints/weights/step_00015000.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ece147cab6a7e002a2a60fb4b3c61b0c1a22d4cedabbb94af397b31e7084661c
|
| 3 |
+
size 10894838421
|
pretrain_droid/config.yaml
ADDED
|
@@ -0,0 +1,214 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
output_dir: ./runs/cotrain_droid_teleop/2026-07-04_01-48-06
|
| 2 |
+
batch_size: 8
|
| 3 |
+
num_workers: 4
|
| 4 |
+
prefetch_factor: 4
|
| 5 |
+
lr_scheduler_type: constant
|
| 6 |
+
learning_rate: 1.0e-05
|
| 7 |
+
num_epochs: null
|
| 8 |
+
log_every: 10
|
| 9 |
+
save_every: 5000
|
| 10 |
+
save_accelerator_state: true
|
| 11 |
+
eval_every: 1000
|
| 12 |
+
eval_num_inference_samples: 8
|
| 13 |
+
eval_num_inference_steps: 40
|
| 14 |
+
eval_video_fps: 16
|
| 15 |
+
gradient_accumulation_steps: 1
|
| 16 |
+
mixed_precision: bf16
|
| 17 |
+
seed: 42
|
| 18 |
+
max_grad_norm: 1.0
|
| 19 |
+
weight_decay: 0.001
|
| 20 |
+
resume: runs/cotrain_droid_teleop/video_only/checkpoints/weights/step_00015000.pt
|
| 21 |
+
resume_reset: true
|
| 22 |
+
train_mode: joint
|
| 23 |
+
wandb:
|
| 24 |
+
enabled: true
|
| 25 |
+
workspace: null
|
| 26 |
+
project: fastwam
|
| 27 |
+
name: finetune
|
| 28 |
+
group: null
|
| 29 |
+
mode: online
|
| 30 |
+
data:
|
| 31 |
+
train:
|
| 32 |
+
_target_: fastwam.datasets.lerobot.lerobot_pretrain_dataset_v3.LerobotPretrainDatasetV3
|
| 33 |
+
data_mix:
|
| 34 |
+
teleop: 1.0
|
| 35 |
+
data_root_per_dataset:
|
| 36 |
+
droid: ./data/datasets/lerobot_v3/droid
|
| 37 |
+
teleop: data/datasets/teleop/pnp202/train
|
| 38 |
+
num_frames: 17
|
| 39 |
+
action_video_freq_ratio: 2
|
| 40 |
+
frame_size:
|
| 41 |
+
- 256
|
| 42 |
+
- 320
|
| 43 |
+
camera_mapping_per_dataset:
|
| 44 |
+
droid:
|
| 45 |
+
primary: observation.images.exterior_image_1_left
|
| 46 |
+
wrist: observation.images.wrist_image_left
|
| 47 |
+
teleop:
|
| 48 |
+
primary: observation.images.right_3pv_camera
|
| 49 |
+
wrist: observation.images.wrist_camera
|
| 50 |
+
action_mapping_per_dataset:
|
| 51 |
+
droid:
|
| 52 |
+
skip: 10
|
| 53 |
+
teleop:
|
| 54 |
+
action.body_eef_xyz: 3
|
| 55 |
+
action.body_eef_rot6d: 6
|
| 56 |
+
action.gripper_state: 1
|
| 57 |
+
proprio_mapping_per_dataset:
|
| 58 |
+
droid:
|
| 59 |
+
skip: 8
|
| 60 |
+
teleop:
|
| 61 |
+
state.joint_pos: 7
|
| 62 |
+
state.gripper_state: 1
|
| 63 |
+
action_norm_mask_per_dataset:
|
| 64 |
+
droid:
|
| 65 |
+
action:
|
| 66 |
+
- true
|
| 67 |
+
- true
|
| 68 |
+
- true
|
| 69 |
+
- true
|
| 70 |
+
- true
|
| 71 |
+
- true
|
| 72 |
+
- false
|
| 73 |
+
- false
|
| 74 |
+
teleop:
|
| 75 |
+
action.body_eef_xyz:
|
| 76 |
+
- true
|
| 77 |
+
- true
|
| 78 |
+
- true
|
| 79 |
+
action.body_eef_rot6d:
|
| 80 |
+
- true
|
| 81 |
+
- true
|
| 82 |
+
- true
|
| 83 |
+
- true
|
| 84 |
+
- true
|
| 85 |
+
- true
|
| 86 |
+
action.gripper_state:
|
| 87 |
+
- false
|
| 88 |
+
proprio_norm_mask_per_dataset:
|
| 89 |
+
droid:
|
| 90 |
+
action:
|
| 91 |
+
- true
|
| 92 |
+
- true
|
| 93 |
+
- true
|
| 94 |
+
- true
|
| 95 |
+
- true
|
| 96 |
+
- true
|
| 97 |
+
- false
|
| 98 |
+
- false
|
| 99 |
+
teleop:
|
| 100 |
+
state.joint_pos:
|
| 101 |
+
- true
|
| 102 |
+
- true
|
| 103 |
+
- true
|
| 104 |
+
- true
|
| 105 |
+
- true
|
| 106 |
+
- true
|
| 107 |
+
- true
|
| 108 |
+
state.gripper_state:
|
| 109 |
+
- false
|
| 110 |
+
emit_raw_video: true
|
| 111 |
+
context_len: 128
|
| 112 |
+
image_aug: false
|
| 113 |
+
sampler_type: hierarchical_chunk
|
| 114 |
+
episode_weighting: uniform
|
| 115 |
+
sampler_num_samples: 1000000000
|
| 116 |
+
val:
|
| 117 |
+
_target_: fastwam.datasets.lerobot.lerobot_dataset_v3.LerobotDatasetV3
|
| 118 |
+
data_root: data/datasets/teleop/pnp202/val
|
| 119 |
+
num_frames: 17
|
| 120 |
+
action_video_freq_ratio: 2
|
| 121 |
+
frame_size:
|
| 122 |
+
- 256
|
| 123 |
+
- 320
|
| 124 |
+
camera_mapping:
|
| 125 |
+
primary: observation.images.right_3pv_camera
|
| 126 |
+
wrist: observation.images.wrist_camera
|
| 127 |
+
action_mapping:
|
| 128 |
+
action.body_eef_xyz: 3
|
| 129 |
+
action.body_eef_rot6d: 6
|
| 130 |
+
action.gripper_state: 1
|
| 131 |
+
proprio_mapping:
|
| 132 |
+
state.joint_pos: 7
|
| 133 |
+
state.gripper_state: 1
|
| 134 |
+
action_norm_mask:
|
| 135 |
+
action.body_eef_xyz:
|
| 136 |
+
- true
|
| 137 |
+
- true
|
| 138 |
+
- true
|
| 139 |
+
action.body_eef_rot6d:
|
| 140 |
+
- true
|
| 141 |
+
- true
|
| 142 |
+
- true
|
| 143 |
+
- true
|
| 144 |
+
- true
|
| 145 |
+
- true
|
| 146 |
+
action.gripper_state:
|
| 147 |
+
- false
|
| 148 |
+
proprio_norm_mask:
|
| 149 |
+
state.joint_pos:
|
| 150 |
+
- true
|
| 151 |
+
- true
|
| 152 |
+
- true
|
| 153 |
+
- true
|
| 154 |
+
- true
|
| 155 |
+
- true
|
| 156 |
+
- true
|
| 157 |
+
state.gripper_state:
|
| 158 |
+
- false
|
| 159 |
+
emit_video: true
|
| 160 |
+
context_len: 128
|
| 161 |
+
image_aug: false
|
| 162 |
+
sampler_type: resumable_epoch
|
| 163 |
+
model:
|
| 164 |
+
_target_: fastwam.runtime.create_fastwam_joint
|
| 165 |
+
model_id: Wan-AI/Wan2.2-TI2V-5B
|
| 166 |
+
tokenizer_model_id: Wan-AI/Wan2.1-T2V-1.3B
|
| 167 |
+
tokenizer_max_len: 128
|
| 168 |
+
load_text_encoder: false
|
| 169 |
+
redirect_common_files: false
|
| 170 |
+
mot_checkpoint_mixed_attn: true
|
| 171 |
+
action_dit_pretrained_path: checkpoints/ActionDiT_linear_interp_Wan22_alphascale_512hdim_adim10.pt
|
| 172 |
+
skip_dit_load_from_pretrain: false
|
| 173 |
+
checkpoint_video_dit_only: true
|
| 174 |
+
num_frames: 17
|
| 175 |
+
video_dit_config:
|
| 176 |
+
patch_size:
|
| 177 |
+
- 1
|
| 178 |
+
- 2
|
| 179 |
+
- 2
|
| 180 |
+
in_dim: 48
|
| 181 |
+
hidden_dim: 3072
|
| 182 |
+
ffn_dim: 14336
|
| 183 |
+
freq_dim: 256
|
| 184 |
+
text_dim: 4096
|
| 185 |
+
out_dim: 48
|
| 186 |
+
num_heads: 24
|
| 187 |
+
attn_head_dim: 128
|
| 188 |
+
num_layers: 30
|
| 189 |
+
eps: 1.0e-06
|
| 190 |
+
use_gradient_checkpointing: true
|
| 191 |
+
action_dit_config:
|
| 192 |
+
proprio_dim: 8
|
| 193 |
+
action_dim: 10
|
| 194 |
+
hidden_dim: 512
|
| 195 |
+
ffn_dim: 2048
|
| 196 |
+
num_heads: 24
|
| 197 |
+
attn_head_dim: 128
|
| 198 |
+
num_layers: 30
|
| 199 |
+
text_dim: 4096
|
| 200 |
+
freq_dim: 256
|
| 201 |
+
eps: 1.0e-06
|
| 202 |
+
use_gradient_checkpointing: true
|
| 203 |
+
video_scheduler:
|
| 204 |
+
train_shift: 3.0
|
| 205 |
+
infer_shift: null
|
| 206 |
+
num_train_timesteps: 1000
|
| 207 |
+
action_scheduler:
|
| 208 |
+
train_shift: 3.0
|
| 209 |
+
infer_shift: null
|
| 210 |
+
num_train_timesteps: 1000
|
| 211 |
+
loss:
|
| 212 |
+
lambda_action: 1.0
|
| 213 |
+
name: cotrain
|
| 214 |
+
max_steps: 50000
|