Upload folder using huggingface_hub

README.md

@@ -4,4 +4,4 @@ datasets:
 - nvidia/PhysicalAI-Robotics-mindmap-Franka-Mug-in-Drawer
 - nvidia/PhysicalAI-Robotics-mindmap-GR1-Drill-in-Box
 - nvidia/PhysicalAI-Robotics-mindmap-GR1-Stick-in-Bin
----
+---

cube_stacking_checkpoint/best.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:10645d835ab47b05dff11b9ad3f4ced100590c343892f9aecdd276bbf7087afe
+oid sha256:201423e49acfe2f4bd60ce8e44e7ef0f1490b93585448b7ba5ab89294345ff2a
 size 27422483

cube_stacking_checkpoint/last.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f09e2a7b864c90f6b4e8eb26af5e8ed21b0da1fe9d0b73211c961ea50552ff66
+oid sha256:5a3dbf17f620045add842d112d328d8d133eec6232e8416c20f918d18c6f6455
 size 27422483

cube_stacking_checkpoint/training_args.json

@@ -4,22 +4,22 @@
     "apply_geometry_noise": 0,
     "apply_random_transforms": 0,
     "base_log_dir": {
-        "_string": "/eval/train_logs",
+        "_string": "/osmo/data/output/train_logs",
         "_type": "python_object (type = PosixPath)",
-        "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
+        "_value": "gASVRgAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5QojAEvlIwEb3Ntb5SMBGRhdGGUjAZvdXRwdXSUjAp0cmFpbl9sb2dzlHSUUpQu"
     },
-    "batch_size": 16,
-    "batch_size_val": 16,
+    "batch_size": 32,
+    "batch_size_val": 32,
     "checkpoint": null,
     "data_type": {
         "_string": "DataType.RGBD_AND_MESH",
         "_type": "python_object (type = DataType)",
         "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
     },
-    "dataset": "/datasets/mindmap_cubes_rgbdmesh_radioB_1cam/",
-    "demos_datagen": "0",
-    "demos_train": "0-7",
-    "demos_valset": "8-9",
+    "dataset": "/osmo/data/input/1",
+    "demos_datagen": "0-149",
+    "demos_train": "0-129",
+    "demos_valset": "130-149",
     "diffusion_dropout": 0.0,
     "diffusion_timesteps": 100,
     "disable_visualizer_wait_on_key": false,
@@ -63,20 +63,20 @@
     "learning_rate_end_factor": 0.5,
     "max_episode_length": 5,
     "max_episodes_per_task": 100,
-    "num_batches_per_test_eval": 1,
-    "num_batches_per_train_eval": 1,
+    "num_batches_per_test_eval": -1,
+    "num_batches_per_train_eval": 10,
     "num_history": 3,
     "num_vertices_to_sample": 2048,
     "num_vis_ins_attn_layers": 2,
-    "num_workers": 0,
-    "num_workers_for_test_dataset": null,
+    "num_workers": 20,
+    "num_workers_for_test_dataset": 0,
     "only_sample_keyposes": false,
     "pos_loss": 30.0,
     "pos_noise_stddev_m": 0.01,
     "prediction_horizon": 1,
     "predictor_dropout": 0.0,
     "print_progress_freq": 100,
-    "print_timers_freq": 1000,
+    "print_timers_freq": 100,
     "projective_appearance_integrator_measurement_weight": null,
     "quaternion_format": "wxyz",
     "random_rpy_range_deg": {
@@ -111,11 +111,8 @@
     },
     "relative_action": 0,
     "reproducibility": {
-        "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/",
-        "git_has_uncommitted_changes": true,
-        "git_root": "/workspaces/mindmap",
-        "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
-        "time": "Tue Sep  9 18:59:03 2025"
+        "command_line": "python run_training.py --dataset /osmo/data/input/1 --base_log_dir /osmo/data/output/train_logs --batch_size 32 --batch_size_val 32 --demos_datagen 0-149 --demos_train 0-129 --demos_valset 130-149 --num_workers 20 --num_workers_for_test_dataset 0 --print_timers_freq 100 --task cube_stacking --train_iters 150000 --val_freq 2500 --wandb_name Release_e2e_cubes_rgbdmesh_radioB_1cam",
+        "time": "Fri Sep 12 10:37:59 2025"
     },
     "rgbd_min_depth_threshold": 0.0,
     "rot_loss": 10.0,
@@ -131,12 +128,12 @@
         "_type": "python_object (type = Tasks)",
         "_value": "gASVNQAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwNY3ViZV9zdGFja2luZ5SFlFKULg=="
     },
-    "train_iters": 24000,
+    "train_iters": 150000,
     "use_fps": 1,
     "use_instruction": 0,
     "use_keyposes": 1,
     "use_shared_feature_encoder": 0,
-    "val_freq": 12,
+    "val_freq": 2500,
     "validate_demos_with_gt_poses": 1,
     "variations": {
         "_type": "tuple",
@@ -173,5 +170,5 @@
     "voxel_size_m": null,
     "wandb_entity": "nv-welcome",
     "wandb_mode": "online",
-    "wandb_name": null
+    "wandb_name": "Release_e2e_cubes_rgbdmesh_radioB_1cam"
 }

drill_in_box_checkpoint/best.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f4527bb0cc33c63f7399deddc506fa69fd08984b3388d90b39026f9dc2c29730
+oid sha256:5adde85fa3c26f214f2f3ff918f1fd77ca507773f61516909f2c606d10dd3dba
 size 28336963

drill_in_box_checkpoint/last.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:266a80a8d232ef7aa5fc32f76bd80bdec86dc62c7633f186b26282733ff416d8
+oid sha256:34bd8d4c67ceb77fb2c01ed39313af0e9810e0a4e8a89b6e50aaf4dc6ca36ff5
 size 28336963

drill_in_box_checkpoint/training_args.json

@@ -4,22 +4,22 @@
     "apply_geometry_noise": 0,
     "apply_random_transforms": 0,
     "base_log_dir": {
-        "_string": "/eval/train_logs",
+        "_string": "/osmo/data/output/train_logs",
         "_type": "python_object (type = PosixPath)",
-        "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
+        "_value": "gASVRgAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5QojAEvlIwEb3Ntb5SMBGRhdGGUjAZvdXRwdXSUjAp0cmFpbl9sb2dzlHSUUpQu"
     },
-    "batch_size": 16,
-    "batch_size_val": 16,
+    "batch_size": 32,
+    "batch_size_val": 32,
     "checkpoint": null,
     "data_type": {
         "_string": "DataType.RGBD_AND_MESH",
         "_type": "python_object (type = DataType)",
         "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
     },
-    "dataset": "/datasets/mindmap_galileo_gr1_rgbdmesh_radioB_1cam/",
-    "demos_datagen": "0",
-    "demos_train": "0-7",
-    "demos_valset": "8-9",
+    "dataset": "/osmo/data/input/1",
+    "demos_datagen": "0-199",
+    "demos_train": "0-99",
+    "demos_valset": "100-119",
     "diffusion_dropout": 0.0,
     "diffusion_timesteps": 100,
     "disable_visualizer_wait_on_key": false,
@@ -63,20 +63,20 @@
     "learning_rate_end_factor": 0.5,
     "max_episode_length": 5,
     "max_episodes_per_task": 100,
-    "num_batches_per_test_eval": 1,
-    "num_batches_per_train_eval": 1,
+    "num_batches_per_test_eval": -1,
+    "num_batches_per_train_eval": 10,
     "num_history": 3,
     "num_vertices_to_sample": 2048,
     "num_vis_ins_attn_layers": 2,
-    "num_workers": 0,
-    "num_workers_for_test_dataset": null,
+    "num_workers": 20,
+    "num_workers_for_test_dataset": 0,
     "only_sample_keyposes": false,
     "pos_loss": 30.0,
     "pos_noise_stddev_m": 0.01,
     "prediction_horizon": 1,
     "predictor_dropout": 0.0,
     "print_progress_freq": 100,
-    "print_timers_freq": 1000,
+    "print_timers_freq": 100,
     "projective_appearance_integrator_measurement_weight": null,
     "quaternion_format": "wxyz",
     "random_rpy_range_deg": {
@@ -111,11 +111,8 @@
     },
     "relative_action": 0,
     "reproducibility": {
-        "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/",
-        "git_has_uncommitted_changes": true,
-        "git_root": "/workspaces/mindmap",
-        "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
-        "time": "Tue Sep  9 19:03:07 2025"
+        "command_line": "python run_training.py --dataset /osmo/data/input/1 --base_log_dir /osmo/data/output/train_logs --batch_size 32 --batch_size_val 32 --demos_datagen 0-199 --demos_train 0-99 --demos_valset 100-119 --num_workers 20 --num_workers_for_test_dataset 0 --print_timers_freq 100 --task drill_in_box --train_iters 150000 --val_freq 2500 --wandb_name Release_e2e_drill_rgbdmesh_radioB_1cam",
+        "time": "Sat Sep 13 01:23:29 2025"
     },
     "rgbd_min_depth_threshold": 0.0,
     "rot_loss": 10.0,
@@ -131,12 +128,12 @@
         "_type": "python_object (type = Tasks)",
         "_value": "gASVNAAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwMZHJpbGxfaW5fYm94lIWUUpQu"
     },
-    "train_iters": 24000,
+    "train_iters": 150000,
     "use_fps": 1,
     "use_instruction": 0,
     "use_keyposes": 1,
     "use_shared_feature_encoder": 0,
-    "val_freq": 12,
+    "val_freq": 2500,
     "validate_demos_with_gt_poses": 1,
     "variations": {
         "_type": "tuple",
@@ -173,5 +170,5 @@
     "voxel_size_m": null,
     "wandb_entity": "nv-welcome",
     "wandb_mode": "online",
-    "wandb_name": null
+    "wandb_name": "Release_e2e_drill_rgbdmesh_radioB_1cam"
 }

mug_in_drawer_checkpoint/best.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d70542250f60764a4736778de6007282c74f5756a535a5ec84794e6aa9ad0df7
+oid sha256:7c9cf214c851e70b292831ab1cab29b9575870735fa8fa4affbea9997e2c271e
 size 27422483

mug_in_drawer_checkpoint/last.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:03ec63d77e46cdc7e807be5a198716232f95a12d37701206bd45effe26fdcbca
+oid sha256:4b3b2be9c119c9a855096ea1c0914bb8a4b5bb99f2d77a5b2dc8bd784c21c89a
 size 27422483

mug_in_drawer_checkpoint/training_args.json

@@ -4,22 +4,22 @@
     "apply_geometry_noise": 0,
     "apply_random_transforms": 0,
     "base_log_dir": {
-        "_string": "/eval/train_logs",
+        "_string": "/osmo/data/output/train_logs",
         "_type": "python_object (type = PosixPath)",
-        "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
+        "_value": "gASVRgAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5QojAEvlIwEb3Ntb5SMBGRhdGGUjAZvdXRwdXSUjAp0cmFpbl9sb2dzlHSUUpQu"
     },
-    "batch_size": 16,
-    "batch_size_val": 16,
+    "batch_size": 32,
+    "batch_size_val": 32,
     "checkpoint": null,
     "data_type": {
         "_string": "DataType.RGBD_AND_MESH",
         "_type": "python_object (type = DataType)",
         "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
     },
-    "dataset": "/datasets/mindmap_mug_rgbdmesh_radioB_1cam/",
-    "demos_datagen": "0",
-    "demos_train": "0-7",
-    "demos_valset": "8-9",
+    "dataset": "/osmo/data/input/1",
+    "demos_datagen": "0-149",
+    "demos_train": "0-129",
+    "demos_valset": "130-149",
     "diffusion_dropout": 0.0,
     "diffusion_timesteps": 100,
     "disable_visualizer_wait_on_key": false,
@@ -63,20 +63,20 @@
     "learning_rate_end_factor": 0.5,
     "max_episode_length": 5,
     "max_episodes_per_task": 100,
-    "num_batches_per_test_eval": 1,
-    "num_batches_per_train_eval": 1,
+    "num_batches_per_test_eval": -1,
+    "num_batches_per_train_eval": 10,
     "num_history": 3,
     "num_vertices_to_sample": 2048,
     "num_vis_ins_attn_layers": 2,
-    "num_workers": 0,
-    "num_workers_for_test_dataset": null,
+    "num_workers": 20,
+    "num_workers_for_test_dataset": 0,
     "only_sample_keyposes": false,
     "pos_loss": 30.0,
     "pos_noise_stddev_m": 0.01,
     "prediction_horizon": 1,
     "predictor_dropout": 0.0,
     "print_progress_freq": 100,
-    "print_timers_freq": 1000,
+    "print_timers_freq": 100,
     "projective_appearance_integrator_measurement_weight": null,
     "quaternion_format": "wxyz",
     "random_rpy_range_deg": {
@@ -111,11 +111,8 @@
     },
     "relative_action": 0,
     "reproducibility": {
-        "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/",
-        "git_has_uncommitted_changes": true,
-        "git_root": "/workspaces/mindmap",
-        "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
-        "time": "Tue Sep  9 19:01:07 2025"
+        "command_line": "python run_training.py --dataset /osmo/data/input/1 --base_log_dir /osmo/data/output/train_logs --batch_size 32 --batch_size_val 32 --demos_datagen 0-149 --demos_train 0-129 --demos_valset 130-149 --num_workers 20 --num_workers_for_test_dataset 0 --print_timers_freq 100 --task mug_in_drawer --train_iters 150000 --val_freq 2500 --wandb_name Release_e2e_mug_rgbdmesh_radioB_1cam",
+        "time": "Sat Sep 13 15:41:22 2025"
     },
     "rgbd_min_depth_threshold": 0.0,
     "rot_loss": 10.0,
@@ -131,12 +128,12 @@
         "_type": "python_object (type = Tasks)",
         "_value": "gASVNQAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwNbXVnX2luX2RyYXdlcpSFlFKULg=="
     },
-    "train_iters": 24000,
+    "train_iters": 150000,
     "use_fps": 1,
     "use_instruction": 0,
     "use_keyposes": 1,
     "use_shared_feature_encoder": 0,
-    "val_freq": 12,
+    "val_freq": 2500,
     "validate_demos_with_gt_poses": 1,
     "variations": {
         "_type": "tuple",
@@ -173,5 +170,5 @@
     "voxel_size_m": null,
     "wandb_entity": "nv-welcome",
     "wandb_mode": "online",
-    "wandb_name": null
+    "wandb_name": "Release_e2e_mug_rgbdmesh_radioB_1cam"
 }

stick_in_bin_checkpoint/best.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dc45a68b3c437dac08b530e65fa138b5522fbe4f1460b1851abe4ae53b13a45f
+oid sha256:6eaa897b795668da0489318d9153a3173acaeef0462141ed9e978fd134ab61c9
 size 28336963

stick_in_bin_checkpoint/last.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:809947b0da2c4486ba8f016d038ac0e6c72056dc2a049c01257be5cc6494c697
+oid sha256:233e08861c09a9d174507ac5befb1c42504df6ed988c8c90724b9d7bf10208ca
 size 28336963

stick_in_bin_checkpoint/training_args.json

@@ -4,22 +4,22 @@
     "apply_geometry_noise": 0,
     "apply_random_transforms": 0,
     "base_log_dir": {
-        "_string": "/eval/train_logs",
+        "_string": "/osmo/data/output/train_logs",
         "_type": "python_object (type = PosixPath)",
-        "_value": "gASVNQAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5SMAS+UjARldmFslIwKdHJhaW5fbG9nc5SHlFKULg=="
+        "_value": "gASVRgAAAAAAAACMB3BhdGhsaWKUjAlQb3NpeFBhdGiUk5QojAEvlIwEb3Ntb5SMBGRhdGGUjAZvdXRwdXSUjAp0cmFpbl9sb2dzlHSUUpQu"
     },
-    "batch_size": 16,
-    "batch_size_val": 16,
+    "batch_size": 32,
+    "batch_size_val": 32,
     "checkpoint": null,
     "data_type": {
         "_string": "DataType.RGBD_AND_MESH",
         "_type": "python_object (type = DataType)",
         "_value": "gASVRAAAAAAAAACMH21pbmRtYXAuZGF0YV9sb2FkaW5nLmRhdGFfdHlwZXOUjAhEYXRhVHlwZZSTlIwNcmdiZF9hbmRfbWVzaJSFlFKULg=="
     },
-    "dataset": "/datasets/mindmap_galileo_gr1_rgbdmesh_radioB_1cam/",
-    "demos_datagen": "0",
-    "demos_train": "0-6",
-    "demos_valset": "0-6",
+    "dataset": "/osmo/data/input/1",
+    "demos_datagen": "0-199",
+    "demos_train": "0-99",
+    "demos_valset": "100-119",
     "diffusion_dropout": 0.0,
     "diffusion_timesteps": 100,
     "disable_visualizer_wait_on_key": false,
@@ -63,20 +63,20 @@
     "learning_rate_end_factor": 0.5,
     "max_episode_length": 5,
     "max_episodes_per_task": 100,
-    "num_batches_per_test_eval": 1,
-    "num_batches_per_train_eval": 1,
+    "num_batches_per_test_eval": -1,
+    "num_batches_per_train_eval": 10,
     "num_history": 3,
     "num_vertices_to_sample": 2048,
     "num_vis_ins_attn_layers": 2,
-    "num_workers": 0,
-    "num_workers_for_test_dataset": null,
+    "num_workers": 20,
+    "num_workers_for_test_dataset": 0,
     "only_sample_keyposes": false,
     "pos_loss": 30.0,
     "pos_noise_stddev_m": 0.01,
     "prediction_horizon": 1,
     "predictor_dropout": 0.0,
     "print_progress_freq": 100,
-    "print_timers_freq": 1000,
+    "print_timers_freq": 100,
     "projective_appearance_integrator_measurement_weight": null,
     "quaternion_format": "wxyz",
     "random_rpy_range_deg": {
@@ -111,11 +111,8 @@
     },
     "relative_action": 0,
     "reproducibility": {
-        "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/",
-        "git_has_uncommitted_changes": true,
-        "git_root": "/workspaces/mindmap",
-        "git_url": "https://gitlab-master.nvidia.com/12051/nvblox/mindmap/tree/12fa2bdc1462db417618cd17a0a08a4a57a883cb",
-        "time": "Tue Sep  9 19:09:08 2025"
+        "command_line": "python run_training.py --dataset /osmo/data/input/1 --base_log_dir /osmo/data/output/train_logs --batch_size 32 --batch_size_val 32 --demos_datagen 0-199 --demos_train 0-99 --demos_valset 100-119 --num_workers 20 --num_workers_for_test_dataset 0 --print_timers_freq 100 --task stick_in_bin --train_iters 150000 --val_freq 2500 --wandb_name Release_e2e_stick_rgbdmesh_radioB_1cam",
+        "time": "Sat Sep 13 12:41:28 2025"
     },
     "rgbd_min_depth_threshold": 0.0,
     "rot_loss": 10.0,
@@ -131,12 +128,12 @@
         "_type": "python_object (type = Tasks)",
         "_value": "gASVNAAAAAAAAACME21pbmRtYXAudGFza3MudGFza3OUjAVUYXNrc5STlIwMc3RpY2tfaW5fYmlulIWUUpQu"
     },
-    "train_iters": 24000,
+    "train_iters": 150000,
     "use_fps": 1,
     "use_instruction": 0,
     "use_keyposes": 1,
     "use_shared_feature_encoder": 0,
-    "val_freq": 12,
+    "val_freq": 2500,
     "validate_demos_with_gt_poses": 1,
     "variations": {
         "_type": "tuple",
@@ -173,5 +170,5 @@
     "voxel_size_m": null,
     "wandb_entity": "nv-welcome",
     "wandb_mode": "online",
-    "wandb_name": null
+    "wandb_name": "Release_e2e_stick_rgbdmesh_radioB_1cam"
 }