aractingi HF Staff commited on
Commit
0683dea
·
verified ·
1 Parent(s): 00023df

Create train_config.json

Browse files
Files changed (1) hide show
  1. rl/train_config.json +285 -0
rl/train_config.json ADDED
@@ -0,0 +1,285 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": null,
3
+ "job_name": "default",
4
+ "resume": false,
5
+ "seed": 1000,
6
+ "num_workers": 4,
7
+ "batch_size": 256,
8
+ "steps": 100000,
9
+ "log_freq": 500,
10
+ "save_checkpoint": true,
11
+ "save_freq": 2000000,
12
+ "wandb": {
13
+ "enable": true,
14
+ "project": "so100_real",
15
+ "disable_artifact": true
16
+ },
17
+ "dataset": {
18
+ "repo_id": "aractingi/pick_lift_cube_pipeline_test_cropped_resized",
19
+ "use_imagenet_stats": false
20
+ },
21
+ "policy": {
22
+ "type": "sac",
23
+ "n_obs_steps": 1,
24
+ "normalization_mapping": {
25
+ "VISUAL": "MEAN_STD",
26
+ "STATE": "MIN_MAX",
27
+ "ENV": "MIN_MAX",
28
+ "ACTION": "MIN_MAX"
29
+ },
30
+ "input_features": {
31
+ "observation.images.top": {
32
+ "type": "VISUAL",
33
+ "shape": [
34
+ 3,
35
+ 128,
36
+ 128
37
+ ]
38
+ },
39
+ "observation.state": {
40
+ "type": "STATE",
41
+ "shape": [
42
+ 18
43
+ ]
44
+ }
45
+ },
46
+ "output_features": {
47
+ "action": {
48
+ "type": "ACTION",
49
+ "shape": [
50
+ 3
51
+ ]
52
+ }
53
+ },
54
+ "device": "cuda",
55
+ "use_amp": false,
56
+ "dataset_stats": {
57
+ "observation.images.top": {
58
+ "mean": [
59
+ 0.485,
60
+ 0.456,
61
+ 0.406
62
+ ],
63
+ "std": [
64
+ 0.229,
65
+ 0.224,
66
+ 0.225
67
+ ]
68
+ },
69
+ "observation.state": {
70
+ "min": [
71
+ -15.0,
72
+ -20.0,
73
+ 10.0,
74
+ 65.0,
75
+ 0.0,
76
+ 0.0,
77
+ -40.0,
78
+ -60.0,
79
+ -50.0,
80
+ -80.0,
81
+ -10.0,
82
+ -110.0,
83
+ 0.26,
84
+ -0.06,
85
+ 0.25
86
+ ],
87
+ "max": [
88
+ 15.0,
89
+ 15.0,
90
+ 40.0,
91
+ 110.0,
92
+ 0.5,
93
+ 31.0,
94
+ 53.0,
95
+ 61.0,
96
+ 67.0,
97
+ 90.0,
98
+ 6.0,
99
+ 100.0,
100
+ 0.32,
101
+ 0.06,
102
+ 0.35
103
+ ]
104
+ },
105
+ "action": {
106
+ "min": [
107
+ -1.0,
108
+ -1.0,
109
+ -1.0
110
+ ],
111
+ "max": [
112
+ 1.0,
113
+ 1.0,
114
+ 1.0
115
+ ]
116
+ }
117
+ },
118
+ "num_discrete_actions": 3,
119
+ "storage_device": "cuda",
120
+ "vision_encoder_name": "helper2424/resnet10",
121
+ "freeze_vision_encoder": true,
122
+ "image_encoder_hidden_dim": 32,
123
+ "shared_encoder": true,
124
+ "online_steps": 1000000,
125
+ "online_env_seed": 10000,
126
+ "online_buffer_capacity": 30000,
127
+ "offline_buffer_capacity": 10000,
128
+ "online_step_before_learning": 10,
129
+ "policy_update_freq": 1,
130
+ "discount": 0.97,
131
+ "async_prefetch": false,
132
+ "temperature_init": 0.01,
133
+ "num_critics": 2,
134
+ "num_subsample_critics": null,
135
+ "critic_lr": 0.0003,
136
+ "actor_lr": 0.0003,
137
+ "temperature_lr": 0.0003,
138
+ "critic_target_update_weight": 0.005,
139
+ "utd_ratio": 2,
140
+ "state_encoder_hidden_dim": 256,
141
+ "latent_dim": 256,
142
+ "target_entropy": null,
143
+ "use_backup_entropy": true,
144
+ "grad_clip_norm": 40.0,
145
+ "critic_network_kwargs": {
146
+ "hidden_dims": [
147
+ 256,
148
+ 256
149
+ ],
150
+ "activate_final": true,
151
+ "final_activation": null
152
+ },
153
+ "actor_network_kwargs": {
154
+ "hidden_dims": [
155
+ 256,
156
+ 256
157
+ ],
158
+ "activate_final": true
159
+ },
160
+ "policy_kwargs": {
161
+ "use_tanh_squash": true,
162
+ "std_min": -5,
163
+ "std_max": 2,
164
+ "init_final": 0.05
165
+ },
166
+ "actor_learner_config": {
167
+ "learner_host": "127.0.0.1",
168
+ "learner_port": 50051,
169
+ "policy_parameters_push_frequency": 4
170
+ },
171
+ "concurrency": {
172
+ "actor": "threads",
173
+ "learner": "threads"
174
+ },
175
+ "repo_id": "aractingi/pick_lift_cube_policy"
176
+ },
177
+ "env": {
178
+ "type": "gym_manipulator",
179
+ "robot": {
180
+ "type": "so100_follower",
181
+ "port": "/dev/ttyACM0",
182
+ "id": "turqoise",
183
+ "cameras": {
184
+ "top": {
185
+ "type": "opencv",
186
+ "index_or_path": 2,
187
+ "height": 480,
188
+ "width": 640,
189
+ "fps": 30
190
+ }
191
+ }
192
+ },
193
+ "teleop": {
194
+ "type": "gamepad",
195
+ "use_gripper": true
196
+ },
197
+ "processor": {
198
+ "control_mode": "gamepad",
199
+ "observation": {
200
+ "display_cameras": false,
201
+ "add_joint_velocity_to_observation": true,
202
+ "add_current_to_observation": true,
203
+ "add_ee_pose_to_observation": true
204
+ },
205
+ "image_preprocessing": {
206
+ "crop_params_dict": {
207
+ "observation.images.top": [49, 189, 262, 305]
208
+ },
209
+ "resize_size": [128, 128]
210
+ },
211
+ "gripper": {
212
+ "use_gripper": true,
213
+ "gripper_penalty": -0.02,
214
+ "gripper_penalty_in_reward": false
215
+ },
216
+ "reset": {
217
+ "fixed_reset_joint_positions": [
218
+ 0.0,
219
+ 0.0,
220
+ 0.0,
221
+ 90.0,
222
+ 0.0,
223
+ 30.0
224
+ ],
225
+ "reset_time_s": 2.5,
226
+ "control_time_s": 20.0,
227
+ "terminate_on_success": true
228
+ },
229
+ "inverse_kinematics": {
230
+ "urdf_path": "/home/rcadene/code/SO-ARM100/Simulation/SO101/so101_new_calib.urdf",
231
+ "target_frame_name": "gripper_frame_link",
232
+ "end_effector_bounds": {
233
+ "min": [
234
+ 0.195,
235
+ -0.065,
236
+ 0.005
237
+ ],
238
+ "max": [
239
+ 0.25,
240
+ 0.065,
241
+ 0.1
242
+ ]
243
+ },
244
+ "end_effector_step_sizes": {
245
+ "x": 0.02,
246
+ "y": 0.02,
247
+ "z": 0.02
248
+ }
249
+ },
250
+ "reward_classifier": {
251
+ "pretrained_path": null,
252
+ "success_threshold": 0.5,
253
+ "success_reward": 1.0
254
+ },
255
+ "max_gripper_pos": 30
256
+ },
257
+ "name": "real_robot",
258
+ "fps": 10,
259
+
260
+ "task": "",
261
+ "features": {
262
+ "observation.images.top": {
263
+ "type": "VISUAL",
264
+ "shape": [
265
+ 3,
266
+ 128,
267
+ 128
268
+ ]
269
+ },
270
+ "observation.state": {
271
+ "type": "STATE",
272
+ "shape": [
273
+ 18
274
+ ]
275
+ },
276
+ "action": {
277
+ "type": "ACTION",
278
+ "shape": [
279
+ 3
280
+ ]
281
+ }
282
+ },
283
+ "features_map": {"observation.images.top": "observation.images.top", "observation.state": "observation.state","action":"action"}
284
+ }
285
+ }