Hugging Face's logo

nvidia
/
PhysicalAI-Robotics-mindmap-Checkpoints

Model card Files
xet
Community
4
remostei commited on
Commit
0b72465
·
verified ·
1 Parent(s): 723e380

Upload 12 files

Browse files
Files changed (12) hide show
  1. cube_stacking_checkpoint/best.pth +3 -0
  2. cube_stacking_checkpoint/last.pth +3 -0
  3. cube_stacking_checkpoint/training_args.json +177 -0
  4. drill_in_box_checkpoint/best.pth +3 -0
  5. drill_in_box_checkpoint/last.pth +3 -0
  6. drill_in_box_checkpoint/training_args.json +177 -0
  7. mug_in_drawer_checkpoint/best.pth +3 -0
  8. mug_in_drawer_checkpoint/last.pth +3 -0
  9. mug_in_drawer_checkpoint/training_args.json +177 -0
  10. stick_in_bin_checkpoint/best.pth +3 -0
  11. stick_in_bin_checkpoint/last.pth +3 -0
  12. stick_in_bin_checkpoint/training_args.json +177 -0
cube_stacking_checkpoint/best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10645d835ab47b05dff11b9ad3f4ced100590c343892f9aecdd276bbf7087afe
3
+ size 27422483
cube_stacking_checkpoint/last.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f09e2a7b864c90f6b4e8eb26af5e8ed21b0da1fe9d0b73211c961ea50552ff66
3
+ size 27422483
cube_stacking_checkpoint/training_args.json ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "accumulate_grad_batches": 1,
3
+ "add_external_cam": false,
4
+ "apply_geometry_noise": 0,
5
+ "apply_random_transforms": 0,
6
+ "base_log_dir": {
7
+ "_string": "/eval/train_logs",
8
+ "_type": "python_object (type = PosixPath)",
9
+ "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
10
+ },
11
+ "batch_size": 16,
12
+ "batch_size_val": 16,
13
+ "checkpoint": null,
14
+ "data_type": {
15
+ "_string": "DataType.RGBD_AND_MESH",
16
+ "_type": "python_object (type = DataType)",
17
+ "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
18
+ },
19
+ "dataset": "/datasets/mindmap_cubes_rgbdmesh_radioB_1cam/",
20
+ "demos_datagen": "0",
21
+ "demos_train": "0-7",
22
+ "demos_valset": "8-9",
23
+ "diffusion_dropout": 0.0,
24
+ "diffusion_timesteps": 100,
25
+ "disable_visualizer_wait_on_key": false,
26
+ "embedding_dim": 120,
27
+ "encode_openness": 1,
28
+ "encoder_dropout": 0.0,
29
+ "eval_only": false,
30
+ "exp_name": "mindmap Training",
31
+ "extra_keyposes_around_grasp_events": null,
32
+ "feature_image_size": {
33
+ "_type": "tuple",
34
+ "_value": [
35
+ 32,
36
+ 32
37
+ ]
38
+ },
39
+ "feature_type": {
40
+ "_string": "FeatureExtractorType.RADIO_V25_B",
41
+ "_type": "python_object (type = FeatureExtractorType)",
42
+ "_value": "gASVWgAAAAAAAACMK21pbmRtYXAuaW1hZ2VfcHJvY2Vzc2luZy5mZWF0dXJlX2V4dHJhY3Rpb26UjBRGZWF0dXJlRXh0cmFjdG9yVHlwZZSTlIwLcmFkaW9fdjI1X2KUhZRSlC4="
43
+ },
44
+ "fpn_checkpoint": null,
45
+ "fps_subsampling_factor": 5,
46
+ "gripper_encoding_mode": "binary",
47
+ "gripper_loss": 1.0,
48
+ "ignore_model_args_json": false,
49
+ "image_size": {
50
+ "_type": "tuple",
51
+ "_value": [
52
+ 512,
53
+ 512
54
+ ]
55
+ },
56
+ "include_dynamic": false,
57
+ "include_failed_demos": false,
58
+ "initial_learning_rate": 0.0001,
59
+ "instructions": null,
60
+ "keypose_detection_mode": null,
61
+ "lang_enhanced": 0,
62
+ "learning_rate_convergence_percentage": 0.75,
63
+ "learning_rate_end_factor": 0.5,
64
+ "max_episode_length": 5,
65
+ "max_episodes_per_task": 100,
66
+ "num_batches_per_test_eval": 1,
67
+ "num_batches_per_train_eval": 1,
68
+ "num_history": 3,
69
+ "num_vertices_to_sample": 2048,
70
+ "num_vis_ins_attn_layers": 2,
71
+ "num_workers": 0,
72
+ "num_workers_for_test_dataset": null,
73
+ "only_sample_keyposes": false,
74
+ "pos_loss": 30.0,
75
+ "pos_noise_stddev_m": 0.01,
76
+ "prediction_horizon": 1,
77
+ "predictor_dropout": 0.0,
78
+ "print_progress_freq": 100,
79
+ "print_timers_freq": 1000,
80
+ "projective_appearance_integrator_measurement_weight": null,
81
+ "quaternion_format": "wxyz",
82
+ "random_rpy_range_deg": {
83
+ "_type": "tuple",
84
+ "_value": [
85
+ [
86
+ 0.0,
87
+ 0.0,
88
+ -90.0
89
+ ],
90
+ [
91
+ 0.0,
92
+ 0.0,
93
+ 90.0
94
+ ]
95
+ ]
96
+ },
97
+ "random_translation_range_m": {
98
+ "_type": "tuple",
99
+ "_value": [
100
+ [
101
+ -0.1,
102
+ -0.1,
103
+ 0.0
104
+ ],
105
+ [
106
+ 0.1,
107
+ 0.1,
108
+ 0.0
109
+ ]
110
+ ]
111
+ },
112
+ "relative_action": 0,
113
+ "reproducibility": {
114
+ "command_line": "python run_training.py --task cube_stacking --data_type rgbd_and_mesh --feature_type radio_v25_b --demos_train 0-7 --demos_valset 8-9 --dataset /datasets/mindmap_cubes_rgbdmesh_radioB_1cam/",
115
+ "git_has_uncommitted_changes": true,
116
+ "git_root": "/workspaces/mindmap",
117
+ "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
118
+ "time": "Tue Sep 9 18:59:03 2025"
119
+ },
120
+ "rgbd_min_depth_threshold": 0.0,
121
+ "rot_loss": 10.0,
122
+ "rot_noise_stddev_deg": 0.01,
123
+ "rotation_parametrization": "6D_from_query",
124
+ "sampling_weighting_type": "uniform",
125
+ "save_checkpoint": true,
126
+ "save_serialized_nvblox_map_to_disk": false,
127
+ "seed": 0,
128
+ "skip_train_val": false,
129
+ "task": {
130
+ "_string": "Tasks.CUBE_STACKING",
131
+ "_type": "python_object (type = Tasks)",
132
+ "_value": "gASVNQAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwNY3ViZV9zdGFja2luZ5SFlFKULg=="
133
+ },
134
+ "train_iters": 24000,
135
+ "use_fps": 1,
136
+ "use_instruction": 0,
137
+ "use_keyposes": 1,
138
+ "use_shared_feature_encoder": 0,
139
+ "val_freq": 12,
140
+ "validate_demos_with_gt_poses": 1,
141
+ "variations": {
142
+ "_type": "tuple",
143
+ "_value": [
144
+ 0
145
+ ]
146
+ },
147
+ "vertex_sampling_method": {
148
+ "_string": "VertexSamplingMethod.RANDOM_WITHOUT_REPLACEMENT",
149
+ "_type": "python_object (type = VertexSamplingMethod)",
150
+ "_value": "gASVYgAAAAAAAACMJG1pbmRtYXAuZGF0YV9sb2FkaW5nLnZlcnRleF9zYW1wbGluZ5SMFFZlcnRleFNhbXBsaW5nTWV0aG9klJOUjBpyYW5kb21fd2l0aG91dF9yZXBsYWNlbWVudJSFlFKULg=="
151
+ },
152
+ "visualize": false,
153
+ "visualize_aabb": false,
154
+ "visualize_attention_weights": false,
155
+ "visualize_backprojected_features": false,
156
+ "visualize_encoded_features": true,
157
+ "visualizer_background_rgb": {
158
+ "_type": "tuple",
159
+ "_value": [
160
+ 0.0,
161
+ 0.0,
162
+ 0.0
163
+ ]
164
+ },
165
+ "visualizer_min_attention_weight": 0.0,
166
+ "visualizer_min_tsdf_weight": 0.0,
167
+ "visualizer_point_size": 3,
168
+ "visualizer_pointcloud_max_distance": null,
169
+ "visualizer_pointclouds_ply_output_dir": null,
170
+ "visualizer_record_camera_output_path": null,
171
+ "visualizer_voxel_size_m": 0.01,
172
+ "viz_freq": 200,
173
+ "voxel_size_m": null,
174
+ "wandb_entity": "nv-welcome",
175
+ "wandb_mode": "online",
176
+ "wandb_name": null
177
+ }
drill_in_box_checkpoint/best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4527bb0cc33c63f7399deddc506fa69fd08984b3388d90b39026f9dc2c29730
3
+ size 28336963
drill_in_box_checkpoint/last.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:266a80a8d232ef7aa5fc32f76bd80bdec86dc62c7633f186b26282733ff416d8
3
+ size 28336963
drill_in_box_checkpoint/training_args.json ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "accumulate_grad_batches": 1,
3
+ "add_external_cam": false,
4
+ "apply_geometry_noise": 0,
5
+ "apply_random_transforms": 0,
6
+ "base_log_dir": {
7
+ "_string": "/eval/train_logs",
8
+ "_type": "python_object (type = PosixPath)",
9
+ "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
10
+ },
11
+ "batch_size": 16,
12
+ "batch_size_val": 16,
13
+ "checkpoint": null,
14
+ "data_type": {
15
+ "_string": "DataType.RGBD_AND_MESH",
16
+ "_type": "python_object (type = DataType)",
17
+ "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
18
+ },
19
+ "dataset": "/datasets/mindmap_galileo_gr1_rgbdmesh_radioB_1cam/",
20
+ "demos_datagen": "0",
21
+ "demos_train": "0-7",
22
+ "demos_valset": "8-9",
23
+ "diffusion_dropout": 0.0,
24
+ "diffusion_timesteps": 100,
25
+ "disable_visualizer_wait_on_key": false,
26
+ "embedding_dim": 120,
27
+ "encode_openness": 1,
28
+ "encoder_dropout": 0.0,
29
+ "eval_only": false,
30
+ "exp_name": "mindmap Training",
31
+ "extra_keyposes_around_grasp_events": null,
32
+ "feature_image_size": {
33
+ "_type": "tuple",
34
+ "_value": [
35
+ 32,
36
+ 32
37
+ ]
38
+ },
39
+ "feature_type": {
40
+ "_string": "FeatureExtractorType.RADIO_V25_B",
41
+ "_type": "python_object (type = FeatureExtractorType)",
42
+ "_value": "gASVWgAAAAAAAACMK21pbmRtYXAuaW1hZ2VfcHJvY2Vzc2luZy5mZWF0dXJlX2V4dHJhY3Rpb26UjBRGZWF0dXJlRXh0cmFjdG9yVHlwZZSTlIwLcmFkaW9fdjI1X2KUhZRSlC4="
43
+ },
44
+ "fpn_checkpoint": null,
45
+ "fps_subsampling_factor": 5,
46
+ "gripper_encoding_mode": "binary",
47
+ "gripper_loss": 1.0,
48
+ "ignore_model_args_json": false,
49
+ "image_size": {
50
+ "_type": "tuple",
51
+ "_value": [
52
+ 512,
53
+ 512
54
+ ]
55
+ },
56
+ "include_dynamic": false,
57
+ "include_failed_demos": false,
58
+ "initial_learning_rate": 0.0001,
59
+ "instructions": null,
60
+ "keypose_detection_mode": null,
61
+ "lang_enhanced": 0,
62
+ "learning_rate_convergence_percentage": 0.75,
63
+ "learning_rate_end_factor": 0.5,
64
+ "max_episode_length": 5,
65
+ "max_episodes_per_task": 100,
66
+ "num_batches_per_test_eval": 1,
67
+ "num_batches_per_train_eval": 1,
68
+ "num_history": 3,
69
+ "num_vertices_to_sample": 2048,
70
+ "num_vis_ins_attn_layers": 2,
71
+ "num_workers": 0,
72
+ "num_workers_for_test_dataset": null,
73
+ "only_sample_keyposes": false,
74
+ "pos_loss": 30.0,
75
+ "pos_noise_stddev_m": 0.01,
76
+ "prediction_horizon": 1,
77
+ "predictor_dropout": 0.0,
78
+ "print_progress_freq": 100,
79
+ "print_timers_freq": 1000,
80
+ "projective_appearance_integrator_measurement_weight": null,
81
+ "quaternion_format": "wxyz",
82
+ "random_rpy_range_deg": {
83
+ "_type": "tuple",
84
+ "_value": [
85
+ [
86
+ 0.0,
87
+ 0.0,
88
+ -90.0
89
+ ],
90
+ [
91
+ 0.0,
92
+ 0.0,
93
+ 90.0
94
+ ]
95
+ ]
96
+ },
97
+ "random_translation_range_m": {
98
+ "_type": "tuple",
99
+ "_value": [
100
+ [
101
+ -0.1,
102
+ -0.1,
103
+ 0.0
104
+ ],
105
+ [
106
+ 0.1,
107
+ 0.1,
108
+ 0.0
109
+ ]
110
+ ]
111
+ },
112
+ "relative_action": 0,
113
+ "reproducibility": {
114
+ "command_line": "python run_training.py --task drill_in_box --data_type rgbd_and_mesh --feature_type radio_v25_b --demos_train 0-7 --demos_valset 8-9 --dataset /datasets/mindmap_galileo_gr1_rgbdmesh_radioB_1cam/",
115
+ "git_has_uncommitted_changes": true,
116
+ "git_root": "/workspaces/mindmap",
117
+ "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
118
+ "time": "Tue Sep 9 19:03:07 2025"
119
+ },
120
+ "rgbd_min_depth_threshold": 0.0,
121
+ "rot_loss": 10.0,
122
+ "rot_noise_stddev_deg": 0.01,
123
+ "rotation_parametrization": "6D_from_query",
124
+ "sampling_weighting_type": "uniform",
125
+ "save_checkpoint": true,
126
+ "save_serialized_nvblox_map_to_disk": false,
127
+ "seed": 0,
128
+ "skip_train_val": false,
129
+ "task": {
130
+ "_string": "Tasks.DRILL_IN_BOX",
131
+ "_type": "python_object (type = Tasks)",
132
+ "_value": "gASVNAAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwMZHJpbGxfaW5fYm94lIWUUpQu"
133
+ },
134
+ "train_iters": 24000,
135
+ "use_fps": 1,
136
+ "use_instruction": 0,
137
+ "use_keyposes": 1,
138
+ "use_shared_feature_encoder": 0,
139
+ "val_freq": 12,
140
+ "validate_demos_with_gt_poses": 1,
141
+ "variations": {
142
+ "_type": "tuple",
143
+ "_value": [
144
+ 0
145
+ ]
146
+ },
147
+ "vertex_sampling_method": {
148
+ "_string": "VertexSamplingMethod.RANDOM_WITHOUT_REPLACEMENT",
149
+ "_type": "python_object (type = VertexSamplingMethod)",
150
+ "_value": "gASVYgAAAAAAAACMJG1pbmRtYXAuZGF0YV9sb2FkaW5nLnZlcnRleF9zYW1wbGluZ5SMFFZlcnRleFNhbXBsaW5nTWV0aG9klJOUjBpyYW5kb21fd2l0aG91dF9yZXBsYWNlbWVudJSFlFKULg=="
151
+ },
152
+ "visualize": false,
153
+ "visualize_aabb": false,
154
+ "visualize_attention_weights": false,
155
+ "visualize_backprojected_features": false,
156
+ "visualize_encoded_features": true,
157
+ "visualizer_background_rgb": {
158
+ "_type": "tuple",
159
+ "_value": [
160
+ 0.0,
161
+ 0.0,
162
+ 0.0
163
+ ]
164
+ },
165
+ "visualizer_min_attention_weight": 0.0,
166
+ "visualizer_min_tsdf_weight": 0.0,
167
+ "visualizer_point_size": 3,
168
+ "visualizer_pointcloud_max_distance": null,
169
+ "visualizer_pointclouds_ply_output_dir": null,
170
+ "visualizer_record_camera_output_path": null,
171
+ "visualizer_voxel_size_m": 0.01,
172
+ "viz_freq": 200,
173
+ "voxel_size_m": null,
174
+ "wandb_entity": "nv-welcome",
175
+ "wandb_mode": "online",
176
+ "wandb_name": null
177
+ }
mug_in_drawer_checkpoint/best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d70542250f60764a4736778de6007282c74f5756a535a5ec84794e6aa9ad0df7
3
+ size 27422483
mug_in_drawer_checkpoint/last.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03ec63d77e46cdc7e807be5a198716232f95a12d37701206bd45effe26fdcbca
3
+ size 27422483
mug_in_drawer_checkpoint/training_args.json ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "accumulate_grad_batches": 1,
3
+ "add_external_cam": false,
4
+ "apply_geometry_noise": 0,
5
+ "apply_random_transforms": 0,
6
+ "base_log_dir": {
7
+ "_string": "/eval/train_logs",
8
+ "_type": "python_object (type = PosixPath)",
9
+ "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
10
+ },
11
+ "batch_size": 16,
12
+ "batch_size_val": 16,
13
+ "checkpoint": null,
14
+ "data_type": {
15
+ "_string": "DataType.RGBD_AND_MESH",
16
+ "_type": "python_object (type = DataType)",
17
+ "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
18
+ },
19
+ "dataset": "/datasets/mindmap_mug_rgbdmesh_radioB_1cam/",
20
+ "demos_datagen": "0",
21
+ "demos_train": "0-7",
22
+ "demos_valset": "8-9",
23
+ "diffusion_dropout": 0.0,
24
+ "diffusion_timesteps": 100,
25
+ "disable_visualizer_wait_on_key": false,
26
+ "embedding_dim": 120,
27
+ "encode_openness": 1,
28
+ "encoder_dropout": 0.0,
29
+ "eval_only": false,
30
+ "exp_name": "mindmap Training",
31
+ "extra_keyposes_around_grasp_events": null,
32
+ "feature_image_size": {
33
+ "_type": "tuple",
34
+ "_value": [
35
+ 32,
36
+ 32
37
+ ]
38
+ },
39
+ "feature_type": {
40
+ "_string": "FeatureExtractorType.RADIO_V25_B",
41
+ "_type": "python_object (type = FeatureExtractorType)",
42
+ "_value": "gASVWgAAAAAAAACMK21pbmRtYXAuaW1hZ2VfcHJvY2Vzc2luZy5mZWF0dXJlX2V4dHJhY3Rpb26UjBRGZWF0dXJlRXh0cmFjdG9yVHlwZZSTlIwLcmFkaW9fdjI1X2KUhZRSlC4="
43
+ },
44
+ "fpn_checkpoint": null,
45
+ "fps_subsampling_factor": 5,
46
+ "gripper_encoding_mode": "binary",
47
+ "gripper_loss": 1.0,
48
+ "ignore_model_args_json": false,
49
+ "image_size": {
50
+ "_type": "tuple",
51
+ "_value": [
52
+ 512,
53
+ 512
54
+ ]
55
+ },
56
+ "include_dynamic": false,
57
+ "include_failed_demos": false,
58
+ "initial_learning_rate": 0.0001,
59
+ "instructions": null,
60
+ "keypose_detection_mode": null,
61
+ "lang_enhanced": 0,
62
+ "learning_rate_convergence_percentage": 0.75,
63
+ "learning_rate_end_factor": 0.5,
64
+ "max_episode_length": 5,
65
+ "max_episodes_per_task": 100,
66
+ "num_batches_per_test_eval": 1,
67
+ "num_batches_per_train_eval": 1,
68
+ "num_history": 3,
69
+ "num_vertices_to_sample": 2048,
70
+ "num_vis_ins_attn_layers": 2,
71
+ "num_workers": 0,
72
+ "num_workers_for_test_dataset": null,
73
+ "only_sample_keyposes": false,
74
+ "pos_loss": 30.0,
75
+ "pos_noise_stddev_m": 0.01,
76
+ "prediction_horizon": 1,
77
+ "predictor_dropout": 0.0,
78
+ "print_progress_freq": 100,
79
+ "print_timers_freq": 1000,
80
+ "projective_appearance_integrator_measurement_weight": null,
81
+ "quaternion_format": "wxyz",
82
+ "random_rpy_range_deg": {
83
+ "_type": "tuple",
84
+ "_value": [
85
+ [
86
+ 0.0,
87
+ 0.0,
88
+ -90.0
89
+ ],
90
+ [
91
+ 0.0,
92
+ 0.0,
93
+ 90.0
94
+ ]
95
+ ]
96
+ },
97
+ "random_translation_range_m": {
98
+ "_type": "tuple",
99
+ "_value": [
100
+ [
101
+ -0.1,
102
+ -0.1,
103
+ 0.0
104
+ ],
105
+ [
106
+ 0.1,
107
+ 0.1,
108
+ 0.0
109
+ ]
110
+ ]
111
+ },
112
+ "relative_action": 0,
113
+ "reproducibility": {
114
+ "command_line": "python run_training.py --task mug_in_drawer --data_type rgbd_and_mesh --feature_type radio_v25_b --demos_train 0-7 --demos_valset 8-9 --dataset /datasets/mindmap_mug_rgbdmesh_radioB_1cam/",
115
+ "git_has_uncommitted_changes": true,
116
+ "git_root": "/workspaces/mindmap",
117
+ "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
118
+ "time": "Tue Sep 9 19:01:07 2025"
119
+ },
120
+ "rgbd_min_depth_threshold": 0.0,
121
+ "rot_loss": 10.0,
122
+ "rot_noise_stddev_deg": 0.01,
123
+ "rotation_parametrization": "6D_from_query",
124
+ "sampling_weighting_type": "uniform",
125
+ "save_checkpoint": true,
126
+ "save_serialized_nvblox_map_to_disk": false,
127
+ "seed": 0,
128
+ "skip_train_val": false,
129
+ "task": {
130
+ "_string": "Tasks.MUG_IN_DRAWER",
131
+ "_type": "python_object (type = Tasks)",
132
+ "_value": "gASVNQAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwNbXVnX2luX2RyYXdlcpSFlFKULg=="
133
+ },
134
+ "train_iters": 24000,
135
+ "use_fps": 1,
136
+ "use_instruction": 0,
137
+ "use_keyposes": 1,
138
+ "use_shared_feature_encoder": 0,
139
+ "val_freq": 12,
140
+ "validate_demos_with_gt_poses": 1,
141
+ "variations": {
142
+ "_type": "tuple",
143
+ "_value": [
144
+ 0
145
+ ]
146
+ },
147
+ "vertex_sampling_method": {
148
+ "_string": "VertexSamplingMethod.RANDOM_WITHOUT_REPLACEMENT",
149
+ "_type": "python_object (type = VertexSamplingMethod)",
150
+ "_value": "gASVYgAAAAAAAACMJG1pbmRtYXAuZGF0YV9sb2FkaW5nLnZlcnRleF9zYW1wbGluZ5SMFFZlcnRleFNhbXBsaW5nTWV0aG9klJOUjBpyYW5kb21fd2l0aG91dF9yZXBsYWNlbWVudJSFlFKULg=="
151
+ },
152
+ "visualize": false,
153
+ "visualize_aabb": false,
154
+ "visualize_attention_weights": false,
155
+ "visualize_backprojected_features": false,
156
+ "visualize_encoded_features": true,
157
+ "visualizer_background_rgb": {
158
+ "_type": "tuple",
159
+ "_value": [
160
+ 0.0,
161
+ 0.0,
162
+ 0.0
163
+ ]
164
+ },
165
+ "visualizer_min_attention_weight": 0.0,
166
+ "visualizer_min_tsdf_weight": 0.0,
167
+ "visualizer_point_size": 3,
168
+ "visualizer_pointcloud_max_distance": null,
169
+ "visualizer_pointclouds_ply_output_dir": null,
170
+ "visualizer_record_camera_output_path": null,
171
+ "visualizer_voxel_size_m": 0.01,
172
+ "viz_freq": 200,
173
+ "voxel_size_m": null,
174
+ "wandb_entity": "nv-welcome",
175
+ "wandb_mode": "online",
176
+ "wandb_name": null
177
+ }
stick_in_bin_checkpoint/best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc45a68b3c437dac08b530e65fa138b5522fbe4f1460b1851abe4ae53b13a45f
3
+ size 28336963
stick_in_bin_checkpoint/last.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:809947b0da2c4486ba8f016d038ac0e6c72056dc2a049c01257be5cc6494c697
3
+ size 28336963
stick_in_bin_checkpoint/training_args.json ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "accumulate_grad_batches": 1,
3
+ "add_external_cam": false,
4
+ "apply_geometry_noise": 0,
5
+ "apply_random_transforms": 0,
6
+ "base_log_dir": {
7
+ "_string": "/eval/train_logs",
8
+ "_type": "python_object (type = PosixPath)",
9
+ "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
10
+ },
11
+ "batch_size": 16,
12
+ "batch_size_val": 16,
13
+ "checkpoint": null,
14
+ "data_type": {
15
+ "_string": "DataType.RGBD_AND_MESH",
16
+ "_type": "python_object (type = DataType)",
17
+ "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
18
+ },
19
+ "dataset": "/datasets/mindmap_galileo_gr1_rgbdmesh_radioB_1cam/",
20
+ "demos_datagen": "0",
21
+ "demos_train": "0-6",
22
+ "demos_valset": "0-6",
23
+ "diffusion_dropout": 0.0,
24
+ "diffusion_timesteps": 100,
25
+ "disable_visualizer_wait_on_key": false,
26
+ "embedding_dim": 120,
27
+ "encode_openness": 1,
28
+ "encoder_dropout": 0.0,
29
+ "eval_only": false,
30
+ "exp_name": "mindmap Training",
31
+ "extra_keyposes_around_grasp_events": null,
32
+ "feature_image_size": {
33
+ "_type": "tuple",
34
+ "_value": [
35
+ 32,
36
+ 32
37
+ ]
38
+ },
39
+ "feature_type": {
40
+ "_string": "FeatureExtractorType.RADIO_V25_B",
41
+ "_type": "python_object (type = FeatureExtractorType)",
42
+ "_value": "gASVWgAAAAAAAACMK21pbmRtYXAuaW1hZ2VfcHJvY2Vzc2luZy5mZWF0dXJlX2V4dHJhY3Rpb26UjBRGZWF0dXJlRXh0cmFjdG9yVHlwZZSTlIwLcmFkaW9fdjI1X2KUhZRSlC4="
43
+ },
44
+ "fpn_checkpoint": null,
45
+ "fps_subsampling_factor": 5,
46
+ "gripper_encoding_mode": "binary",
47
+ "gripper_loss": 1.0,
48
+ "ignore_model_args_json": false,
49
+ "image_size": {
50
+ "_type": "tuple",
51
+ "_value": [
52
+ 512,
53
+ 512
54
+ ]
55
+ },
56
+ "include_dynamic": false,
57
+ "include_failed_demos": false,
58
+ "initial_learning_rate": 0.0001,
59
+ "instructions": null,
60
+ "keypose_detection_mode": null,
61
+ "lang_enhanced": 0,
62
+ "learning_rate_convergence_percentage": 0.75,
63
+ "learning_rate_end_factor": 0.5,
64
+ "max_episode_length": 5,
65
+ "max_episodes_per_task": 100,
66
+ "num_batches_per_test_eval": 1,
67
+ "num_batches_per_train_eval": 1,
68
+ "num_history": 3,
69
+ "num_vertices_to_sample": 2048,
70
+ "num_vis_ins_attn_layers": 2,
71
+ "num_workers": 0,
72
+ "num_workers_for_test_dataset": null,
73
+ "only_sample_keyposes": false,
74
+ "pos_loss": 30.0,
75
+ "pos_noise_stddev_m": 0.01,
76
+ "prediction_horizon": 1,
77
+ "predictor_dropout": 0.0,
78
+ "print_progress_freq": 100,
79
+ "print_timers_freq": 1000,
80
+ "projective_appearance_integrator_measurement_weight": null,
81
+ "quaternion_format": "wxyz",
82
+ "random_rpy_range_deg": {
83
+ "_type": "tuple",
84
+ "_value": [
85
+ [
86
+ 0.0,
87
+ 0.0,
88
+ -90.0
89
+ ],
90
+ [
91
+ 0.0,
92
+ 0.0,
93
+ 90.0
94
+ ]
95
+ ]
96
+ },
97
+ "random_translation_range_m": {
98
+ "_type": "tuple",
99
+ "_value": [
100
+ [
101
+ -0.1,
102
+ -0.1,
103
+ 0.0
104
+ ],
105
+ [
106
+ 0.1,
107
+ 0.1,
108
+ 0.0
109
+ ]
110
+ ]
111
+ },
112
+ "relative_action": 0,
113
+ "reproducibility": {
114
+ "command_line": "python run_training.py --task stick_in_bin --data_type rgbd_and_mesh --feature_type radio_v25_b --demos_train 0-6 --demos_valset 0-6 --dataset /datasets/mindmap_galileo_gr1_rgbdmesh_radioB_1cam/",
115
+ "git_has_uncommitted_changes": true,
116
+ "git_root": "/workspaces/mindmap",
117
+ "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
118
+ "time": "Tue Sep 9 19:09:08 2025"
119
+ },
120
+ "rgbd_min_depth_threshold": 0.0,
121
+ "rot_loss": 10.0,
122
+ "rot_noise_stddev_deg": 0.01,
123
+ "rotation_parametrization": "6D_from_query",
124
+ "sampling_weighting_type": "uniform",
125
+ "save_checkpoint": true,
126
+ "save_serialized_nvblox_map_to_disk": false,
127
+ "seed": 0,
128
+ "skip_train_val": false,
129
+ "task": {
130
+ "_string": "Tasks.STICK_IN_BIN",
131
+ "_type": "python_object (type = Tasks)",
132
+ "_value": "gASVNAAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwMc3RpY2tfaW5fYmlulIWUUpQu"
133
+ },
134
+ "train_iters": 24000,
135
+ "use_fps": 1,
136
+ "use_instruction": 0,
137
+ "use_keyposes": 1,
138
+ "use_shared_feature_encoder": 0,
139
+ "val_freq": 12,
140
+ "validate_demos_with_gt_poses": 1,
141
+ "variations": {
142
+ "_type": "tuple",
143
+ "_value": [
144
+ 0
145
+ ]
146
+ },
147
+ "vertex_sampling_method": {
148
+ "_string": "VertexSamplingMethod.RANDOM_WITHOUT_REPLACEMENT",
149
+ "_type": "python_object (type = VertexSamplingMethod)",
150
+ "_value": "gASVYgAAAAAAAACMJG1pbmRtYXAuZGF0YV9sb2FkaW5nLnZlcnRleF9zYW1wbGluZ5SMFFZlcnRleFNhbXBsaW5nTWV0aG9klJOUjBpyYW5kb21fd2l0aG91dF9yZXBsYWNlbWVudJSFlFKULg=="
151
+ },
152
+ "visualize": false,
153
+ "visualize_aabb": false,
154
+ "visualize_attention_weights": false,
155
+ "visualize_backprojected_features": false,
156
+ "visualize_encoded_features": true,
157
+ "visualizer_background_rgb": {
158
+ "_type": "tuple",
159
+ "_value": [
160
+ 0.0,
161
+ 0.0,
162
+ 0.0
163
+ ]
164
+ },
165
+ "visualizer_min_attention_weight": 0.0,
166
+ "visualizer_min_tsdf_weight": 0.0,
167
+ "visualizer_point_size": 3,
168
+ "visualizer_pointcloud_max_distance": null,
169
+ "visualizer_pointclouds_ply_output_dir": null,
170
+ "visualizer_record_camera_output_path": null,
171
+ "visualizer_voxel_size_m": 0.01,
172
+ "viz_freq": 200,
173
+ "voxel_size_m": null,
174
+ "wandb_entity": "nv-welcome",
175
+ "wandb_mode": "online",
176
+ "wandb_name": null
177
+ }