Hugging Face's logo

Whalswp
/
experiment_v1

Safetensors
Model card Files
xet
 Community
Whalswp commited on
Commit
cd041bb
·
verified ·
1 Parent(s): 915056f

Add files using upload-large-folder tool

Browse files
Files changed (23) hide show
  1. .gitattributes +2 -0
  2. MGRKD/launch_logs/mgrkd_raw_action_flatten_frozen_mlp_gpu5_bs128_20260622_165623.log +0 -0
  3. MGRKD/logs/mgrkd_da_flatten_action_encoder_gpu4_bs128_20260622_134555.log +0 -0
  4. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/config.json +84 -0
  5. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/model-00001-of-00002.safetensors +3 -0
  6. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/model-00002-of-00002.safetensors +3 -0
  7. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/model.safetensors.index.json +0 -0
  8. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/task_fm_loss_1024_seed42.json +244 -0
  9. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/task_fm_loss_smoke16.json +244 -0
  10. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/trainer_state.json +0 -0
  11. MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/training_args.bin +3 -0
  12. MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/mgrkd_da_flatten_action_encoder_phase3_plus30k.yaml +36 -0
  13. MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/phase3/experiment_cfg/metadata.json +431 -0
  14. MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/phase3_bad_aux_reset_20260624_120659/experiment_cfg/metadata.json +431 -0
  15. MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/resolved_config.yaml +37 -0
  16. MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/config.json +90 -0
  17. MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/model-00001-of-00002.safetensors +3 -0
  18. MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/model-00002-of-00002.safetensors +3 -0
  19. MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/model.safetensors.index.json +0 -0
  20. MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/trainer_state.json +0 -0
  21. MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/training_args.bin +3 -0
  22. MGRKD/task_fm_loss_baseline120k_rkdv21a_mgrkdae_1024_seed42.tsv +28 -0
  23. rkd_v2_1/rkd_v2_1_angle/default/phase3/task_fm_loss_1024_seed42.json +244 -0
.gitattributes CHANGED
@@ -39,3 +39,5 @@ VLM_Only/MGD_Raw_VLM/train.log filter=lfs diff=lfs merge=lfs -text
39
  experiment_cfg/processing_line_only_v2/MGD_v2/mgd_v2_mask_ratio_gpu01_lane.log filter=lfs diff=lfs merge=lfs -text
40
  rkd_v2_1/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
  rkd_v2_1/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
39
  experiment_cfg/processing_line_only_v2/MGD_v2/mgd_v2_mask_ratio_gpu01_lane.log filter=lfs diff=lfs merge=lfs -text
40
  rkd_v2_1/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
  rkd_v2_1/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
+ MGRKD/logs/mgrkd_da_flatten_action_encoder_gpu4_bs128_20260622_134555.log filter=lfs diff=lfs merge=lfs -text
43
+ MGRKD/launch_logs/mgrkd_raw_action_flatten_frozen_mlp_gpu5_bs128_20260622_165623.log filter=lfs diff=lfs merge=lfs -text
MGRKD/launch_logs/mgrkd_raw_action_flatten_frozen_mlp_gpu5_bs128_20260622_165623.log CHANGED
The diff for this file is too large to render. See raw diff
 
MGRKD/logs/mgrkd_da_flatten_action_encoder_gpu4_bs128_20260622_134555.log CHANGED
The diff for this file is too large to render. See raw diff
 
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/config.json ADDED
@@ -0,0 +1,84 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_projector_dim": 512,
63
+ "rkd_action_encoder_projector_enabled": true,
64
+ "rkd_action_encoder_projector_pooling": "flatten",
65
+ "rkd_action_temp": 0.1,
66
+ "rkd_angle_loss_weight": 2.0,
67
+ "rkd_distance_loss_weight": 1.0,
68
+ "rkd_enabled": true,
69
+ "rkd_exclude_diagonal": true,
70
+ "rkd_fm_loss_weight": 1.0,
71
+ "rkd_loss_type": "distance_angle",
72
+ "rkd_loss_weight": 0.5,
73
+ "rkd_loss_weight_end": 0.0,
74
+ "rkd_loss_weight_schedule": null,
75
+ "rkd_loss_weight_start": 0.1,
76
+ "rkd_relation_mode": "flatten",
77
+ "rkd_student_reconstruction_head_dim": 512,
78
+ "rkd_student_reconstruction_head_enabled": true,
79
+ "rkd_student_reconstruction_head_hidden_dim": 512,
80
+ "rkd_teacher_source": "action_encoder",
81
+ "rkd_vlm_temp": 0.07,
82
+ "torch_dtype": "bfloat16",
83
+ "transformers_version": "4.51.3"
84
+ }
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:026ee17fbf4ac0165da522d4085cee881b4d1488480b270da583c2c2f3c93987
3
+ size 4999367032
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a6b71606494587a3b50fbd25f71d3ffe5e77cc629f4cdad2c83aeac77f4da19
3
+ size 2643339852
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/task_fm_loss_1024_seed42.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_path": "/home/ext_minje/groot_robocasa/robocasa_v2/MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3",
3
+ "config": "/home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/processing_line_only_v2/MGRKD/mgrkd_da_flatten_action_encoder_phase3_plus30k.yaml",
4
+ "batch_size": 32,
5
+ "max_samples_per_task": 1024,
6
+ "seed": 42,
7
+ "mean_task_fm_loss": 0.014533606417199526,
8
+ "results": [
9
+ {
10
+ "task": "CloseBlenderLid",
11
+ "fm_loss": 0.014584552016458474,
12
+ "samples": 1024,
13
+ "batches": 32,
14
+ "dataset_len": 34953,
15
+ "sampled": true,
16
+ "planned_samples": 1024
17
+ },
18
+ {
19
+ "task": "CloseFridge",
20
+ "fm_loss": 0.015641167206922546,
21
+ "samples": 1024,
22
+ "batches": 32,
23
+ "dataset_len": 25750,
24
+ "sampled": true,
25
+ "planned_samples": 1024
26
+ },
27
+ {
28
+ "task": "CloseToasterOvenDoor",
29
+ "fm_loss": 0.012825453974073753,
30
+ "samples": 1024,
31
+ "batches": 32,
32
+ "dataset_len": 17858,
33
+ "sampled": true,
34
+ "planned_samples": 1024
35
+ },
36
+ {
37
+ "task": "CoffeeSetupMug",
38
+ "fm_loss": 0.012992071235203184,
39
+ "samples": 1024,
40
+ "batches": 32,
41
+ "dataset_len": 22529,
42
+ "sampled": true,
43
+ "planned_samples": 1024
44
+ },
45
+ {
46
+ "task": "NavigateKitchen",
47
+ "fm_loss": 0.01587945010396652,
48
+ "samples": 1024,
49
+ "batches": 32,
50
+ "dataset_len": 16346,
51
+ "sampled": true,
52
+ "planned_samples": 1024
53
+ },
54
+ {
55
+ "task": "OpenCabinet",
56
+ "fm_loss": 0.014442342275287956,
57
+ "samples": 1024,
58
+ "batches": 32,
59
+ "dataset_len": 35393,
60
+ "sampled": true,
61
+ "planned_samples": 1024
62
+ },
63
+ {
64
+ "task": "OpenDrawer",
65
+ "fm_loss": 0.010897883112193085,
66
+ "samples": 1024,
67
+ "batches": 32,
68
+ "dataset_len": 20027,
69
+ "sampled": true,
70
+ "planned_samples": 1024
71
+ },
72
+ {
73
+ "task": "OpenStandMixerHead",
74
+ "fm_loss": 0.015409716623253189,
75
+ "samples": 1024,
76
+ "batches": 32,
77
+ "dataset_len": 12400,
78
+ "sampled": true,
79
+ "planned_samples": 1024
80
+ },
81
+ {
82
+ "task": "PickPlaceCounterToCabinet",
83
+ "fm_loss": 0.01651429443154484,
84
+ "samples": 1024,
85
+ "batches": 32,
86
+ "dataset_len": 22582,
87
+ "sampled": true,
88
+ "planned_samples": 1024
89
+ },
90
+ {
91
+ "task": "PickPlaceCounterToStove",
92
+ "fm_loss": 0.013721773124416359,
93
+ "samples": 1024,
94
+ "batches": 32,
95
+ "dataset_len": 22279,
96
+ "sampled": true,
97
+ "planned_samples": 1024
98
+ },
99
+ {
100
+ "task": "PickPlaceDrawerToCounter",
101
+ "fm_loss": 0.015555709018372,
102
+ "samples": 1024,
103
+ "batches": 32,
104
+ "dataset_len": 30785,
105
+ "sampled": true,
106
+ "planned_samples": 1024
107
+ },
108
+ {
109
+ "task": "PickPlaceSinkToCounter",
110
+ "fm_loss": 0.013916867799707688,
111
+ "samples": 1024,
112
+ "batches": 32,
113
+ "dataset_len": 24529,
114
+ "sampled": true,
115
+ "planned_samples": 1024
116
+ },
117
+ {
118
+ "task": "PickPlaceToasterToCounter",
119
+ "fm_loss": 0.014703709646710195,
120
+ "samples": 1024,
121
+ "batches": 32,
122
+ "dataset_len": 25630,
123
+ "sampled": true,
124
+ "planned_samples": 1024
125
+ },
126
+ {
127
+ "task": "PickPlaceCabinetToCounter",
128
+ "fm_loss": 0.015287194430129603,
129
+ "samples": 1024,
130
+ "batches": 32,
131
+ "dataset_len": 19090,
132
+ "sampled": true,
133
+ "planned_samples": 1024
134
+ },
135
+ {
136
+ "task": "TurnOnElectricKettle",
137
+ "fm_loss": 0.019986314291600138,
138
+ "samples": 1024,
139
+ "batches": 32,
140
+ "dataset_len": 11625,
141
+ "sampled": true,
142
+ "planned_samples": 1024
143
+ },
144
+ {
145
+ "task": "TurnOnMicrowave",
146
+ "fm_loss": 0.020842015743255615,
147
+ "samples": 1024,
148
+ "batches": 32,
149
+ "dataset_len": 13038,
150
+ "sampled": true,
151
+ "planned_samples": 1024
152
+ },
153
+ {
154
+ "task": "TurnOffMicrowave",
155
+ "fm_loss": 0.01835524899070151,
156
+ "samples": 1024,
157
+ "batches": 32,
158
+ "dataset_len": 14175,
159
+ "sampled": true,
160
+ "planned_samples": 1024
161
+ },
162
+ {
163
+ "task": "OpenElectricKettleLid",
164
+ "fm_loss": 0.014266238911659457,
165
+ "samples": 1024,
166
+ "batches": 32,
167
+ "dataset_len": 9928,
168
+ "sampled": true,
169
+ "planned_samples": 1024
170
+ },
171
+ {
172
+ "task": "SlideDishwasherRack",
173
+ "fm_loss": 0.0117580722871935,
174
+ "samples": 1024,
175
+ "batches": 32,
176
+ "dataset_len": 19052,
177
+ "sampled": true,
178
+ "planned_samples": 1024
179
+ },
180
+ {
181
+ "task": "TurnOnSinkFaucet",
182
+ "fm_loss": 0.008850100843119435,
183
+ "samples": 1024,
184
+ "batches": 32,
185
+ "dataset_len": 22234,
186
+ "sampled": true,
187
+ "planned_samples": 1024
188
+ },
189
+ {
190
+ "task": "TurnSinkSpout",
191
+ "fm_loss": 0.01594742490851786,
192
+ "samples": 1024,
193
+ "batches": 32,
194
+ "dataset_len": 10219,
195
+ "sampled": true,
196
+ "planned_samples": 1024
197
+ },
198
+ {
199
+ "task": "TurnOffSinkFaucet",
200
+ "fm_loss": 0.016457595964311622,
201
+ "samples": 1024,
202
+ "batches": 32,
203
+ "dataset_len": 11626,
204
+ "sampled": true,
205
+ "planned_samples": 1024
206
+ },
207
+ {
208
+ "task": "AdjustWaterTemperature",
209
+ "fm_loss": 0.010618240354233421,
210
+ "samples": 1024,
211
+ "batches": 32,
212
+ "dataset_len": 19725,
213
+ "sampled": true,
214
+ "planned_samples": 1024
215
+ },
216
+ {
217
+ "task": "TurnOffStove",
218
+ "fm_loss": 0.011689437960740179,
219
+ "samples": 1024,
220
+ "batches": 32,
221
+ "dataset_len": 29821,
222
+ "sampled": true,
223
+ "planned_samples": 1024
224
+ },
225
+ {
226
+ "task": "TurnOnStove",
227
+ "fm_loss": 0.013856435543857515,
228
+ "samples": 1024,
229
+ "batches": 32,
230
+ "dataset_len": 17322,
231
+ "sampled": true,
232
+ "planned_samples": 1024
233
+ },
234
+ {
235
+ "task": "AdjustToasterOvenTemperature",
236
+ "fm_loss": 0.01287445604975801,
237
+ "samples": 1024,
238
+ "batches": 32,
239
+ "dataset_len": 19846,
240
+ "sampled": true,
241
+ "planned_samples": 1024
242
+ }
243
+ ]
244
+ }
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/task_fm_loss_smoke16.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_path": "/home/ext_minje/groot_robocasa/robocasa_v2/MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3",
3
+ "config": "/home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/processing_line_only_v2/MGRKD/mgrkd_da_flatten_action_encoder_phase3_plus30k.yaml",
4
+ "batch_size": 8,
5
+ "max_samples_per_task": 16,
6
+ "seed": 42,
7
+ "mean_task_fm_loss": 0.013483350194292143,
8
+ "results": [
9
+ {
10
+ "task": "CloseBlenderLid",
11
+ "fm_loss": 0.01670990791171789,
12
+ "samples": 16,
13
+ "batches": 2,
14
+ "dataset_len": 34953,
15
+ "sampled": true,
16
+ "planned_samples": 16
17
+ },
18
+ {
19
+ "task": "CloseFridge",
20
+ "fm_loss": 0.012572316452860832,
21
+ "samples": 16,
22
+ "batches": 2,
23
+ "dataset_len": 25750,
24
+ "sampled": true,
25
+ "planned_samples": 16
26
+ },
27
+ {
28
+ "task": "CloseToasterOvenDoor",
29
+ "fm_loss": 0.005728181917220354,
30
+ "samples": 16,
31
+ "batches": 2,
32
+ "dataset_len": 17858,
33
+ "sampled": true,
34
+ "planned_samples": 16
35
+ },
36
+ {
37
+ "task": "CoffeeSetupMug",
38
+ "fm_loss": 0.004615404701326042,
39
+ "samples": 16,
40
+ "batches": 2,
41
+ "dataset_len": 22529,
42
+ "sampled": true,
43
+ "planned_samples": 16
44
+ },
45
+ {
46
+ "task": "NavigateKitchen",
47
+ "fm_loss": 0.03536599408835173,
48
+ "samples": 16,
49
+ "batches": 2,
50
+ "dataset_len": 16346,
51
+ "sampled": true,
52
+ "planned_samples": 16
53
+ },
54
+ {
55
+ "task": "OpenCabinet",
56
+ "fm_loss": 0.0077230054885149,
57
+ "samples": 16,
58
+ "batches": 2,
59
+ "dataset_len": 35393,
60
+ "sampled": true,
61
+ "planned_samples": 16
62
+ },
63
+ {
64
+ "task": "OpenDrawer",
65
+ "fm_loss": 0.009860041551291943,
66
+ "samples": 16,
67
+ "batches": 2,
68
+ "dataset_len": 20027,
69
+ "sampled": true,
70
+ "planned_samples": 16
71
+ },
72
+ {
73
+ "task": "OpenStandMixerHead",
74
+ "fm_loss": 0.015443287091329694,
75
+ "samples": 16,
76
+ "batches": 2,
77
+ "dataset_len": 12400,
78
+ "sampled": true,
79
+ "planned_samples": 16
80
+ },
81
+ {
82
+ "task": "PickPlaceCounterToCabinet",
83
+ "fm_loss": 0.030835176818072796,
84
+ "samples": 16,
85
+ "batches": 2,
86
+ "dataset_len": 22582,
87
+ "sampled": true,
88
+ "planned_samples": 16
89
+ },
90
+ {
91
+ "task": "PickPlaceCounterToStove",
92
+ "fm_loss": 0.009551360504701734,
93
+ "samples": 16,
94
+ "batches": 2,
95
+ "dataset_len": 22279,
96
+ "sampled": true,
97
+ "planned_samples": 16
98
+ },
99
+ {
100
+ "task": "PickPlaceDrawerToCounter",
101
+ "fm_loss": 0.008698298479430377,
102
+ "samples": 16,
103
+ "batches": 2,
104
+ "dataset_len": 30785,
105
+ "sampled": true,
106
+ "planned_samples": 16
107
+ },
108
+ {
109
+ "task": "PickPlaceSinkToCounter",
110
+ "fm_loss": 0.006355189252644777,
111
+ "samples": 16,
112
+ "batches": 2,
113
+ "dataset_len": 24529,
114
+ "sampled": true,
115
+ "planned_samples": 16
116
+ },
117
+ {
118
+ "task": "PickPlaceToasterToCounter",
119
+ "fm_loss": 0.005177703453227878,
120
+ "samples": 16,
121
+ "batches": 2,
122
+ "dataset_len": 25630,
123
+ "sampled": true,
124
+ "planned_samples": 16
125
+ },
126
+ {
127
+ "task": "PickPlaceCabinetToCounter",
128
+ "fm_loss": 0.011685180012136698,
129
+ "samples": 16,
130
+ "batches": 2,
131
+ "dataset_len": 19090,
132
+ "sampled": true,
133
+ "planned_samples": 16
134
+ },
135
+ {
136
+ "task": "TurnOnElectricKettle",
137
+ "fm_loss": 0.01364654814824462,
138
+ "samples": 16,
139
+ "batches": 2,
140
+ "dataset_len": 11625,
141
+ "sampled": true,
142
+ "planned_samples": 16
143
+ },
144
+ {
145
+ "task": "TurnOnMicrowave",
146
+ "fm_loss": 0.009197523118928075,
147
+ "samples": 16,
148
+ "batches": 2,
149
+ "dataset_len": 13038,
150
+ "sampled": true,
151
+ "planned_samples": 16
152
+ },
153
+ {
154
+ "task": "TurnOffMicrowave",
155
+ "fm_loss": 0.013085138285532594,
156
+ "samples": 16,
157
+ "batches": 2,
158
+ "dataset_len": 14175,
159
+ "sampled": true,
160
+ "planned_samples": 16
161
+ },
162
+ {
163
+ "task": "OpenElectricKettleLid",
164
+ "fm_loss": 0.02448701485991478,
165
+ "samples": 16,
166
+ "batches": 2,
167
+ "dataset_len": 9928,
168
+ "sampled": true,
169
+ "planned_samples": 16
170
+ },
171
+ {
172
+ "task": "SlideDishwasherRack",
173
+ "fm_loss": 0.01759397378191352,
174
+ "samples": 16,
175
+ "batches": 2,
176
+ "dataset_len": 19052,
177
+ "sampled": true,
178
+ "planned_samples": 16
179
+ },
180
+ {
181
+ "task": "TurnOnSinkFaucet",
182
+ "fm_loss": 0.012656820472329855,
183
+ "samples": 16,
184
+ "batches": 2,
185
+ "dataset_len": 22234,
186
+ "sampled": true,
187
+ "planned_samples": 16
188
+ },
189
+ {
190
+ "task": "TurnSinkSpout",
191
+ "fm_loss": 0.010076357051730156,
192
+ "samples": 16,
193
+ "batches": 2,
194
+ "dataset_len": 10219,
195
+ "sampled": true,
196
+ "planned_samples": 16
197
+ },
198
+ {
199
+ "task": "TurnOffSinkFaucet",
200
+ "fm_loss": 0.011767837218940258,
201
+ "samples": 16,
202
+ "batches": 2,
203
+ "dataset_len": 11626,
204
+ "sampled": true,
205
+ "planned_samples": 16
206
+ },
207
+ {
208
+ "task": "AdjustWaterTemperature",
209
+ "fm_loss": 0.007991486811079085,
210
+ "samples": 16,
211
+ "batches": 2,
212
+ "dataset_len": 19725,
213
+ "sampled": true,
214
+ "planned_samples": 16
215
+ },
216
+ {
217
+ "task": "TurnOffStove",
218
+ "fm_loss": 0.018952951300889254,
219
+ "samples": 16,
220
+ "batches": 2,
221
+ "dataset_len": 29821,
222
+ "sampled": true,
223
+ "planned_samples": 16
224
+ },
225
+ {
226
+ "task": "TurnOnStove",
227
+ "fm_loss": 0.008545647375285625,
228
+ "samples": 16,
229
+ "batches": 2,
230
+ "dataset_len": 17322,
231
+ "sampled": true,
232
+ "planned_samples": 16
233
+ },
234
+ {
235
+ "task": "AdjustToasterOvenTemperature",
236
+ "fm_loss": 0.022244758903980255,
237
+ "samples": 16,
238
+ "batches": 2,
239
+ "dataset_len": 19846,
240
+ "sampled": true,
241
+ "planned_samples": 16
242
+ }
243
+ ]
244
+ }
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef72ba30b63e015c8ef5b9d5c3dfdfefc8d06160a9599f02ac6a102a7484241e
3
+ size 5969
MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/mgrkd_da_flatten_action_encoder_phase3_plus30k.yaml ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k_aux_fixed_0p5
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model: null
12
+ phases:
13
+ - name: phase3_fm_mgrkd_da_fixed_0p5
14
+ max_steps: 30000
15
+ save_steps: 0
16
+ trainable:
17
+ tune_llm: false
18
+ tune_visual: false
19
+ tune_projector: true
20
+ tune_diffusion_model: true
21
+ losses:
22
+ rkd_enabled: true
23
+ rkd_fm_loss_weight: 1.0
24
+ rkd_loss_weight: 0.5
25
+ rkd_relation_mode: flatten
26
+ rkd_loss_type: distance_angle
27
+ rkd_teacher_source: action_encoder
28
+ rkd_action_encoder_projector_enabled: true
29
+ rkd_action_encoder_projector_dim: 512
30
+ rkd_action_encoder_projector_pooling: flatten
31
+ rkd_student_reconstruction_head_enabled: true
32
+ rkd_student_reconstruction_head_dim: 512
33
+ rkd_student_reconstruction_head_hidden_dim: 512
34
+ rkd_distance_loss_weight: 1.0
35
+ rkd_angle_loss_weight: 2.0
36
+ rkd_exclude_diagonal: true
MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/phase3/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/phase3_bad_aux_reset_20260624_120659/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k/default/resolved_config.yaml ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k_aux_fixed_0p5
2
+ policy_type: groot_MGRKD
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/MGRKD/mgrkd_distance_angle_flatten_action_encoder/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/MGRKD/mgrkd_distance_angle_flatten_action_encoder_phase3_plus30k
5
+ dataset:
6
+ dataset_soup: my_atomic26_human
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 128
10
+ seed: 42
11
+ model: null
12
+ phases:
13
+ - name: phase3_fm_mgrkd_da_fixed_0p5
14
+ max_steps: 30000
15
+ save_steps: 0
16
+ trainable:
17
+ tune_llm: false
18
+ tune_visual: false
19
+ tune_projector: true
20
+ tune_diffusion_model: true
21
+ losses:
22
+ rkd_enabled: true
23
+ rkd_fm_loss_weight: 1.0
24
+ rkd_loss_weight: 0.5
25
+ rkd_relation_mode: flatten
26
+ rkd_loss_type: distance_angle
27
+ rkd_teacher_source: action_encoder
28
+ rkd_action_encoder_projector_enabled: true
29
+ rkd_action_encoder_projector_dim: 512
30
+ rkd_action_encoder_projector_pooling: flatten
31
+ rkd_student_reconstruction_head_enabled: true
32
+ rkd_student_reconstruction_head_dim: 512
33
+ rkd_student_reconstruction_head_hidden_dim: 512
34
+ rkd_distance_loss_weight: 1.0
35
+ rkd_angle_loss_weight: 2.0
36
+ rkd_exclude_diagonal: true
37
+ resolved_sweep: {}
MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/config.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_enabled": true,
72
+ "rkd_exclude_diagonal": true,
73
+ "rkd_fm_loss_weight": 1.0,
74
+ "rkd_loss_type": "distance_angle",
75
+ "rkd_loss_weight": 0.5,
76
+ "rkd_loss_weight_end": 0.0,
77
+ "rkd_loss_weight_schedule": null,
78
+ "rkd_loss_weight_start": 0.1,
79
+ "rkd_raw_action_projector_dim": 512,
80
+ "rkd_raw_action_projector_enabled": true,
81
+ "rkd_raw_action_projector_pooling": "flatten",
82
+ "rkd_relation_mode": "flatten",
83
+ "rkd_student_reconstruction_head_dim": 512,
84
+ "rkd_student_reconstruction_head_enabled": true,
85
+ "rkd_student_reconstruction_head_hidden_dim": 512,
86
+ "rkd_teacher_source": "raw_action",
87
+ "rkd_vlm_temp": 0.07,
88
+ "torch_dtype": "bfloat16",
89
+ "transformers_version": "4.51.3"
90
+ }
MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55b50388fa8d66dc64aa02b2fe6e717625f3a1d3a853d1bd74f2088636dbddfa
3
+ size 4999367032
MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:243ea990f27ca6533a150a1b9015b4efe8d08456acd3e8042333d526ae1a0808
3
+ size 2660191348
MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
MGRKD/mgrkd_distance_angle_flatten_raw_action/default/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40a35ab41df0ab7a54205eca37d1b046b0d122a76c2a00bdd5bddbdbf483a61f
3
+ size 5969
MGRKD/task_fm_loss_baseline120k_rkdv21a_mgrkdae_1024_seed42.tsv ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Task Baseline120K RKD v2.1 A MGRKD AE
2
+ CloseBlenderLid 0.013299564 0.012893491 0.014584552
3
+ CloseFridge 0.017638529 0.018061224 0.015641167
4
+ CloseToasterOvenDoor 0.011711232 0.012771354 0.012825454
5
+ CoffeeSetupMug 0.013539590 0.013651895 0.012992071
6
+ NavigateKitchen 0.014088925 0.013925644 0.015879450
7
+ OpenCabinet 0.015281203 0.014759833 0.014442342
8
+ OpenDrawer 0.012880304 0.013188228 0.010897883
9
+ OpenStandMixerHead 0.014427024 0.015075979 0.015409717
10
+ PickPlaceCounterToCabinet 0.016352756 0.016090463 0.016514294
11
+ PickPlaceCounterToStove 0.014688284 0.014552780 0.013721773
12
+ PickPlaceDrawerToCounter 0.016490812 0.016062375 0.015555709
13
+ PickPlaceSinkToCounter 0.015005368 0.015319367 0.013916868
14
+ PickPlaceToasterToCounter 0.013634393 0.013780935 0.014703710
15
+ PickPlaceCabinetToCounter 0.018807061 0.019037108 0.015287194
16
+ TurnOnElectricKettle 0.018448099 0.018178111 0.019986314
17
+ TurnOnMicrowave 0.019436894 0.019411736 0.020842016
18
+ TurnOffMicrowave 0.020420188 0.020360561 0.018355249
19
+ OpenElectricKettleLid 0.013036789 0.013258201 0.014266239
20
+ SlideDishwasherRack 0.010529016 0.011744259 0.011758072
21
+ TurnOnSinkFaucet 0.009266675 0.009852688 0.008850101
22
+ TurnSinkSpout 0.017223676 0.017699005 0.015947425
23
+ TurnOffSinkFaucet 0.015480601 0.015963105 0.016457596
24
+ AdjustWaterTemperature 0.009746092 0.010782943 0.010618240
25
+ TurnOffStove 0.011625355 0.011809507 0.011689438
26
+ TurnOnStove 0.014633293 0.015040388 0.013856436
27
+ AdjustToasterOvenTemperature 0.012944635 0.013481709 0.012874456
28
+ AVG 0.014639860 0.014875111 0.014533606
rkd_v2_1/rkd_v2_1_angle/default/phase3/task_fm_loss_1024_seed42.json ADDED
@@ -0,0 +1,244 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_path": "/home/ext_minje/groot_robocasa/robocasa_v2/rkd_v2_1/rkd_v2_1_angle/default/phase3",
3
+ "config": "/home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/processing_line_only_v2/RKD_v2_1/rkd_v2_1_a.yaml",
4
+ "batch_size": 32,
5
+ "max_samples_per_task": 1024,
6
+ "seed": 42,
7
+ "mean_task_fm_loss": 0.01487511117403091,
8
+ "results": [
9
+ {
10
+ "task": "CloseBlenderLid",
11
+ "fm_loss": 0.012893490929855034,
12
+ "samples": 1024,
13
+ "batches": 32,
14
+ "dataset_len": 34953,
15
+ "sampled": true,
16
+ "planned_samples": 1024
17
+ },
18
+ {
19
+ "task": "CloseFridge",
20
+ "fm_loss": 0.01806122387642972,
21
+ "samples": 1024,
22
+ "batches": 32,
23
+ "dataset_len": 25750,
24
+ "sampled": true,
25
+ "planned_samples": 1024
26
+ },
27
+ {
28
+ "task": "CloseToasterOvenDoor",
29
+ "fm_loss": 0.012771354464348406,
30
+ "samples": 1024,
31
+ "batches": 32,
32
+ "dataset_len": 17858,
33
+ "sampled": true,
34
+ "planned_samples": 1024
35
+ },
36
+ {
37
+ "task": "CoffeeSetupMug",
38
+ "fm_loss": 0.013651895205839537,
39
+ "samples": 1024,
40
+ "batches": 32,
41
+ "dataset_len": 22529,
42
+ "sampled": true,
43
+ "planned_samples": 1024
44
+ },
45
+ {
46
+ "task": "NavigateKitchen",
47
+ "fm_loss": 0.01392564426350873,
48
+ "samples": 1024,
49
+ "batches": 32,
50
+ "dataset_len": 16346,
51
+ "sampled": true,
52
+ "planned_samples": 1024
53
+ },
54
+ {
55
+ "task": "OpenCabinet",
56
+ "fm_loss": 0.01475983289128635,
57
+ "samples": 1024,
58
+ "batches": 32,
59
+ "dataset_len": 35393,
60
+ "sampled": true,
61
+ "planned_samples": 1024
62
+ },
63
+ {
64
+ "task": "OpenDrawer",
65
+ "fm_loss": 0.013188227690989152,
66
+ "samples": 1024,
67
+ "batches": 32,
68
+ "dataset_len": 20027,
69
+ "sampled": true,
70
+ "planned_samples": 1024
71
+ },
72
+ {
73
+ "task": "OpenStandMixerHead",
74
+ "fm_loss": 0.015075979434186593,
75
+ "samples": 1024,
76
+ "batches": 32,
77
+ "dataset_len": 12400,
78
+ "sampled": true,
79
+ "planned_samples": 1024
80
+ },
81
+ {
82
+ "task": "PickPlaceCounterToCabinet",
83
+ "fm_loss": 0.01609046253724955,
84
+ "samples": 1024,
85
+ "batches": 32,
86
+ "dataset_len": 22582,
87
+ "sampled": true,
88
+ "planned_samples": 1024
89
+ },
90
+ {
91
+ "task": "PickPlaceCounterToStove",
92
+ "fm_loss": 0.014552779990481213,
93
+ "samples": 1024,
94
+ "batches": 32,
95
+ "dataset_len": 22279,
96
+ "sampled": true,
97
+ "planned_samples": 1024
98
+ },
99
+ {
100
+ "task": "PickPlaceDrawerToCounter",
101
+ "fm_loss": 0.01606237472151406,
102
+ "samples": 1024,
103
+ "batches": 32,
104
+ "dataset_len": 30785,
105
+ "sampled": true,
106
+ "planned_samples": 1024
107
+ },
108
+ {
109
+ "task": "PickPlaceSinkToCounter",
110
+ "fm_loss": 0.015319367055781186,
111
+ "samples": 1024,
112
+ "batches": 32,
113
+ "dataset_len": 24529,
114
+ "sampled": true,
115
+ "planned_samples": 1024
116
+ },
117
+ {
118
+ "task": "PickPlaceToasterToCounter",
119
+ "fm_loss": 0.01378093523089774,
120
+ "samples": 1024,
121
+ "batches": 32,
122
+ "dataset_len": 25630,
123
+ "sampled": true,
124
+ "planned_samples": 1024
125
+ },
126
+ {
127
+ "task": "PickPlaceCabinetToCounter",
128
+ "fm_loss": 0.01903710834449157,
129
+ "samples": 1024,
130
+ "batches": 32,
131
+ "dataset_len": 19090,
132
+ "sampled": true,
133
+ "planned_samples": 1024
134
+ },
135
+ {
136
+ "task": "TurnOnElectricKettle",
137
+ "fm_loss": 0.01817811105865985,
138
+ "samples": 1024,
139
+ "batches": 32,
140
+ "dataset_len": 11625,
141
+ "sampled": true,
142
+ "planned_samples": 1024
143
+ },
144
+ {
145
+ "task": "TurnOnMicrowave",
146
+ "fm_loss": 0.01941173564409837,
147
+ "samples": 1024,
148
+ "batches": 32,
149
+ "dataset_len": 13038,
150
+ "sampled": true,
151
+ "planned_samples": 1024
152
+ },
153
+ {
154
+ "task": "TurnOffMicrowave",
155
+ "fm_loss": 0.020360561029519886,
156
+ "samples": 1024,
157
+ "batches": 32,
158
+ "dataset_len": 14175,
159
+ "sampled": true,
160
+ "planned_samples": 1024
161
+ },
162
+ {
163
+ "task": "OpenElectricKettleLid",
164
+ "fm_loss": 0.013258201302960515,
165
+ "samples": 1024,
166
+ "batches": 32,
167
+ "dataset_len": 9928,
168
+ "sampled": true,
169
+ "planned_samples": 1024
170
+ },
171
+ {
172
+ "task": "SlideDishwasherRack",
173
+ "fm_loss": 0.011744259303668514,
174
+ "samples": 1024,
175
+ "batches": 32,
176
+ "dataset_len": 19052,
177
+ "sampled": true,
178
+ "planned_samples": 1024
179
+ },
180
+ {
181
+ "task": "TurnOnSinkFaucet",
182
+ "fm_loss": 0.009852687799138948,
183
+ "samples": 1024,
184
+ "batches": 32,
185
+ "dataset_len": 22234,
186
+ "sampled": true,
187
+ "planned_samples": 1024
188
+ },
189
+ {
190
+ "task": "TurnSinkSpout",
191
+ "fm_loss": 0.017699005256872624,
192
+ "samples": 1024,
193
+ "batches": 32,
194
+ "dataset_len": 10219,
195
+ "sampled": true,
196
+ "planned_samples": 1024
197
+ },
198
+ {
199
+ "task": "TurnOffSinkFaucet",
200
+ "fm_loss": 0.015963105455739424,
201
+ "samples": 1024,
202
+ "batches": 32,
203
+ "dataset_len": 11626,
204
+ "sampled": true,
205
+ "planned_samples": 1024
206
+ },
207
+ {
208
+ "task": "AdjustWaterTemperature",
209
+ "fm_loss": 0.010782943121739663,
210
+ "samples": 1024,
211
+ "batches": 32,
212
+ "dataset_len": 19725,
213
+ "sampled": true,
214
+ "planned_samples": 1024
215
+ },
216
+ {
217
+ "task": "TurnOffStove",
218
+ "fm_loss": 0.011809507021098398,
219
+ "samples": 1024,
220
+ "batches": 32,
221
+ "dataset_len": 29821,
222
+ "sampled": true,
223
+ "planned_samples": 1024
224
+ },
225
+ {
226
+ "task": "TurnOnStove",
227
+ "fm_loss": 0.01504038815619424,
228
+ "samples": 1024,
229
+ "batches": 32,
230
+ "dataset_len": 17322,
231
+ "sampled": true,
232
+ "planned_samples": 1024
233
+ },
234
+ {
235
+ "task": "AdjustToasterOvenTemperature",
236
+ "fm_loss": 0.013481708738254383,
237
+ "samples": 1024,
238
+ "batches": 32,
239
+ "dataset_len": 19846,
240
+ "sampled": true,
241
+ "planned_samples": 1024
242
+ }
243
+ ]
244
+ }