Hugging Face's logo

cijerezg
/
jack_cube_task_step_4800

Safetensors
Model card Files
xet
 Community
cijerezg commited on
Commit
f6bb052
·
verified ·
1 Parent(s): 4121c9c

Upload folder using huggingface_hub

Browse files
Files changed (7) hide show
  1. config.json +166 -0
  2. model.safetensors +3 -0
  3. policy_postprocessor.json +33 -0
  4. policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  5. policy_preprocessor.json +88 -0
  6. policy_preprocessor_step_2_normalizer_processor.safetensors +3 -0
  7. train_config.json +476 -0
config.json ADDED
@@ -0,0 +1,166 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "pi05_rl",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.side": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 224,
10
+ 224
11
+ ]
12
+ },
13
+ "observation.images.top": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 224,
18
+ 224
19
+ ]
20
+ },
21
+ "observation.state": {
22
+ "type": "STATE",
23
+ "shape": [
24
+ 6
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": true,
40
+ "repo_id": "cijerezg/multi-task-toys-merged-v2",
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "paligemma_variant": "gemma_2b",
46
+ "action_expert_variant": "gemma_300m",
47
+ "dtype": "bfloat16",
48
+ "chunk_size": 50,
49
+ "n_action_steps": 50,
50
+ "max_state_dim": 6,
51
+ "max_action_dim": 32,
52
+ "num_inference_steps": 5,
53
+ "time_sampling_beta_alpha": 1.5,
54
+ "time_sampling_beta_beta": 1.0,
55
+ "time_sampling_scale": 0.999,
56
+ "time_sampling_offset": 0.001,
57
+ "min_period": 0.004,
58
+ "max_period": 4.0,
59
+ "rtc_config": {
60
+ "enabled": true,
61
+ "prefix_attention_schedule": "LINEAR",
62
+ "max_guidance_weight": 10.0,
63
+ "execution_horizon": 10,
64
+ "debug": false,
65
+ "debug_maxlen": 100
66
+ },
67
+ "image_resolution": [
68
+ 224,
69
+ 224
70
+ ],
71
+ "empty_cameras": 0,
72
+ "use_dataset_stats": false,
73
+ "normalization_mapping": {
74
+ "VISUAL": "IDENTITY",
75
+ "STATE": "MIN_MAX",
76
+ "ENV": "MIN_MAX",
77
+ "ACTION": "QUANTILES"
78
+ },
79
+ "action_tokenizer_name": "physical-intelligence/fast",
80
+ "text_tokenizer_name": "google/paligemma-3b-pt-224",
81
+ "max_action_tokens": 256,
82
+ "fast_skip_tokens": 128,
83
+ "max_decoding_steps": 200,
84
+ "temperature": 0.0,
85
+ "subtask_regeneration_interval": 1.5,
86
+ "gradient_checkpointing": true,
87
+ "compile_model": false,
88
+ "compile_mode": "max-autotune",
89
+ "freeze_vision_encoder": false,
90
+ "train_expert_only": false,
91
+ "knowledge_insulation": true,
92
+ "action_encoding": "anchor",
93
+ "loss_weight_flow": 1.0,
94
+ "loss_weight_action_ce": 1.0,
95
+ "loss_weight_subtask_ce": 1.0,
96
+ "optimizer_lr": 2.5e-05,
97
+ "optimizer_betas": [
98
+ 0.9,
99
+ 0.95
100
+ ],
101
+ "optimizer_eps": 1e-08,
102
+ "optimizer_weight_decay": 0.1,
103
+ "optimizer_grad_clip_norm": 1.0,
104
+ "scheduler_warmup_steps": 1000,
105
+ "scheduler_decay_steps": 30000,
106
+ "scheduler_decay_lr": 2.5e-06,
107
+ "tokenizer_max_length": 64,
108
+ "task": "Pick up the orange cube and place it on the black X marker",
109
+ "action_dim": 6,
110
+ "drop_n_last_frames": 2,
111
+ "critic_target_update_weight": 0.005,
112
+ "num_critics": 1,
113
+ "discount": 0.97,
114
+ "reward_normalization_constant": 5.0,
115
+ "terminal_failure_reward": -16.0,
116
+ "online_steps": 20000,
117
+ "online_buffer_capacity": 5000,
118
+ "offline_buffer_capacity": 50000,
119
+ "async_prefetch": false,
120
+ "online_step_before_learning": 10,
121
+ "policy_update_freq": 1,
122
+ "grad_clip_norm": 2.0,
123
+ "gradient_accumulation_steps": 16,
124
+ "critic_lr": 5e-05,
125
+ "actor_lr": 5e-05,
126
+ "utd_ratio": 1,
127
+ "actor_device": "cuda:0",
128
+ "learner_device": "cuda:0",
129
+ "use_separate_critic": true,
130
+ "critic_llm_depth": 6,
131
+ "critic_network_kwargs": {
132
+ "hidden_dims": [
133
+ 256,
134
+ 256
135
+ ],
136
+ "activate_final": true
137
+ },
138
+ "trainable_params": {
139
+ "vision_encoder_from_layer": {
140
+ "vision_tower": 5,
141
+ "multi_modal_projector": true
142
+ },
143
+ "language_from_layer": 0,
144
+ "critic_language_from_layer": 5
145
+ },
146
+ "offline_steps": 10000,
147
+ "inference_advantage": 1.0,
148
+ "advantage_scaling": 0.2,
149
+ "pi05_checkpoint": "outputs/pi05_base",
150
+ "action_encoding_stats_path": "outputs/stats_jack/action_stats_anchor_jack_cube.pt",
151
+ "dataset_stats": null,
152
+ "storage_device": "cpu",
153
+ "shared_encoder": false,
154
+ "num_discrete_actions": null,
155
+ "vision_encoder_name": null,
156
+ "actor_learner_config": {
157
+ "learner_host": "192.168.50.1",
158
+ "learner_port": 50051,
159
+ "policy_parameters_push_frequency": 180,
160
+ "queue_get_timeout": 2
161
+ },
162
+ "concurrency": {
163
+ "actor": "threads",
164
+ "learner": "threads"
165
+ }
166
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcd85218fb14c4783d7347a974ca2ef3cc2bb7ef6c24b1b9d051282ac7bed6fe
3
+ size 9371429196
policy_postprocessor.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_postprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "unnormalizer_processor",
6
+ "config": {
7
+ "eps": 1e-08,
8
+ "features": {
9
+ "action": {
10
+ "type": "ACTION",
11
+ "shape": [
12
+ 6
13
+ ]
14
+ }
15
+ },
16
+ "norm_map": {
17
+ "VISUAL": "IDENTITY",
18
+ "STATE": "MIN_MAX",
19
+ "ENV": "MIN_MAX",
20
+ "ACTION": "QUANTILES"
21
+ }
22
+ },
23
+ "state_file": "policy_postprocessor_step_0_unnormalizer_processor.safetensors"
24
+ },
25
+ {
26
+ "registry_name": "device_processor",
27
+ "config": {
28
+ "device": "cpu",
29
+ "float_dtype": null
30
+ }
31
+ }
32
+ ]
33
+ }
policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55c11e62ea79ad9d94d3940555548249042f3c145b7e37a78da5c1fba3d3af3d
3
+ size 14436
policy_preprocessor.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_preprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "rename_observations_processor",
6
+ "config": {
7
+ "rename_map": {}
8
+ }
9
+ },
10
+ {
11
+ "registry_name": "to_batch_processor",
12
+ "config": {}
13
+ },
14
+ {
15
+ "registry_name": "normalizer_processor",
16
+ "config": {
17
+ "eps": 1e-08,
18
+ "features": {
19
+ "observation.images.side": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 224,
24
+ 224
25
+ ]
26
+ },
27
+ "observation.images.top": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 224,
32
+ 224
33
+ ]
34
+ },
35
+ "observation.state": {
36
+ "type": "STATE",
37
+ "shape": [
38
+ 6
39
+ ]
40
+ },
41
+ "action": {
42
+ "type": "ACTION",
43
+ "shape": [
44
+ 6
45
+ ]
46
+ }
47
+ },
48
+ "norm_map": {
49
+ "VISUAL": "IDENTITY",
50
+ "STATE": "MIN_MAX",
51
+ "ENV": "MIN_MAX",
52
+ "ACTION": "QUANTILES"
53
+ }
54
+ },
55
+ "state_file": "policy_preprocessor_step_2_normalizer_processor.safetensors"
56
+ },
57
+ {
58
+ "registry_name": "pi05_full_prepare_state_tokenizer_processor_step",
59
+ "config": {}
60
+ },
61
+ {
62
+ "registry_name": "tokenizer_processor",
63
+ "config": {
64
+ "max_length": 64,
65
+ "task_key": "task",
66
+ "padding_side": "right",
67
+ "padding": "max_length",
68
+ "truncation": true,
69
+ "tokenizer_name": "google/paligemma-3b-pt-224"
70
+ }
71
+ },
72
+ {
73
+ "registry_name": "action_tokenizer_processor",
74
+ "config": {
75
+ "trust_remote_code": true,
76
+ "max_action_tokens": 256,
77
+ "action_tokenizer_name": "physical-intelligence/fast"
78
+ }
79
+ },
80
+ {
81
+ "registry_name": "device_processor",
82
+ "config": {
83
+ "device": "cuda",
84
+ "float_dtype": null
85
+ }
86
+ }
87
+ ]
88
+ }
policy_preprocessor_step_2_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55c11e62ea79ad9d94d3940555548249042f3c145b7e37a78da5c1fba3d3af3d
3
+ size 14436
train_config.json ADDED
@@ -0,0 +1,476 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "cijerezg/yellow-car-offline-training-v2",
4
+ "root": "outputs/jack_cube",
5
+ "episodes": null,
6
+ "max_episodes": null,
7
+ "image_transforms": {
8
+ "enable": false,
9
+ "max_num_transforms": 3,
10
+ "random_order": false,
11
+ "tfs": {
12
+ "brightness": {
13
+ "weight": 1.0,
14
+ "type": "ColorJitter",
15
+ "kwargs": {
16
+ "brightness": [
17
+ 0.8,
18
+ 1.2
19
+ ]
20
+ }
21
+ },
22
+ "contrast": {
23
+ "weight": 1.0,
24
+ "type": "ColorJitter",
25
+ "kwargs": {
26
+ "contrast": [
27
+ 0.8,
28
+ 1.2
29
+ ]
30
+ }
31
+ },
32
+ "saturation": {
33
+ "weight": 1.0,
34
+ "type": "ColorJitter",
35
+ "kwargs": {
36
+ "saturation": [
37
+ 0.5,
38
+ 1.5
39
+ ]
40
+ }
41
+ },
42
+ "hue": {
43
+ "weight": 1.0,
44
+ "type": "ColorJitter",
45
+ "kwargs": {
46
+ "hue": [
47
+ -0.05,
48
+ 0.05
49
+ ]
50
+ }
51
+ },
52
+ "sharpness": {
53
+ "weight": 1.0,
54
+ "type": "SharpnessJitter",
55
+ "kwargs": {
56
+ "sharpness": [
57
+ 0.5,
58
+ 1.5
59
+ ]
60
+ }
61
+ },
62
+ "affine": {
63
+ "weight": 1.0,
64
+ "type": "RandomAffine",
65
+ "kwargs": {
66
+ "degrees": [
67
+ -5.0,
68
+ 5.0
69
+ ],
70
+ "translate": [
71
+ 0.05,
72
+ 0.05
73
+ ]
74
+ }
75
+ }
76
+ }
77
+ },
78
+ "revision": null,
79
+ "use_imagenet_stats": false,
80
+ "video_backend": "pyav",
81
+ "return_uint8": false,
82
+ "streaming": false,
83
+ "additional_offline_dataset_paths": []
84
+ },
85
+ "env": {
86
+ "type": "gym_manipulator",
87
+ "task": "Pick up the red truck and put it in the bowl",
88
+ "fps": 30,
89
+ "features": {
90
+ "observation.images.side": {
91
+ "type": "VISUAL",
92
+ "shape": [
93
+ 3,
94
+ 128,
95
+ 128
96
+ ]
97
+ },
98
+ "observation.images.top": {
99
+ "type": "VISUAL",
100
+ "shape": [
101
+ 3,
102
+ 128,
103
+ 128
104
+ ]
105
+ },
106
+ "observation.state": {
107
+ "type": "STATE",
108
+ "shape": [
109
+ 6
110
+ ]
111
+ },
112
+ "action": {
113
+ "type": "ACTION",
114
+ "shape": [
115
+ 6
116
+ ]
117
+ }
118
+ },
119
+ "features_map": {
120
+ "observation.images.side": "observation.images.side",
121
+ "observation.images.top": "observation.images.top",
122
+ "observation.state": "observation.state",
123
+ "action": "action"
124
+ },
125
+ "max_parallel_tasks": 1,
126
+ "disable_env_checker": true,
127
+ "robot": {
128
+ "type": "so100_follower",
129
+ "port": "/dev/ttyACM0",
130
+ "disable_torque_on_disconnect": true,
131
+ "max_relative_target": null,
132
+ "cameras": {
133
+ "side": {
134
+ "type": "opencv",
135
+ "fps": 30,
136
+ "width": 640,
137
+ "height": 480,
138
+ "index_or_path": 0,
139
+ "color_mode": "rgb",
140
+ "rotation": 0,
141
+ "warmup_s": 1,
142
+ "fourcc": null,
143
+ "backend": 0
144
+ },
145
+ "top": {
146
+ "type": "opencv",
147
+ "fps": 30,
148
+ "width": 640,
149
+ "height": 480,
150
+ "index_or_path": 2,
151
+ "color_mode": "rgb",
152
+ "rotation": 0,
153
+ "warmup_s": 1,
154
+ "fourcc": null,
155
+ "backend": 0
156
+ }
157
+ },
158
+ "use_degrees": true,
159
+ "id": "follower_arm_v2",
160
+ "calibration_dir": null
161
+ },
162
+ "teleop": {
163
+ "type": "so100_leader",
164
+ "port": "/dev/ttyACM1",
165
+ "use_degrees": true,
166
+ "id": "leader_arm_v2",
167
+ "calibration_dir": null
168
+ },
169
+ "processor": {
170
+ "control_mode": "leader",
171
+ "observation": {
172
+ "add_joint_velocity_to_observation": false,
173
+ "add_current_to_observation": false,
174
+ "add_ee_pose_to_observation": false,
175
+ "display_cameras": false
176
+ },
177
+ "image_preprocessing": {
178
+ "crop_params_dict": null,
179
+ "resize_size": [
180
+ 224,
181
+ 224
182
+ ]
183
+ },
184
+ "gripper": {
185
+ "use_gripper": true,
186
+ "gripper_penalty": 0.0
187
+ },
188
+ "reset": {
189
+ "fixed_reset_joint_positions": [
190
+ 0.54,
191
+ -90.69,
192
+ 99.55,
193
+ 73.7,
194
+ -50.23,
195
+ 42.71
196
+ ],
197
+ "reset_time_s": 10.0,
198
+ "control_time_s": 200.0,
199
+ "terminate_on_success": true
200
+ },
201
+ "inverse_kinematics": null,
202
+ "reward_classifier": {
203
+ "pretrained_path": null,
204
+ "success_threshold": 0.5,
205
+ "success_reward": 1.0
206
+ },
207
+ "max_gripper_pos": 30.0
208
+ },
209
+ "name": "real_robot"
210
+ },
211
+ "policy": {
212
+ "type": "pi05_rl",
213
+ "n_obs_steps": 1,
214
+ "input_features": {
215
+ "observation.images.side": {
216
+ "type": "VISUAL",
217
+ "shape": [
218
+ 3,
219
+ 224,
220
+ 224
221
+ ]
222
+ },
223
+ "observation.images.top": {
224
+ "type": "VISUAL",
225
+ "shape": [
226
+ 3,
227
+ 224,
228
+ 224
229
+ ]
230
+ },
231
+ "observation.state": {
232
+ "type": "STATE",
233
+ "shape": [
234
+ 6
235
+ ]
236
+ }
237
+ },
238
+ "output_features": {
239
+ "action": {
240
+ "type": "ACTION",
241
+ "shape": [
242
+ 6
243
+ ]
244
+ }
245
+ },
246
+ "device": "cuda",
247
+ "use_amp": false,
248
+ "use_peft": false,
249
+ "push_to_hub": true,
250
+ "repo_id": "cijerezg/multi-task-toys-merged-v2",
251
+ "private": null,
252
+ "tags": null,
253
+ "license": null,
254
+ "pretrained_path": null,
255
+ "paligemma_variant": "gemma_2b",
256
+ "action_expert_variant": "gemma_300m",
257
+ "dtype": "bfloat16",
258
+ "chunk_size": 50,
259
+ "n_action_steps": 50,
260
+ "max_state_dim": 6,
261
+ "max_action_dim": 32,
262
+ "num_inference_steps": 5,
263
+ "time_sampling_beta_alpha": 1.5,
264
+ "time_sampling_beta_beta": 1.0,
265
+ "time_sampling_scale": 0.999,
266
+ "time_sampling_offset": 0.001,
267
+ "min_period": 0.004,
268
+ "max_period": 4.0,
269
+ "rtc_config": {
270
+ "enabled": true,
271
+ "prefix_attention_schedule": "LINEAR",
272
+ "max_guidance_weight": 10.0,
273
+ "execution_horizon": 10,
274
+ "debug": false,
275
+ "debug_maxlen": 100
276
+ },
277
+ "image_resolution": [
278
+ 224,
279
+ 224
280
+ ],
281
+ "empty_cameras": 0,
282
+ "use_dataset_stats": false,
283
+ "normalization_mapping": {
284
+ "VISUAL": "IDENTITY",
285
+ "STATE": "MIN_MAX",
286
+ "ENV": "MIN_MAX",
287
+ "ACTION": "QUANTILES"
288
+ },
289
+ "action_tokenizer_name": "physical-intelligence/fast",
290
+ "text_tokenizer_name": "google/paligemma-3b-pt-224",
291
+ "max_action_tokens": 256,
292
+ "fast_skip_tokens": 128,
293
+ "max_decoding_steps": 200,
294
+ "temperature": 0.0,
295
+ "subtask_regeneration_interval": 1.5,
296
+ "gradient_checkpointing": true,
297
+ "compile_model": false,
298
+ "compile_mode": "max-autotune",
299
+ "freeze_vision_encoder": false,
300
+ "train_expert_only": false,
301
+ "knowledge_insulation": true,
302
+ "action_encoding": "anchor",
303
+ "loss_weight_flow": 1.0,
304
+ "loss_weight_action_ce": 1.0,
305
+ "loss_weight_subtask_ce": 1.0,
306
+ "optimizer_lr": 2.5e-05,
307
+ "optimizer_betas": [
308
+ 0.9,
309
+ 0.95
310
+ ],
311
+ "optimizer_eps": 1e-08,
312
+ "optimizer_weight_decay": 0.1,
313
+ "optimizer_grad_clip_norm": 1.0,
314
+ "scheduler_warmup_steps": 1000,
315
+ "scheduler_decay_steps": 30000,
316
+ "scheduler_decay_lr": 2.5e-06,
317
+ "tokenizer_max_length": 64,
318
+ "task": "Pick up the orange cube and place it on the black X marker",
319
+ "action_dim": 6,
320
+ "drop_n_last_frames": 2,
321
+ "critic_target_update_weight": 0.005,
322
+ "num_critics": 1,
323
+ "discount": 0.97,
324
+ "reward_normalization_constant": 5.0,
325
+ "terminal_failure_reward": -16.0,
326
+ "online_steps": 20000,
327
+ "online_buffer_capacity": 5000,
328
+ "offline_buffer_capacity": 50000,
329
+ "async_prefetch": false,
330
+ "online_step_before_learning": 10,
331
+ "policy_update_freq": 1,
332
+ "grad_clip_norm": 2.0,
333
+ "gradient_accumulation_steps": 16,
334
+ "critic_lr": 5e-05,
335
+ "actor_lr": 5e-05,
336
+ "utd_ratio": 1,
337
+ "actor_device": "cuda:0",
338
+ "learner_device": "cuda:0",
339
+ "use_separate_critic": true,
340
+ "critic_llm_depth": 6,
341
+ "critic_network_kwargs": {
342
+ "hidden_dims": [
343
+ 256,
344
+ 256
345
+ ],
346
+ "activate_final": true
347
+ },
348
+ "trainable_params": {
349
+ "vision_encoder_from_layer": {
350
+ "vision_tower": 5,
351
+ "multi_modal_projector": true
352
+ },
353
+ "language_from_layer": 0,
354
+ "critic_language_from_layer": 5
355
+ },
356
+ "offline_steps": 10000,
357
+ "inference_advantage": 1.0,
358
+ "advantage_scaling": 0.2,
359
+ "pi05_checkpoint": "outputs/pi05_base",
360
+ "action_encoding_stats_path": "outputs/stats_jack/action_stats_anchor_jack_cube.pt",
361
+ "dataset_stats": null,
362
+ "storage_device": "cpu",
363
+ "shared_encoder": false,
364
+ "num_discrete_actions": null,
365
+ "vision_encoder_name": null,
366
+ "actor_learner_config": {
367
+ "learner_host": "192.168.50.1",
368
+ "learner_port": 50051,
369
+ "policy_parameters_push_frequency": 180,
370
+ "queue_get_timeout": 2
371
+ },
372
+ "concurrency": {
373
+ "actor": "threads",
374
+ "learner": "threads"
375
+ }
376
+ },
377
+ "reward_model": null,
378
+ "output_dir": "outputs/jack_pi05_full_offline_training_val_anchor_action_v1",
379
+ "job_name": "default",
380
+ "resume": false,
381
+ "seed": 42,
382
+ "cudnn_deterministic": false,
383
+ "num_workers": 4,
384
+ "batch_size": 8,
385
+ "prefetch_factor": 4,
386
+ "persistent_workers": true,
387
+ "steps": 100000,
388
+ "eval_freq": 20000,
389
+ "log_freq": 20,
390
+ "tolerance_s": 0.0001,
391
+ "save_checkpoint": true,
392
+ "save_freq": 100,
393
+ "use_policy_training_preset": true,
394
+ "optimizer": {
395
+ "type": "multi_adam",
396
+ "lr": 0.001,
397
+ "weight_decay": 0.1,
398
+ "grad_clip_norm": 10.0,
399
+ "optimizer_groups": {
400
+ "actor": {
401
+ "lr": 5e-05
402
+ },
403
+ "critic": {
404
+ "lr": 5e-05
405
+ }
406
+ }
407
+ },
408
+ "scheduler": null,
409
+ "eval": {
410
+ "n_episodes": 50,
411
+ "batch_size": 22,
412
+ "use_async_envs": true
413
+ },
414
+ "wandb": {
415
+ "enable": true,
416
+ "disable_artifact": true,
417
+ "project": "so101_real_offline-v1",
418
+ "entity": null,
419
+ "notes": null,
420
+ "run_id": "bfi8fh35",
421
+ "mode": null,
422
+ "offline_project": "so101_real_offline-v1",
423
+ "add_tags": true
424
+ },
425
+ "peft": null,
426
+ "sample_weighting": null,
427
+ "rename_map": {},
428
+ "checkpoint_path": null,
429
+ "offline_output_dir": "outputs/jack_pi05_full_offline_training_val_anchor_action_v1",
430
+ "offline_save_freq": 400,
431
+ "buffer_cache_dir": null,
432
+ "use_rerun": true,
433
+ "video_logging_cameras": [
434
+ "top",
435
+ "wrist"
436
+ ],
437
+ "episode_logging_freq": 4,
438
+ "episode_save_freq": 10,
439
+ "probe_parameters": {
440
+ "enable_actions": true,
441
+ "enable_representations": true,
442
+ "enable_attention": true,
443
+ "enable_offline_inference": true,
444
+ "enable_spatial_memorization": true,
445
+ "enable_action_drift_jacobian": true,
446
+ "enable_spatial_memorization_jacobian": true,
447
+ "output_dir": "outputs/probe",
448
+ "mode": "all",
449
+ "max_episodes": 1,
450
+ "n_frames_per_episode": 128,
451
+ "offline_inference_n_frames": 5,
452
+ "random_seed": 42,
453
+ "timestep": 0.5,
454
+ "ref_max_episodes": 2,
455
+ "ref_n_frames_per_episode": 256,
456
+ "action_pca_dims": 50,
457
+ "repr_pca_dims": 100,
458
+ "umap_n_neighbors": 15,
459
+ "umap_min_dist": 0.1,
460
+ "umap_seed": 42,
461
+ "sites": "prefix,suffix",
462
+ "ep_3d_a": 0,
463
+ "ep_3d_b": 1,
464
+ "subtask_injection": false,
465
+ "validation_batch_size": 32,
466
+ "attn_eval_episodes": null,
467
+ "attn_eval_subsample": 2,
468
+ "spatial_layers": "0,9,17",
469
+ "spatial_n_frames": 32
470
+ },
471
+ "val_dataset_path": "outputs/annotated_dataset_validation",
472
+ "val_split": 0.0,
473
+ "val_freq": 400,
474
+ "val_on_start": false,
475
+ "skip_critic": true
476
+ }