Aero-Hand/logs_r34_stdout.txt at main · luck2shi/Aero-Hand · 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
140
141
142
143
144
nohup: 忽略输入
Environment Config:
action_repeat: 1
action_scale:
- 1.4
- 0.8
- 1.4
- 1.4
- 1.4
- 1.4
ctrl_dt: 0.05
early_termination: true
episode_length: 800
force_history_len: 1
history_len: 1
impl: jax
noise_config:
  force_ema_alpha: 0.85
  level: 0.3
  scales:
    hw_force: 0.05
    hw_pos: 0.005
reset_config:
  hand_qpos_noise_scale: 0.02
reward_config:
  finger_active_threshold: 0.08
  force_contact_saturation: 1.0
  force_contact_threshold: 0.06
  force_overload_soft_width: 1.5
  force_overload_threshold: 2.8
  scales:
    action_accel: -0.005
    action_rate: -0.01
    approach: 15.0
    closure: 40.0
    contact: 10.0
    drop_risk: -5.0
    finger_participation: 0.0
    force_balance: 0.0
    force_contact: 20.0
    force_overload: 0.0
    grip_force: 8.0
    height: 3.0
    hold_position: 30.0
    human_pose: 5.0
    idle_follow: 0.0
    pip_closure: 15.0
    progressive_hold: 0.0
    soft_contact: 0.0
    stable_hold: 0.0
    survival: 1.0
    termination: -80.0
    thumb_engage: 40.0
    torques: -3.0e-05
  soft_contact_fmax: 2.5
  soft_contact_fmin: 0.1
sim_dt: 0.01
spawn_config:
  cube_jitter:
  - 0.003
  - 0.005
  - 0.003
  cube_pos:
  - 0.0
  - -0.04
  - 0.0955
  support_enabled: true
support_config:
  min_release_active_fingers: 3
  min_release_force: 0.2
  release_after_sec: 2.0
  release_ramp_sec: 2.0
  support_hidden_pos:
  - 0.0
  - 0.0
  - -10.0
  support_pos:
  - 0.0
  - -0.04
  - 0.08
tactile_config:
  force_saturation_n: 3.0
  obs_force_ema_alpha: 0.7
  taxel_weights:
  - 0.7
  - 1.0
  - 1.0
  - 0.7
  - 1.0
  - 1.4
  - 1.4
  - 1.0
  - 1.0
  - 1.4
  - 1.4
  - 1.0
  - 0.7
  - 1.0
  - 1.0
  - 0.7
  use_pooled_obs: true
  use_real_tactile: true

PPO Training Parameters:
action_repeat: 1
batch_size: 256
discounting: 0.97
entropy_cost: 0.01
episode_length: 800
learning_rate: 0.0001
network_factory:
  policy_hidden_layer_sizes: &id001 !!python/tuple
  - 512
  - 256
  - 128
  policy_obs_key: state
  value_hidden_layer_sizes: *id001
  value_obs_key: privileged_state
normalize_observations: true
num_envs: 8192
num_evals: 5
num_minibatches: 32
num_resets_per_eval: 1
num_timesteps: 20000000
num_updates_per_batch: 4
reward_scaling: 1.0
unroll_length: 40

Experiment name: AeroCubeGraspV2Force-20260414-150028-v2_R34_continue_R33
Logs are being stored in: /home/ll/SRTP/Aero-Hand/sim_rl/mujoco_playground/logs/AeroCubeGraspV2Force-20260414-150028-v2_R34_continue_R33
Restoring from: /home/ll/SRTP/Aero-Hand/sim_rl/mujoco_playground/logs/AeroCubeGraspV2Force-20260414-135824-v2_R33_ramp2s/checkpoints/000020971520
Checkpoint path: /home/ll/SRTP/Aero-Hand/sim_rl/mujoco_playground/logs/AeroCubeGraspV2Force-20260414-150028-v2_R34_continue_R33/checkpoints
0: reward=77.415
5242880: reward=75.012
10485760: reward=77.046
15728640: reward=76.832
20971520: reward=90.816
Done training.
Time to JIT compile: 45.72809476699331
Time to train: 3322.3381047250004
Starting inference...
FPS for rendering: 10.0
  0%|          | 0/400 [00:00<?, ?it/s]  6%|▋         | 25/400 [00:00<00:01, 249.60it/s] 13%|█▎        | 51/400 [00:00<00:01, 254.46it/s] 20%|█▉        | 78/400 [00:00<00:01, 257.77it/s] 26%|██▌       | 104/400 [00:00<00:01, 255.81it/s] 32%|███▎      | 130/400 [00:00<00:01, 252.86it/s] 39%|███▉      | 156/400 [00:00<00:00, 248.60it/s] 45%|████▌     | 181/400 [00:00<00:00, 247.38it/s] 52%|█████▏    | 208/400 [00:00<00:00, 252.00it/s] 58%|█████▊    | 234/400 [00:00<00:00, 247.03it/s] 65%|██████▌   | 260/400 [00:01<00:00, 248.63it/s] 71%|███████▏  | 285/400 [00:01<00:00, 246.62it/s] 78%|███████▊  | 310/400 [00:01<00:00, 240.26it/s] 84%|████████▍ | 335/400 [00:01<00:00, 241.31it/s] 90%|█████████ | 360/400 [00:01<00:00, 240.45it/s] 96%|█████████▋| 386/400 [00:01<00:00, 244.69it/s]100%|██████████| 400/400 [00:01<00:00, 246.81it/s]
Rollout video saved as 'rollout0.mp4'.