ducido commited on
Commit
a15f191
·
verified ·
1 Parent(s): 6075f22

Initial commit

Browse files
.gitattributes CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ run-20260125_043035-g0u01bsh/files/output.log filter=lfs diff=lfs merge=lfs -text
37
+ run-20260125_043035-g0u01bsh/run-g0u01bsh.wandb filter=lfs diff=lfs merge=lfs -text
debug-internal.log ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-01-25T04:30:35.337188511Z","level":"INFO","msg":"stream: starting","core version":"0.21.4"}
2
+ {"time":"2026-01-25T04:30:35.352330898Z","level":"ERROR","msg":"monitor: failed to initialize GPU resource: monitor: could not create portfile"}
3
+ {"time":"2026-01-25T04:30:35.516712677Z","level":"INFO","msg":"stream: created new stream","id":"g0u01bsh"}
4
+ {"time":"2026-01-25T04:30:35.516771215Z","level":"INFO","msg":"stream: started","id":"g0u01bsh"}
5
+ {"time":"2026-01-25T04:30:35.516797314Z","level":"INFO","msg":"handler: started","stream_id":"g0u01bsh"}
6
+ {"time":"2026-01-25T04:30:35.51678996Z","level":"INFO","msg":"writer: started","stream_id":"g0u01bsh"}
7
+ {"time":"2026-01-25T04:30:35.516816019Z","level":"INFO","msg":"sender: started","stream_id":"g0u01bsh"}
8
+ {"time":"2026-01-25T12:11:53.737869083Z","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/graphql","body":"\n<html><head>\n<meta http-equiv=\"content-type\" content=\"text/html;charset=utf-8\">\n<title>502 Server Error</title>\n</head>\n<body text=#000000 bgcolor=#ffffff>\n<h1>Error: Server Error</h1>\n<h2>The server encountered a temporary error and could not complete your request.<p>Please try again in 30 seconds.</h2>\n<h2></h2>\n</body></html>\n"}
9
+ {"time":"2026-01-26T03:42:47.247976354Z","level":"INFO","msg":"stream: closing","id":"g0u01bsh"}
10
+ {"time":"2026-01-26T03:43:02.531387332Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
11
+ {"time":"2026-01-26T03:43:03.114011174Z","level":"INFO","msg":"handler: closed","stream_id":"g0u01bsh"}
12
+ {"time":"2026-01-26T03:43:03.116713969Z","level":"INFO","msg":"sender: closed","stream_id":"g0u01bsh"}
13
+ {"time":"2026-01-26T03:43:03.116749606Z","level":"INFO","msg":"stream: closed","id":"g0u01bsh"}
debug.log ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Current SDK version is 0.21.4
2
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Configure stats pid to 1165115
3
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Loading settings from /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/.config/wandb/settings
4
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Loading settings from /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/wandb/settings
5
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Loading settings from environment variables
6
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:setup_run_log_directory():686] Logging user logs to /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN/wandb/run-20260125_043035-g0u01bsh/logs/debug.log
7
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:setup_run_log_directory():687] Logging internal logs to /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN/wandb/run-20260125_043035-g0u01bsh/logs/debug-internal.log
8
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:init():813] calling init triggers
9
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:init():818] wandb.init called with sweep_config: {}
10
+ config: {'dataset': {'repo_id': 'ducido/calvin_task_D_D_scale_100_lerobo_format', 'root': '/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/ALOHA/aloha_3_tasks_large_100_original_synthetic_high_view', 'episodes': None, 'image_transforms': {'enable': True, 'max_num_transforms': 3, 'random_order': False, 'image_tfs': {'hue': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'hue': [-0.05, 0.05]}}, 'contrast': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'contrast': [0.8, 1.2]}}, 'sharpness': {'weight': 1.0, 'type': 'SharpnessJitter', 'kwargs': {'sharpness': [0.5, 1.5]}}, 'brightness': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'brightness': [0.8, 1.2]}}, 'saturation': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'saturation': [0.5, 1.5]}}, 'crop_resize': {'weight': 1.0, 'type': 'RandomResizedCrop', 'kwargs': {'size': [256, 256], 'ratio': [1, 1], 'scale': [0.9, 0.95]}}, 'rotate': {'weight': 1.0, 'type': 'RandomRotate', 'kwargs': {'degrees': [-5, 5]}}}, 'wrist_tfs': {'hue': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'hue': [-0.05, 0.05]}}, 'contrast': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'contrast': [0.8, 1.2]}}, 'crop_resize': {'weight': 1.0, 'type': 'RandomResizedCrop', 'kwargs': {'size': [256, 256], 'ratio': [1, 1], 'scale': [0.9, 0.95]}}, 'sharpness': {'weight': 1.0, 'type': 'SharpnessJitter', 'kwargs': {'sharpness': [0.5, 1.5]}}, 'brightness': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'brightness': [0.8, 1.2]}}, 'saturation': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'saturation': [0.5, 1.5]}}}}, 'revision': None, 'use_imagenet_stats': True, 'video_backend': 'torchcodec', 'vqa_data_path': None}, 'env': None, 'policy': {'type': 'pi0', 'n_obs_steps': 1, 'normalization_mapping': {'VISUAL': <NormalizationMode.IDENTITY: 'IDENTITY'>, 'STATE': <NormalizationMode.MEAN_STD: 'MEAN_STD'>, 'ACTION': <NormalizationMode.MEAN_STD: 'MEAN_STD'>}, 'input_features': {}, 'output_features': {}, 'device': 'cuda', 'use_amp': False, 'gradient_accumulation_steps': 2, 'chunk_size': 50, 'n_action_steps': 50, 'visual_prompt': False, 'use_both_origin': False, 'zero_init': False, 'coordinate_text_prod': False, 'coordinate_posEmbed_prod': False, 'posEmbed_to_preTrained': False, 'num_obs_layers': 30, 'last_obs_layers': False, 'add_posEmbed': False, 'not_used_coord': False, 'use_depth': False, 'depth_model_type': 'frozen_siglip', 'use_track': False, 'use_detect_object': False, 'track_mask': False, 'use_slot_att': 'none', 'future_obj': 'none', 'original_pi0': False, 'max_state_dim': 32, 'max_action_dim': 32, 'resize_imgs_with_padding': [224, 224], 'empty_cameras': 0, 'adapt_to_pi_aloha': False, 'use_delta_joint_actions_aloha': False, 'tokenizer_max_length': 48, 'proj_width': 1024, 'num_steps': 10, 'use_cache': True, 'attention_implementation': 'eager', 'freeze_vision_encoder': True, 'train_expert_only': False, 'train_state_proj': True, 'optimizer_lr': 0.0001, 'optimizer_betas': [0.9, 0.95], 'optimizer_eps': 1e-08, 'optimizer_weight_decay': 1e-10, 'scheduler_warmup_steps': 1000, 'scheduler_decay_steps': 240000, 'scheduler_decay_lr': 2.5e-06}, 'output_dir': '/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN', 'job_name': 'aloha_100%_merged_synthetic_high_view_PRETRAIN', 'visual_prompt': False, 'use_both_origin': False, 'zero_init': False, 'coordinate_text_prod': False, 'coordinate_posEmbed_prod': False, 'posEmbed_to_preTrained': False, 'num_obs_layers': 30, 'last_obs_layers': False, 'add_posEmbed': False, 'not_used_coord': False, 'train_expert_only': False, 'use_depth': False, 'use_track': True, 'use_detect_object': False, 'track_mask': False, 'use_slot_att': 'none', 'future_obj': 'all', 'depth_model_type': 'frozen_siglip', 'original_pi0': False, 'resume': False, 'seed': 42, 'num_workers': 8, 'batch_size': 13, 'steps': 100000, 'eval_freq': 20000, 'log_freq': 200, 'save_checkpoint': True, 'save_freq': 10000, 'use_policy_training_preset': True, 'optimizer': {'type': 'adamw', 'lr': 0.0001, 'weight_decay': 1e-10, 'grad_clip_norm': 10.0, 'betas': [0.9, 0.95], 'eps': 1e-08}, 'scheduler': {'type': 'cosine_decay_with_warmup', 'num_warmup_steps': 1000, 'num_decay_steps': 240000, 'peak_lr': 0.0001, 'decay_lr': 2.5e-06}, 'eval': {'n_episodes': 50, 'batch_size': 50, 'use_async_envs': False}, 'wandb': {'enable': True, 'disable_artifact': True, 'project': 'pi0_lerobot', 'entity': 'Robotics_VLA', 'notes': None, 'run_id': None, 'mode': 'online'}, '_wandb': {}}
11
+ 2026-01-25 04:30:35,122 INFO MainThread:1165115 [wandb_init.py:init():854] starting backend
12
+ 2026-01-25 04:30:35,327 INFO MainThread:1165115 [wandb_init.py:init():857] sending inform_init request
13
+ 2026-01-25 04:30:35,332 INFO MainThread:1165115 [wandb_init.py:init():865] backend started and connected
14
+ 2026-01-25 04:30:35,334 INFO MainThread:1165115 [wandb_init.py:init():936] updated telemetry
15
+ 2026-01-25 04:30:35,344 INFO MainThread:1165115 [wandb_init.py:init():960] communicating run to backend with 90.0 second timeout
16
+ 2026-01-25 04:30:35,923 INFO MainThread:1165115 [wandb_init.py:init():1011] starting run threads in backend
17
+ 2026-01-25 04:30:36,126 INFO MainThread:1165115 [wandb_run.py:_console_start():2506] atexit reg
18
+ 2026-01-25 04:30:36,127 INFO MainThread:1165115 [wandb_run.py:_redirect():2354] redirect: wrap_raw
19
+ 2026-01-25 04:30:36,127 INFO MainThread:1165115 [wandb_run.py:_redirect():2423] Wrapping output streams.
20
+ 2026-01-25 04:30:36,127 INFO MainThread:1165115 [wandb_run.py:_redirect():2446] Redirects installed.
21
+ 2026-01-25 04:30:36,129 INFO MainThread:1165115 [wandb_init.py:init():1049] run started, returning control to user process
22
+ 2026-01-26 03:42:47,248 INFO wandb-AsyncioManager-main:1165115 [service_client.py:_forward_responses():84] Reached EOF.
23
+ 2026-01-26 03:42:47,248 INFO wandb-AsyncioManager-main:1165115 [mailbox.py:close():137] Closing mailbox, abandoning 1 handles.
run-20260125_043035-g0u01bsh/files/config.yaml ADDED
@@ -0,0 +1,353 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ _wandb:
2
+ value:
3
+ cli_version: 0.21.4
4
+ e:
5
+ xfmd715jagdfpknw5q3bwhtgeese05na:
6
+ args:
7
+ - --policy.path=/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/temp/baseline
8
+ - --dataset.root=/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/ALOHA/aloha_3_tasks_large_100_original_synthetic_high_view
9
+ - --output_dir=/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN
10
+ - --job_name=aloha_100%_merged_synthetic_high_view_PRETRAIN
11
+ - --config_path=configs/libero_config/default.json
12
+ - --batch_size=13
13
+ - --policy.gradient_accumulation_steps=2
14
+ - --save_freq=10000
15
+ - --wandb.mode=online
16
+ - --visual_prompt=false
17
+ - --zero_init=false
18
+ - --use_both_origin=false
19
+ - --coordinate_text_prod=false
20
+ - --coordinate_posEmbed_prod=false
21
+ - --posEmbed_to_preTrained=false
22
+ - --add_posEmbed=false
23
+ - --not_used_coord=false
24
+ - --train_expert_only=false
25
+ - --use_depth=false
26
+ - --depth_model_type=frozen_siglip
27
+ - --use_track=true
28
+ - --use_detect_object=false
29
+ - --track_mask=false
30
+ - --future_obj=all
31
+ - --use_slot_att=none
32
+ - --original_pi0=false
33
+ codePath: lerobot/scripts/train_accelerate.py
34
+ codePathLocal: lerobot/scripts/train_accelerate.py
35
+ cpu_count: 32
36
+ cpu_count_logical: 32
37
+ disk:
38
+ /:
39
+ total: "10737418240"
40
+ used: "788828160"
41
+ email: ducido.w@gmail.com
42
+ executable: /usr/bin/python
43
+ git:
44
+ commit: f32c82f37b2b6f59a8b2c1e8367d7d70b203e23b
45
+ remote: https://github.com/jibby2803/VLA_custom
46
+ host: 085ccf349861
47
+ memory:
48
+ total: "2163816681472"
49
+ os: Linux-4.18.0-513.5.1.el8_9.x86_64-x86_64-with-glibc2.35
50
+ program: /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/lerobot/scripts/train_accelerate.py
51
+ python: CPython 3.10.12
52
+ root: /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN
53
+ startedAt: "2026-01-25T04:30:35.116725Z"
54
+ writerId: xfmd715jagdfpknw5q3bwhtgeese05na
55
+ m: []
56
+ python_version: 3.10.12
57
+ t:
58
+ "1":
59
+ - 1
60
+ - 41
61
+ - 49
62
+ - 51
63
+ - 71
64
+ "2":
65
+ - 1
66
+ - 5
67
+ - 6
68
+ - 11
69
+ - 41
70
+ - 49
71
+ - 51
72
+ - 53
73
+ - 63
74
+ - 71
75
+ - 98
76
+ "3":
77
+ - 13
78
+ - 15
79
+ - 16
80
+ - 61
81
+ "4": 3.10.12
82
+ "5": 0.21.4
83
+ "12": 0.21.4
84
+ "13": linux-x86_64
85
+ add_posEmbed:
86
+ value: false
87
+ batch_size:
88
+ value: 13
89
+ coordinate_posEmbed_prod:
90
+ value: false
91
+ coordinate_text_prod:
92
+ value: false
93
+ dataset:
94
+ value:
95
+ episodes: null
96
+ image_transforms:
97
+ enable: true
98
+ image_tfs:
99
+ brightness:
100
+ kwargs:
101
+ brightness:
102
+ - 0.8
103
+ - 1.2
104
+ type: ColorJitter
105
+ weight: 1
106
+ contrast:
107
+ kwargs:
108
+ contrast:
109
+ - 0.8
110
+ - 1.2
111
+ type: ColorJitter
112
+ weight: 1
113
+ crop_resize:
114
+ kwargs:
115
+ ratio:
116
+ - 1
117
+ - 1
118
+ scale:
119
+ - 0.9
120
+ - 0.95
121
+ size:
122
+ - 256
123
+ - 256
124
+ type: RandomResizedCrop
125
+ weight: 1
126
+ hue:
127
+ kwargs:
128
+ hue:
129
+ - -0.05
130
+ - 0.05
131
+ type: ColorJitter
132
+ weight: 1
133
+ rotate:
134
+ kwargs:
135
+ degrees:
136
+ - -5
137
+ - 5
138
+ type: RandomRotate
139
+ weight: 1
140
+ saturation:
141
+ kwargs:
142
+ saturation:
143
+ - 0.5
144
+ - 1.5
145
+ type: ColorJitter
146
+ weight: 1
147
+ sharpness:
148
+ kwargs:
149
+ sharpness:
150
+ - 0.5
151
+ - 1.5
152
+ type: SharpnessJitter
153
+ weight: 1
154
+ max_num_transforms: 3
155
+ random_order: false
156
+ wrist_tfs:
157
+ brightness:
158
+ kwargs:
159
+ brightness:
160
+ - 0.8
161
+ - 1.2
162
+ type: ColorJitter
163
+ weight: 1
164
+ contrast:
165
+ kwargs:
166
+ contrast:
167
+ - 0.8
168
+ - 1.2
169
+ type: ColorJitter
170
+ weight: 1
171
+ crop_resize:
172
+ kwargs:
173
+ ratio:
174
+ - 1
175
+ - 1
176
+ scale:
177
+ - 0.9
178
+ - 0.95
179
+ size:
180
+ - 256
181
+ - 256
182
+ type: RandomResizedCrop
183
+ weight: 1
184
+ hue:
185
+ kwargs:
186
+ hue:
187
+ - -0.05
188
+ - 0.05
189
+ type: ColorJitter
190
+ weight: 1
191
+ saturation:
192
+ kwargs:
193
+ saturation:
194
+ - 0.5
195
+ - 1.5
196
+ type: ColorJitter
197
+ weight: 1
198
+ sharpness:
199
+ kwargs:
200
+ sharpness:
201
+ - 0.5
202
+ - 1.5
203
+ type: SharpnessJitter
204
+ weight: 1
205
+ repo_id: ducido/calvin_task_D_D_scale_100_lerobo_format
206
+ revision: null
207
+ root: /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/ALOHA/aloha_3_tasks_large_100_original_synthetic_high_view
208
+ use_imagenet_stats: true
209
+ video_backend: torchcodec
210
+ vqa_data_path: null
211
+ depth_model_type:
212
+ value: frozen_siglip
213
+ env:
214
+ value: null
215
+ eval:
216
+ value:
217
+ batch_size: 50
218
+ n_episodes: 50
219
+ use_async_envs: false
220
+ eval_freq:
221
+ value: 20000
222
+ future_obj:
223
+ value: all
224
+ job_name:
225
+ value: aloha_100%_merged_synthetic_high_view_PRETRAIN
226
+ last_obs_layers:
227
+ value: false
228
+ log_freq:
229
+ value: 200
230
+ not_used_coord:
231
+ value: false
232
+ num_obs_layers:
233
+ value: 30
234
+ num_workers:
235
+ value: 8
236
+ optimizer:
237
+ value:
238
+ betas:
239
+ - 0.9
240
+ - 0.95
241
+ eps: 1e-08
242
+ grad_clip_norm: 10
243
+ lr: 0.0001
244
+ type: adamw
245
+ weight_decay: 1e-10
246
+ original_pi0:
247
+ value: false
248
+ output_dir:
249
+ value: /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN
250
+ policy:
251
+ value:
252
+ adapt_to_pi_aloha: false
253
+ add_posEmbed: false
254
+ attention_implementation: eager
255
+ chunk_size: 50
256
+ coordinate_posEmbed_prod: false
257
+ coordinate_text_prod: false
258
+ depth_model_type: frozen_siglip
259
+ device: cuda
260
+ empty_cameras: 0
261
+ freeze_vision_encoder: true
262
+ future_obj: none
263
+ gradient_accumulation_steps: 2
264
+ last_obs_layers: false
265
+ max_action_dim: 32
266
+ max_state_dim: 32
267
+ n_action_steps: 50
268
+ n_obs_steps: 1
269
+ normalization_mapping:
270
+ ACTION: MEAN_STD
271
+ STATE: MEAN_STD
272
+ VISUAL: IDENTITY
273
+ not_used_coord: false
274
+ num_obs_layers: 30
275
+ num_steps: 10
276
+ optimizer_betas:
277
+ - 0.9
278
+ - 0.95
279
+ optimizer_eps: 1e-08
280
+ optimizer_lr: 0.0001
281
+ optimizer_weight_decay: 1e-10
282
+ original_pi0: false
283
+ posEmbed_to_preTrained: false
284
+ proj_width: 1024
285
+ resize_imgs_with_padding:
286
+ - 224
287
+ - 224
288
+ scheduler_decay_lr: 2.5e-06
289
+ scheduler_decay_steps: 240000
290
+ scheduler_warmup_steps: 1000
291
+ tokenizer_max_length: 48
292
+ track_mask: false
293
+ train_expert_only: false
294
+ train_state_proj: true
295
+ type: pi0
296
+ use_amp: false
297
+ use_both_origin: false
298
+ use_cache: true
299
+ use_delta_joint_actions_aloha: false
300
+ use_depth: false
301
+ use_detect_object: false
302
+ use_slot_att: none
303
+ use_track: false
304
+ visual_prompt: false
305
+ zero_init: false
306
+ posEmbed_to_preTrained:
307
+ value: false
308
+ resume:
309
+ value: false
310
+ save_checkpoint:
311
+ value: true
312
+ save_freq:
313
+ value: 10000
314
+ scheduler:
315
+ value:
316
+ decay_lr: 2.5e-06
317
+ num_decay_steps: 240000
318
+ num_warmup_steps: 1000
319
+ peak_lr: 0.0001
320
+ type: cosine_decay_with_warmup
321
+ seed:
322
+ value: 42
323
+ steps:
324
+ value: 100000
325
+ track_mask:
326
+ value: false
327
+ train_expert_only:
328
+ value: false
329
+ use_both_origin:
330
+ value: false
331
+ use_depth:
332
+ value: false
333
+ use_detect_object:
334
+ value: false
335
+ use_policy_training_preset:
336
+ value: true
337
+ use_slot_att:
338
+ value: none
339
+ use_track:
340
+ value: true
341
+ visual_prompt:
342
+ value: false
343
+ wandb:
344
+ value:
345
+ disable_artifact: true
346
+ enable: true
347
+ entity: Robotics_VLA
348
+ mode: online
349
+ notes: null
350
+ project: pi0_lerobot
351
+ run_id: null
352
+ zero_init:
353
+ value: false
run-20260125_043035-g0u01bsh/files/output.log ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e36d42bf36f483f2544ef9be82b7a6dba7184007b1a2453f57238c8f982862e3
3
+ size 112715330
run-20260125_043035-g0u01bsh/files/requirements.txt ADDED
@@ -0,0 +1,426 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ determined==0.35.0
2
+ nvidia-ml-py==13.590.44
3
+ wcwidth==0.2.14
4
+ psutil==7.2.1
5
+ blessed==1.25.0
6
+ gpustat==1.1.1
7
+ h5py==3.15.1
8
+ decord==0.6.0
9
+ flash_attn==2.5.8
10
+ timm==1.0.24
11
+ slot_attention==1.4.0
12
+ peft==0.18.1
13
+ quadprog==0.1.13
14
+ qpsolvers==4.8.2
15
+ mink==0.0.5
16
+ hidapi==0.15.0
17
+ pygame==2.6.1
18
+ protobuf==3.19.6
19
+ numpy==1.23.3
20
+ lxml==6.0.2
21
+ llvmlite==0.39.1
22
+ numba==0.56.4
23
+ tianshou==0.4.10
24
+ robocasa==0.2.0
25
+ mujoco==3.2.6
26
+ PyOpenGL==3.1.4
27
+ opencv-python==4.8.0.74
28
+ bddl==3.6.0
29
+ docstring_parser==0.17.0
30
+ easydict==1.13
31
+ etils==1.13.0
32
+ glfw==2.10.0
33
+ gym==0.26.2
34
+ gym-notices==0.1.0
35
+ importlib_resources==6.5.2
36
+ mujoco==3.4.0
37
+ nltk==3.9.2
38
+ numpy==1.26.4
39
+ robosuite==1.4.0
40
+ typeguard==4.4.4
41
+ tyro==1.0.3
42
+ datasets==3.0.0
43
+ deepspeed==0.14.4
44
+ dill==0.3.8
45
+ iso8601==2.1.0
46
+ nvitop==1.6.1
47
+ python-dotenv==1.2.1
48
+ serial==0.0.97
49
+ tokenizers==0.21.4
50
+ transformers==4.48.1
51
+ Farama-Notifications==0.0.4
52
+ accelerate==1.12.0
53
+ av==15.1.0
54
+ cmake==4.1.3
55
+ deepdiff==8.6.1
56
+ diffusers==0.35.2
57
+ draccus==0.10.0
58
+ einops==0.8.1
59
+ evdev==1.9.2
60
+ gymnasium==1.2.3
61
+ hf_transfer==0.1.9
62
+ hf-xet==1.2.0
63
+ huggingface-hub==0.35.3
64
+ ImageIO==2.37.2
65
+ imageio-ffmpeg==0.6.0
66
+ inquirerpy==0.3.4
67
+ jsonlines==4.0.0
68
+ lerobot==0.4.3
69
+ mergedeep==1.3.4
70
+ multiprocess==0.70.16
71
+ mypy_extensions==1.1.0
72
+ nvidia-cublas-cu12==12.6.4.1
73
+ nvidia-cuda-cupti-cu12==12.6.80
74
+ nvidia-cuda-nvrtc-cu12==12.6.77
75
+ nvidia-cuda-runtime-cu12==12.6.77
76
+ nvidia-cudnn-cu12==9.5.1.17
77
+ nvidia-cufft-cu12==11.3.0.4
78
+ nvidia-cufile-cu12==1.11.1.6
79
+ nvidia-curand-cu12==10.3.7.77
80
+ nvidia-cusolver-cu12==11.7.1.2
81
+ nvidia-cusparse-cu12==12.5.4.2
82
+ nvidia-cusparselt-cu12==0.6.3
83
+ nvidia-nccl-cu12==2.26.2
84
+ nvidia-nvjitlink-cu12==12.6.85
85
+ nvidia-nvtx-cu12==12.6.77
86
+ opencv-python-headless==4.12.0.88
87
+ orderly-set==5.5.0
88
+ packaging==25.0
89
+ pfzy==0.3.4
90
+ pyarrow==22.0.0
91
+ pynput==1.8.1
92
+ pyserial==3.5
93
+ python-xlib==0.33
94
+ pyyaml-include==1.4.1
95
+ requests==2.32.5
96
+ rerun-sdk==0.26.2
97
+ safetensors==0.7.0
98
+ sentry-sdk==2.48.0
99
+ setuptools==80.9.0
100
+ sympy==1.14.0
101
+ torch==2.7.1
102
+ torchcodec==0.5
103
+ torchvision==0.22.1
104
+ tqdm==4.67.1
105
+ triton==3.3.1
106
+ typing-inspect==0.9.0
107
+ wandb==0.21.4
108
+ xxhash==3.6.0
109
+ pip==25.3
110
+ hjson==3.1.0
111
+ py-cpuinfo==9.0.0
112
+ pydantic==1.10.11
113
+ anyio==4.11.0
114
+ arrow==1.4.0
115
+ async-lru==2.0.5
116
+ babel==2.17.0
117
+ fqdn==1.5.1
118
+ ipywidgets==8.1.8
119
+ isoduration==20.11.0
120
+ jsonpointer==3.0.0
121
+ jupyter-archive==3.4.0
122
+ jupyter-events==0.12.0
123
+ jupyter-lsp==2.3.0
124
+ jupyter_server==2.17.0
125
+ jupyter_server_terminals==0.5.3
126
+ jupyterlab==4.0.13
127
+ jupyterlab_server==2.28.0
128
+ jupyterlab_widgets==3.0.16
129
+ notebook==7.0.8
130
+ notebook_shim==0.2.4
131
+ overrides==7.7.0
132
+ python-json-logger==4.0.0
133
+ rfc3339-validator==0.1.4
134
+ rfc3986-validator==0.1.1
135
+ sniffio==1.3.1
136
+ tzdata==2025.2
137
+ uri-template==1.3.0
138
+ webcolors==25.10.0
139
+ websocket-client==1.9.0
140
+ widgetsnbextension==4.0.15
141
+ Keras-Preprocessing==1.1.2
142
+ Pillow==9.5.0
143
+ attrdict3==2.0.2
144
+ future==1.0.0
145
+ nvidia-ml-py==13.580.82
146
+ petname==2.6
147
+ promise==2.3
148
+ protobuf==3.20.3
149
+ tensorflow-datasets==1.3.2
150
+ tensorflow-metadata==1.16.1
151
+ torch-tb-profiler==0.4.3
152
+ wrapt==2.0.0
153
+ analytics-python==1.4.post1
154
+ appdirs==1.4.4
155
+ argcomplete==3.6.3
156
+ azure-core==1.36.0
157
+ azure-storage-blob==12.27.1
158
+ backoff==1.10.0
159
+ bcrypt==5.0.0
160
+ boto3==1.40.67
161
+ botocore==1.40.67
162
+ cffi==2.0.0
163
+ cryptography==46.0.3
164
+ docker==7.1.0
165
+ gitdb==4.0.12
166
+ GitPython==3.1.45
167
+ google-api-core==2.28.1
168
+ google-api-python-client==2.187.0
169
+ google-auth-httplib2==0.2.1
170
+ google-cloud-core==2.5.0
171
+ google-cloud-storage==3.5.0
172
+ google-crc32c==1.7.1
173
+ google-resumable-media==2.7.2
174
+ googleapis-common-protos==1.71.0
175
+ httplib2==0.31.0
176
+ invoke==2.2.1
177
+ isodate==0.7.2
178
+ jmespath==1.0.1
179
+ lomond==0.3.3
180
+ monotonic==1.6
181
+ paramiko==4.0.0
182
+ pathspec==0.12.1
183
+ proto-plus==1.26.1
184
+ PyNaCl==1.6.0
185
+ pyOpenSSL==25.3.0
186
+ ruamel.yaml==0.18.16
187
+ ruamel.yaml.clib==0.2.14
188
+ s3transfer==0.14.0
189
+ smmap==5.0.2
190
+ termcolor==3.2.0
191
+ typing_extensions==4.15.0
192
+ uritemplate==4.2.0
193
+ transformer-engine==1.4.0+0fbc76a
194
+ flash-attn==2.4.2
195
+ torch-tensorrt==2.3.0a0
196
+ nvidia-pyindex==1.0.9
197
+ polygraphy==0.49.7
198
+ prettytable==3.10.0
199
+ pytorch-quantization==2.1.2
200
+ sphinx_glpi_theme==0.6
201
+ graphsurgeon==0.4.6
202
+ tensorrt==8.6.3
203
+ uff==0.6.9
204
+ aiohttp==3.9.3
205
+ aiosignal==1.3.1
206
+ async-timeout==4.0.3
207
+ cloudpickle==3.0.0
208
+ cuda-python==12.4.0rc7+3.ge75c8a9.dirty
209
+ cudf==24.2.0
210
+ cugraph==24.2.0
211
+ cugraph-dgl==24.2.0
212
+ cugraph-service-client==24.2.0
213
+ cugraph-service-server==24.2.0
214
+ cuml==24.2.0
215
+ cupy-cuda12x==13.0.0
216
+ dask==2024.1.1
217
+ dask-cuda==24.2.0
218
+ dask-cudf==24.2.0
219
+ distributed==2024.1.1
220
+ fastrlock==0.8.2
221
+ frozenlist==1.4.1
222
+ importlib_metadata==7.0.2
223
+ locket==1.0.0
224
+ multidict==6.0.5
225
+ networkx==2.6.3
226
+ numba==0.59.0+1.g20ae2b56c
227
+ nvtx==0.2.5
228
+ pandas==1.5.3
229
+ partd==1.4.1
230
+ ply==3.11
231
+ psutil==5.9.4
232
+ pylibcugraph==24.2.0
233
+ pylibcugraphops==24.2.0
234
+ pylibraft==24.2.0
235
+ pynvjitlink==0.1.13
236
+ pynvml==11.4.1
237
+ pytz==2024.1
238
+ raft-dask==24.2.0
239
+ rapids-dask-dependency==24.2.0a0
240
+ rich==13.7.1
241
+ rmm==24.2.0
242
+ scikit-learn==1.2.0
243
+ scipy==1.12.0
244
+ tblib==3.0.0
245
+ thriftpy2==0.4.17
246
+ toolz==0.12.1
247
+ treelite==4.0.0
248
+ ucx-py==0.36.0
249
+ urllib3==1.26.18
250
+ xgboost==2.0.3
251
+ yarl==1.9.4
252
+ zict==3.0.0
253
+ zipp==3.17.0
254
+ tabulate==0.9.0
255
+ igraph==0.11.4
256
+ lightning-thunder==0.0.0.dev0
257
+ lightning-utilities==0.10.1
258
+ looseversion==1.3.0
259
+ opt-einsum==3.3.0
260
+ texttable==1.7.0
261
+ apex==0.1
262
+ cudnn==1.1.2
263
+ ninja==1.11.1.1
264
+ onnx==1.15.0rc2
265
+ pybind11-global==2.11.1
266
+ torchdata==0.7.1a0
267
+ torchtext==0.17.0a0
268
+ contourpy==1.2.0
269
+ cycler==0.12.1
270
+ fonttools==4.49.0
271
+ kiwisolver==1.4.5
272
+ matplotlib==3.8.3
273
+ pycocotools==2.0+nv0.8.0
274
+ pyparsing==3.1.2
275
+ dm-tree==0.1.8
276
+ gast==0.5.4
277
+ nvidia-dali-cuda120==1.35.0
278
+ pytorch-triton==2.2.0+e28a256d7
279
+ lark==1.1.9
280
+ optree==0.10.0
281
+ types-dataclasses==0.6.6
282
+ filelock==3.13.1
283
+ fsspec==2024.2.0
284
+ mpmath==1.3.0
285
+ opencv==4.7.0
286
+ jupyter-tensorboard==0.2.0
287
+ jupytext==1.16.1
288
+ markdown-it-py==3.0.0
289
+ mdit-py-plugins==0.4.0
290
+ mdurl==0.1.2
291
+ toml==0.10.2
292
+ Markdown==3.5.2
293
+ Send2Trash==1.8.2
294
+ absl-py==2.1.0
295
+ argon2-cffi==23.1.0
296
+ argon2-cffi-bindings==21.2.0
297
+ asttokens==2.4.1
298
+ beautifulsoup4==4.12.3
299
+ bleach==6.1.0
300
+ cachetools==5.3.3
301
+ comm==0.2.1
302
+ debugpy==1.8.1
303
+ defusedxml==0.7.1
304
+ executing==2.0.1
305
+ fastjsonschema==2.19.1
306
+ google-auth==2.28.2
307
+ google-auth-oauthlib==0.4.6
308
+ grpcio==1.62.1
309
+ ipykernel==6.29.3
310
+ ipython==8.21.0
311
+ ipython-genutils==0.2.0
312
+ jedi==0.19.1
313
+ json5==0.9.22
314
+ jsonschema==4.21.1
315
+ jsonschema-specifications==2023.12.1
316
+ jupyter_client==8.6.0
317
+ jupyter_core==5.7.1
318
+ jupyterlab_pygments==0.3.0
319
+ matplotlib-inline==0.1.6
320
+ mistune==3.0.2
321
+ nbclient==0.9.0
322
+ nbconvert==7.16.2
323
+ nbformat==5.9.2
324
+ nest-asyncio==1.6.0
325
+ oauthlib==3.2.2
326
+ pandocfilters==1.5.1
327
+ parso==0.8.3
328
+ pexpect==4.9.0
329
+ prometheus_client==0.20.0
330
+ prompt-toolkit==3.0.43
331
+ ptyprocess==0.7.0
332
+ pure-eval==0.2.2
333
+ pyasn1==0.5.1
334
+ pyasn1-modules==0.3.0
335
+ Pygments==2.17.2
336
+ python-dateutil==2.9.0.post0
337
+ python-hostlist==1.23.0
338
+ pyzmq==25.1.2
339
+ referencing==0.33.0
340
+ requests-oauthlib==1.3.1
341
+ rpds-py==0.18.0
342
+ rsa==4.9
343
+ soupsieve==2.5
344
+ stack-data==0.6.3
345
+ tensorboard==2.9.0
346
+ tensorboard-data-server==0.6.1
347
+ tensorboard-plugin-wit==1.8.1
348
+ terminado==0.18.0
349
+ tinycss2==1.2.1
350
+ tornado==6.4
351
+ traitlets==5.9.0
352
+ wcwidth==0.2.13
353
+ webencodings==0.5.1
354
+ Werkzeug==3.0.1
355
+ Cython==3.0.9
356
+ Jinja2==3.1.3
357
+ MarkupSafe==2.1.5
358
+ PyYAML==6.0.1
359
+ annotated-types==0.6.0
360
+ astunparse==1.6.3
361
+ attrs==23.2.0
362
+ audioread==3.0.1
363
+ blis==0.7.11
364
+ catalogue==2.0.10
365
+ certifi==2024.2.2
366
+ charset-normalizer==3.3.2
367
+ click==8.1.7
368
+ cloudpathlib==0.16.0
369
+ confection==0.1.4
370
+ cymem==2.0.8
371
+ decorator==5.1.1
372
+ exceptiongroup==1.2.0
373
+ execnet==2.0.2
374
+ expecttest==0.1.3
375
+ hypothesis==5.35.1
376
+ idna==3.6
377
+ iniconfig==2.0.0
378
+ intel-openmp==2021.4.0
379
+ joblib==1.3.2
380
+ langcodes==3.3.0
381
+ lazy_loader==0.3
382
+ librosa==0.10.1
383
+ llvmlite==0.42.0
384
+ mkl==2021.1.1
385
+ mkl-devel==2021.1.1
386
+ mkl-include==2021.1.1
387
+ mock==5.1.0
388
+ msgpack==1.0.8
389
+ murmurhash==1.0.10
390
+ platformdirs==4.2.0
391
+ pluggy==1.4.0
392
+ pooch==1.8.1
393
+ preshed==3.0.9
394
+ pybind11==2.11.1
395
+ pycparser==2.21
396
+ pydantic_core==2.16.3
397
+ pytest==8.0.2
398
+ pytest-flakefinder==1.1.0
399
+ pytest-rerunfailures==13.0
400
+ pytest-shard==0.1.2
401
+ pytest-xdist==3.5.0
402
+ regex==2023.12.25
403
+ six==1.16.0
404
+ smart-open==6.4.0
405
+ sortedcontainers==2.4.0
406
+ soundfile==0.12.1
407
+ soxr==0.3.7
408
+ spacy==3.7.4
409
+ spacy-legacy==3.0.12
410
+ spacy-loggers==1.0.5
411
+ srsly==2.4.8
412
+ tbb==2021.11.0
413
+ thinc==8.2.3
414
+ threadpoolctl==3.3.0
415
+ tomli==2.0.1
416
+ typer==0.9.0
417
+ wasabi==1.1.2
418
+ weasel==0.3.4
419
+ xdoctest==1.0.2
420
+ wheel==0.42.0
421
+ nvfuser==0.1.5a0+f73ff1bc6a
422
+ dbus-python==1.2.18
423
+ devscripts==2.22.1ubuntu1
424
+ distro-info==1.1+ubuntu0.2
425
+ python-apt==2.4.0+ubuntu4
426
+ unattended-upgrades==0.1
run-20260125_043035-g0u01bsh/files/wandb-metadata.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "os": "Linux-4.18.0-513.5.1.el8_9.x86_64-x86_64-with-glibc2.35",
3
+ "python": "CPython 3.10.12",
4
+ "startedAt": "2026-01-25T04:30:35.116725Z",
5
+ "args": [
6
+ "--policy.path=/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/temp/baseline",
7
+ "--dataset.root=/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/ALOHA/aloha_3_tasks_large_100_original_synthetic_high_view",
8
+ "--output_dir=/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN",
9
+ "--job_name=aloha_100%_merged_synthetic_high_view_PRETRAIN",
10
+ "--config_path=configs/libero_config/default.json",
11
+ "--batch_size=13",
12
+ "--policy.gradient_accumulation_steps=2",
13
+ "--save_freq=10000",
14
+ "--wandb.mode=online",
15
+ "--visual_prompt=false",
16
+ "--zero_init=false",
17
+ "--use_both_origin=false",
18
+ "--coordinate_text_prod=false",
19
+ "--coordinate_posEmbed_prod=false",
20
+ "--posEmbed_to_preTrained=false",
21
+ "--add_posEmbed=false",
22
+ "--not_used_coord=false",
23
+ "--train_expert_only=false",
24
+ "--use_depth=false",
25
+ "--depth_model_type=frozen_siglip",
26
+ "--use_track=true",
27
+ "--use_detect_object=false",
28
+ "--track_mask=false",
29
+ "--future_obj=all",
30
+ "--use_slot_att=none",
31
+ "--original_pi0=false"
32
+ ],
33
+ "program": "/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/lerobot/scripts/train_accelerate.py",
34
+ "codePath": "lerobot/scripts/train_accelerate.py",
35
+ "codePathLocal": "lerobot/scripts/train_accelerate.py",
36
+ "git": {
37
+ "remote": "https://github.com/jibby2803/VLA_custom",
38
+ "commit": "f32c82f37b2b6f59a8b2c1e8367d7d70b203e23b"
39
+ },
40
+ "email": "ducido.w@gmail.com",
41
+ "root": "/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN",
42
+ "host": "085ccf349861",
43
+ "executable": "/usr/bin/python",
44
+ "cpu_count": 32,
45
+ "cpu_count_logical": 32,
46
+ "disk": {
47
+ "/": {
48
+ "total": "10737418240",
49
+ "used": "788828160"
50
+ }
51
+ },
52
+ "memory": {
53
+ "total": "2163816681472"
54
+ },
55
+ "writerId": "xfmd715jagdfpknw5q3bwhtgeese05na"
56
+ }
run-20260125_043035-g0u01bsh/files/wandb-summary.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"train/episodes":8836.419809921328,"train/synthetic_high_imi_loss":1.0248169898986816,"train/losses_after_in_ep_bound":{"_type":"histogram","values":[19131,960,341,126,75,45,27,21,15,12,9,9,7,2,2,5,2,3,1,1,2,1,0,0,0,1,1,0,0,0,0,1],"bins":[0,1.310009241104126,2.620018482208252,3.930027723312378,5.240036964416504,6.550045967102051,7.860055446624756,9.170064926147461,10.480073928833008,11.790082931518555,13.100091934204102,14.410101890563965,15.720110893249512,17.030120849609375,18.340129852294922,19.65013885498047,20.960147857666016,22.270156860351562,23.58016586303711,24.890174865722656,26.200183868408203,27.510194778442383,28.82020378112793,30.130212783813477,31.440221786499023,32.7502326965332,34.06024169921875,35.3702507019043,36.680259704589844,37.99026870727539,39.30027770996094,40.610286712646484,41.92029571533203]},"_runtime":83531.322330155,"train/dataloading_s":0.005119632259011269,"train/update_s":1.3159701265580952,"train/losses_after_forward":{"values":[18841,1168,391,152,78,47,28,21,15,12,9,9,7,2,2,5,2,3,1,1,2,1,0,0,0,1,1,0,0,0,0,1],"bins":[1.3219647598816664e-11,1.310009241104126,2.620018482208252,3.930027723312378,5.240036964416504,6.550046443939209,7.860055446624756,9.170064926147461,10.480073928833008,11.790082931518555,13.100092887878418,14.410101890563965,15.720110893249512,17.030120849609375,18.340129852294922,19.65013885498047,20.960147857666016,22.270156860351562,23.58016586303711,24.890174865722656,26.200185775756836,27.510194778442383,28.82020378112793,30.130212783813477,31.440221786499023,32.7502326965332,34.06024169921875,35.3702507019043,36.680259704589844,37.99026870727539,39.30027770996094,40.610286712646484,41.92029571533203],"_type":"histogram"},"train/loss_track":0.060302734375,"_step":62800,"train/lr":2.5000000000000053e-06,"train/real_right_wrist_imi_loss":1.197863221168518,"train/steps":62800,"train/l1_infer_loss":0.6042347550392151,"train/real_high_imi_loss":1.0556011199951172,"_timestamp":1.769398965451009e+09,"train/l2_loss":0,"train/epochs":8.352003601059858,"_wandb":{"runtime":83531},"train/loss":2.6474615955352783,"train/samples":3265600,"train/real_left_wrist_imi_loss":1.2639321088790894,"train/grad_norm":0.3074590889364481,"train/losses_after_rm_padding":{"values":[6612,1174,542,278,160,120,82,42,33,10,12,7,6,2,2,3,1,3,5,0,4,0,0,1,0,0,0,0,0,0,0,1],"bins":[0,0.5477999448776245,1.095599889755249,1.6433998346328735,2.191199779510498,2.738999843597412,3.286799669265747,3.834599494934082,4.382399559020996,4.93019962310791,5.477999687194824,6.02579927444458,6.573599338531494,7.121399402618408,7.669198989868164,8.216999053955078,8.764799118041992,9.312599182128906,9.86039924621582,10.408199310302734,10.955999374389648,11.503798484802246,12.05159854888916,12.599398612976074,13.147198677062988,13.694998741149902,14.242798805236816,14.79059886932373,15.338397979736328,15.886198043823242,16.433998107910156,16.98179817199707,17.529598236083984],"_type":"histogram"}}
run-20260125_043035-g0u01bsh/logs/debug-core.log ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-01-25T04:30:35.136465183Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp3djoyr6n/port-1165115.txt","pid":1165115,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false}
2
+ {"time":"2026-01-25T04:30:35.136728914Z","level":"WARN","msg":"server/listeners: couldn't open Unix socket in os.TempDir()","error":"server/listeners: failed to make tempdir for Unix socket: stat /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/cache: no such file or directory"}
3
+ {"time":"2026-01-25T04:30:35.137206101Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1165115}
4
+ {"time":"2026-01-25T04:30:35.137152893Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1165115-1165147-3906656257/socket","Net":"unix"}}
5
+ {"time":"2026-01-25T04:30:35.327304085Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"}
6
+ {"time":"2026-01-25T04:30:35.336694973Z","level":"INFO","msg":"handleInformInit: received","streamId":"g0u01bsh","id":"1(@)"}
7
+ {"time":"2026-01-25T04:30:35.516780964Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"g0u01bsh","id":"1(@)"}
8
+ {"time":"2026-01-26T03:42:47.247907085Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"}
9
+ {"time":"2026-01-26T03:42:47.247978558Z","level":"INFO","msg":"server is shutting down"}
10
+ {"time":"2026-01-26T03:42:47.247973368Z","level":"INFO","msg":"connection: closing","id":"1(@)"}
11
+ {"time":"2026-01-26T03:42:47.248092069Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"}
12
+ {"time":"2026-01-26T03:42:47.248111625Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1165115-1165147-3906656257/socket","Net":"unix"}}
13
+ {"time":"2026-01-26T03:43:03.117487938Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"}
14
+ {"time":"2026-01-26T03:43:03.117520539Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"}
15
+ {"time":"2026-01-26T03:43:03.117538643Z","level":"INFO","msg":"server is closed"}
run-20260125_043035-g0u01bsh/logs/debug-internal.log ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-01-25T04:30:35.337188511Z","level":"INFO","msg":"stream: starting","core version":"0.21.4"}
2
+ {"time":"2026-01-25T04:30:35.352330898Z","level":"ERROR","msg":"monitor: failed to initialize GPU resource: monitor: could not create portfile"}
3
+ {"time":"2026-01-25T04:30:35.516712677Z","level":"INFO","msg":"stream: created new stream","id":"g0u01bsh"}
4
+ {"time":"2026-01-25T04:30:35.516771215Z","level":"INFO","msg":"stream: started","id":"g0u01bsh"}
5
+ {"time":"2026-01-25T04:30:35.516797314Z","level":"INFO","msg":"handler: started","stream_id":"g0u01bsh"}
6
+ {"time":"2026-01-25T04:30:35.51678996Z","level":"INFO","msg":"writer: started","stream_id":"g0u01bsh"}
7
+ {"time":"2026-01-25T04:30:35.516816019Z","level":"INFO","msg":"sender: started","stream_id":"g0u01bsh"}
8
+ {"time":"2026-01-25T12:11:53.737869083Z","level":"INFO","msg":"api: retrying HTTP error","status":502,"url":"https://api.wandb.ai/graphql","body":"\n<html><head>\n<meta http-equiv=\"content-type\" content=\"text/html;charset=utf-8\">\n<title>502 Server Error</title>\n</head>\n<body text=#000000 bgcolor=#ffffff>\n<h1>Error: Server Error</h1>\n<h2>The server encountered a temporary error and could not complete your request.<p>Please try again in 30 seconds.</h2>\n<h2></h2>\n</body></html>\n"}
9
+ {"time":"2026-01-26T03:42:47.247976354Z","level":"INFO","msg":"stream: closing","id":"g0u01bsh"}
10
+ {"time":"2026-01-26T03:43:02.531387332Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
11
+ {"time":"2026-01-26T03:43:03.114011174Z","level":"INFO","msg":"handler: closed","stream_id":"g0u01bsh"}
12
+ {"time":"2026-01-26T03:43:03.116713969Z","level":"INFO","msg":"sender: closed","stream_id":"g0u01bsh"}
13
+ {"time":"2026-01-26T03:43:03.116749606Z","level":"INFO","msg":"stream: closed","id":"g0u01bsh"}
run-20260125_043035-g0u01bsh/logs/debug.log ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Current SDK version is 0.21.4
2
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Configure stats pid to 1165115
3
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Loading settings from /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/.config/wandb/settings
4
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Loading settings from /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/wandb/settings
5
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_setup.py:_flush():81] Loading settings from environment variables
6
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:setup_run_log_directory():686] Logging user logs to /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN/wandb/run-20260125_043035-g0u01bsh/logs/debug.log
7
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:setup_run_log_directory():687] Logging internal logs to /pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN/wandb/run-20260125_043035-g0u01bsh/logs/debug-internal.log
8
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:init():813] calling init triggers
9
+ 2026-01-25 04:30:35,121 INFO MainThread:1165115 [wandb_init.py:init():818] wandb.init called with sweep_config: {}
10
+ config: {'dataset': {'repo_id': 'ducido/calvin_task_D_D_scale_100_lerobo_format', 'root': '/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/ALOHA/aloha_3_tasks_large_100_original_synthetic_high_view', 'episodes': None, 'image_transforms': {'enable': True, 'max_num_transforms': 3, 'random_order': False, 'image_tfs': {'hue': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'hue': [-0.05, 0.05]}}, 'contrast': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'contrast': [0.8, 1.2]}}, 'sharpness': {'weight': 1.0, 'type': 'SharpnessJitter', 'kwargs': {'sharpness': [0.5, 1.5]}}, 'brightness': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'brightness': [0.8, 1.2]}}, 'saturation': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'saturation': [0.5, 1.5]}}, 'crop_resize': {'weight': 1.0, 'type': 'RandomResizedCrop', 'kwargs': {'size': [256, 256], 'ratio': [1, 1], 'scale': [0.9, 0.95]}}, 'rotate': {'weight': 1.0, 'type': 'RandomRotate', 'kwargs': {'degrees': [-5, 5]}}}, 'wrist_tfs': {'hue': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'hue': [-0.05, 0.05]}}, 'contrast': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'contrast': [0.8, 1.2]}}, 'crop_resize': {'weight': 1.0, 'type': 'RandomResizedCrop', 'kwargs': {'size': [256, 256], 'ratio': [1, 1], 'scale': [0.9, 0.95]}}, 'sharpness': {'weight': 1.0, 'type': 'SharpnessJitter', 'kwargs': {'sharpness': [0.5, 1.5]}}, 'brightness': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'brightness': [0.8, 1.2]}}, 'saturation': {'weight': 1.0, 'type': 'ColorJitter', 'kwargs': {'saturation': [0.5, 1.5]}}}}, 'revision': None, 'use_imagenet_stats': True, 'video_backend': 'torchcodec', 'vqa_data_path': None}, 'env': None, 'policy': {'type': 'pi0', 'n_obs_steps': 1, 'normalization_mapping': {'VISUAL': <NormalizationMode.IDENTITY: 'IDENTITY'>, 'STATE': <NormalizationMode.MEAN_STD: 'MEAN_STD'>, 'ACTION': <NormalizationMode.MEAN_STD: 'MEAN_STD'>}, 'input_features': {}, 'output_features': {}, 'device': 'cuda', 'use_amp': False, 'gradient_accumulation_steps': 2, 'chunk_size': 50, 'n_action_steps': 50, 'visual_prompt': False, 'use_both_origin': False, 'zero_init': False, 'coordinate_text_prod': False, 'coordinate_posEmbed_prod': False, 'posEmbed_to_preTrained': False, 'num_obs_layers': 30, 'last_obs_layers': False, 'add_posEmbed': False, 'not_used_coord': False, 'use_depth': False, 'depth_model_type': 'frozen_siglip', 'use_track': False, 'use_detect_object': False, 'track_mask': False, 'use_slot_att': 'none', 'future_obj': 'none', 'original_pi0': False, 'max_state_dim': 32, 'max_action_dim': 32, 'resize_imgs_with_padding': [224, 224], 'empty_cameras': 0, 'adapt_to_pi_aloha': False, 'use_delta_joint_actions_aloha': False, 'tokenizer_max_length': 48, 'proj_width': 1024, 'num_steps': 10, 'use_cache': True, 'attention_implementation': 'eager', 'freeze_vision_encoder': True, 'train_expert_only': False, 'train_state_proj': True, 'optimizer_lr': 0.0001, 'optimizer_betas': [0.9, 0.95], 'optimizer_eps': 1e-08, 'optimizer_weight_decay': 1e-10, 'scheduler_warmup_steps': 1000, 'scheduler_decay_steps': 240000, 'scheduler_decay_lr': 2.5e-06}, 'output_dir': '/pfss/mlde/workspaces/mlde_wsp_PI_Hauschild/VLA/duc/VLA_custom/outputs/train/2026-01-25/04-30-23_aloha_100%_merged_synthetic_high_view_PRETRAIN', 'job_name': 'aloha_100%_merged_synthetic_high_view_PRETRAIN', 'visual_prompt': False, 'use_both_origin': False, 'zero_init': False, 'coordinate_text_prod': False, 'coordinate_posEmbed_prod': False, 'posEmbed_to_preTrained': False, 'num_obs_layers': 30, 'last_obs_layers': False, 'add_posEmbed': False, 'not_used_coord': False, 'train_expert_only': False, 'use_depth': False, 'use_track': True, 'use_detect_object': False, 'track_mask': False, 'use_slot_att': 'none', 'future_obj': 'all', 'depth_model_type': 'frozen_siglip', 'original_pi0': False, 'resume': False, 'seed': 42, 'num_workers': 8, 'batch_size': 13, 'steps': 100000, 'eval_freq': 20000, 'log_freq': 200, 'save_checkpoint': True, 'save_freq': 10000, 'use_policy_training_preset': True, 'optimizer': {'type': 'adamw', 'lr': 0.0001, 'weight_decay': 1e-10, 'grad_clip_norm': 10.0, 'betas': [0.9, 0.95], 'eps': 1e-08}, 'scheduler': {'type': 'cosine_decay_with_warmup', 'num_warmup_steps': 1000, 'num_decay_steps': 240000, 'peak_lr': 0.0001, 'decay_lr': 2.5e-06}, 'eval': {'n_episodes': 50, 'batch_size': 50, 'use_async_envs': False}, 'wandb': {'enable': True, 'disable_artifact': True, 'project': 'pi0_lerobot', 'entity': 'Robotics_VLA', 'notes': None, 'run_id': None, 'mode': 'online'}, '_wandb': {}}
11
+ 2026-01-25 04:30:35,122 INFO MainThread:1165115 [wandb_init.py:init():854] starting backend
12
+ 2026-01-25 04:30:35,327 INFO MainThread:1165115 [wandb_init.py:init():857] sending inform_init request
13
+ 2026-01-25 04:30:35,332 INFO MainThread:1165115 [wandb_init.py:init():865] backend started and connected
14
+ 2026-01-25 04:30:35,334 INFO MainThread:1165115 [wandb_init.py:init():936] updated telemetry
15
+ 2026-01-25 04:30:35,344 INFO MainThread:1165115 [wandb_init.py:init():960] communicating run to backend with 90.0 second timeout
16
+ 2026-01-25 04:30:35,923 INFO MainThread:1165115 [wandb_init.py:init():1011] starting run threads in backend
17
+ 2026-01-25 04:30:36,126 INFO MainThread:1165115 [wandb_run.py:_console_start():2506] atexit reg
18
+ 2026-01-25 04:30:36,127 INFO MainThread:1165115 [wandb_run.py:_redirect():2354] redirect: wrap_raw
19
+ 2026-01-25 04:30:36,127 INFO MainThread:1165115 [wandb_run.py:_redirect():2423] Wrapping output streams.
20
+ 2026-01-25 04:30:36,127 INFO MainThread:1165115 [wandb_run.py:_redirect():2446] Redirects installed.
21
+ 2026-01-25 04:30:36,129 INFO MainThread:1165115 [wandb_init.py:init():1049] run started, returning control to user process
22
+ 2026-01-26 03:42:47,248 INFO wandb-AsyncioManager-main:1165115 [service_client.py:_forward_responses():84] Reached EOF.
23
+ 2026-01-26 03:42:47,248 INFO wandb-AsyncioManager-main:1165115 [mailbox.py:close():137] Closing mailbox, abandoning 1 handles.
run-20260125_043035-g0u01bsh/run-g0u01bsh.wandb ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8369ba6d67514cf817f8bcfe48ca51cb3260712d17734d2821d2cab1743a0af9
3
+ size 211436453