Add raw checkpoints for exact selected inference reproduction

README.md

@@ -19,8 +19,11 @@ Diffusion Policy for Coverage Path Planning**.
 
 ## Released models
 
-Each directory contains the EMA weights selected by the original seed-42
-training run, the exact inference configuration, and provenance metadata.
+Each root category directory contains the EMA weights selected by the original
+seed-42 training run. `raw/<category>/` contains a tensor-only exact export of
+the raw `state_dicts.model` entry loaded by the archived selected-visualization
+test script. Every directory includes its inference configuration and
+provenance metadata.
 
 | Directory | Dataset | Training run | Top-k selection PCD ↓ |
 |:--|:--|:--|--:|
@@ -29,6 +32,13 @@ training run, the exact inference configuration, and provenance metadata.
 | `shelves/` | `shelves-v2` | `52VCU-S42` | 10.069027 |
 | `containers/` | `containers-v2` | `ODAV4-S42` | 347.932620 |
 
+| Selected inference case | Raw tensor path | Test split item |
+|:--|:--|:--|
+| Windows | `raw/windows/model.safetensors` | index 5, `810_wr1fr_1` |
+| Cuboids | `raw/cuboids/model.safetensors` | index 3, `669_cube_1001_1285_1263` |
+| Shelves | `raw/shelves/model.safetensors` | index 4, `box_h620_w500_d220.0_sh1.0_sv2.0` |
+| Containers | `raw/containers/model.safetensors` | index 1, `spoegcr3gv` |
+
 The values above are the per-run validation monitor used for top-k checkpoint
 selection. They are not the three-seed test results reported in the paper.
 Containers is a separate low-data experiment; `ODAV4-S42` is released because
@@ -37,43 +47,61 @@ evaluation scripts.
 
 ## Download and evaluate
 
-Install the Hugging Face CLI and download one category:
+From the code repository, download one category plus release manifests:
 
 ```bash
-hf download ChenyuanC/3D-CovDiffusion \
-  --include "windows/*" \
-  --local-dir checkpoints/3d-covdiffusion
+python reproduce.py download --category windows
+python reproduce.py smoke --category windows --device cuda
 ```
 
-With the public code and evaluation data configured:
+With the separate raw evaluation meshes, trajectories, and fixed splits:
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python evaluate.py \
-  --checkpoint_path checkpoints/3d-covdiffusion/windows/model.safetensors \
-  --config_path checkpoints/3d-covdiffusion/windows/config.yaml \
-  --dataset_name windows-v2 \
-  --dataset_split test \
-  --eval_episodes 20 \
-  --workers 4
+python reproduce.py evaluate \
+  --category windows \
+  --eval-root /absolute/path/to/evaluation-data \
+  --episodes 0
 ```
 
-Replace `windows` / `windows-v2` with `cuboids`, `shelves`, or `containers`.
-The adjacent `config.yaml` is discovered automatically, so `--config_path` can
-be omitted when the original directory layout is preserved.
+Replace `windows` with `cuboids`, `shelves`, or `containers`. Numeric evaluation
+is metrics-only by default. The train-ready dataset is train-only and cannot be
+used as the evaluation root.
+
+To reproduce one selected project-page visualization with the exact raw tensor
+variant, use the tagged `inference-v1` code release. It replays the archived
+`GT_Cond -> Pred_Cond` RNG order and reports the prediction-conditioned result:
+
+```bash
+python reproduce.py download \
+  --category windows \
+  --weight-variant raw \
+  --models-only
+
+python reproduce.py inference \
+  --category windows \
+  --eval-root /absolute/path/to/raw-evaluation-data \
+  --save-artifacts
+```
+
+The full checkpoint/config/test-index matrix and SHA-256 regression hashes are
+in the code repository's `docs/INFERENCE.md` and
+`configs/inference/seed42_selected_episodes.json`.
 
 ## Format and integrity
 
-- `model.safetensors` contains the complete EMA policy state, including the
-  action and point-cloud normalizer tensors.
-- Optimizer state, raw non-EMA weights, Python/Dill pickles, W&B metadata, and
+- Root `<category>/model.safetensors` files contain complete EMA policy states.
+- `raw/<category>/model.safetensors` files contain the complete raw policy
+  states used by the archived selected-visualization tests.
+- Both variants include the action and point-cloud normalizer tensors.
+- Optimizer state, full Python/Dill training checkpoints, W&B metadata, and
   machine-local paths are intentionally excluded.
 - `metrics.json` records the source run, epoch, step, selection metric, source
   checkpoint digest, release digest, and exact tensor-roundtrip validation.
 - `manifest.json` and `SHA256SUMS` provide repository-wide integrity metadata.
 
-The original training checkpoints must be treated as trusted pickle files. The
-files in this repository use the tensor-only safetensors format and are the
-recommended artifacts for inference.
+The original training checkpoints must be treated as trusted pickle files. All
+files in this repository use tensor-only safetensors; users never need to load
+the source `.ckpt` files.
 
 ## Intended use and limitations
 
@@ -84,6 +112,10 @@ certified motion-planning or robot-safety system. Validate collision handling,
 kinematic feasibility, workcell constraints, and emergency behavior before any
 physical deployment.
 
+Only the selected seed-42 checkpoint is currently released for each category.
+The paper's three-seed mean and standard deviation cannot be regenerated until
+the other checkpoints or their per-seed result JSON files are published.
+
 ## License
 
 No standalone repository or model-weight license has been selected yet.

SHA256SUMS

@@ -1,11 +1,11 @@
-829a9c0e7c0e93ad101ab7bf8a5d6a587e80d4799384b02cdde62d52fa83b948  README.md
+de65562f7a1fdad52dd85ef2ac129b331c6da107db3f1b28b6c52bd9832b057c  README.md
 921479d3af21384ecc0362c9516b019df209f593a1b3ceed985f4d85482e42cc  containers/config.yaml
 7945015f5b38b13551ec707cfa36b39aa1b62233afa9265471c9c864362f04fb  containers/metrics.json
 19559bfecdba3375d8ebe6fc043a53c10297cedfa226023b484bc2fad8f30102  containers/model.safetensors
 a405bcbaa8bd80a73353d786e37972ff93f91d5e075aa171d443adec45e3cc4d  cuboids/config.yaml
 7587101b368c5c0de556aa93c1a2e36767428bd073e0f21128733ead1c6496c1  cuboids/metrics.json
 0116ddc97f31055d8becf3b6ce89d00d66bbfa6b27bf3cdb8ca8fcd525f0c76d  cuboids/model.safetensors
-c19e9403f130432d1599cfc98b3eeb2a4a819fabf9e46536a0ad34c7700dc935  manifest.json
+41d5349254b46a4a28cd64123a93fbe37e4b912cb1f9c3ffc6d2977b4d4ad346  manifest.json
 a33f9643c53f85d7d6f55a105cf19e6844d194e3bd6535c36701a37c5b002395  shelves/config.yaml
 c2473be1acf7b900db676b87fdbe73688b864523451b1ac246ee5d1a1794dcc6  shelves/metrics.json
 aa84b6c73a37bb3906df9b60cf69c4942999f0520ef85977e91b1120338e3256  shelves/model.safetensors

manifest.json

@@ -127,8 +127,8 @@
   },
   "files": {
     "README.md": {
-      "bytes": 3741,
-      "sha256": "829a9c0e7c0e93ad101ab7bf8a5d6a587e80d4799384b02cdde62d52fa83b948"
+      "bytes": 5200,
+      "sha256": "de65562f7a1fdad52dd85ef2ac129b331c6da107db3f1b28b6c52bd9832b057c"
     },
     "containers/config.yaml": {
       "bytes": 1192,

raw/README.md

@@ -0,0 +1,135 @@
+---
+library_name: pytorch
+tags:
+  - robotics
+  - diffusion-policy
+  - 3d-point-cloud
+  - coverage-path-planning
+  - safetensors
+---
+
+# 3D-CovDiffusion checkpoints
+
+Category-specific pretrained policies for **3D-CovDiffusion: 3D-Aware
+Diffusion Policy for Coverage Path Planning**.
+
+- Project page: https://crystalccy1.github.io/3D-CovDiffusion/
+- Code: https://github.com/crystalccy1/3D-CovDiffusion
+- Train-ready dataset: https://huggingface.co/datasets/ChenyuanC/3D-CovDiffusion-Train-Ready
+
+## Released models
+
+Each root category directory contains the EMA weights selected by the original
+seed-42 training run. `raw/<category>/` contains a tensor-only exact export of
+the raw `state_dicts.model` entry loaded by the archived selected-visualization
+test script. Every directory includes its inference configuration and
+provenance metadata.
+
+| Directory | Dataset | Training run | Top-k selection PCD ↓ |
+|:--|:--|:--|--:|
+| `windows/` | `windows-v2` | `TML4Q-S42` | 10.410878 |
+| `cuboids/` | `cuboids-v2` | `X1PD1-S42` | 6.612324 |
+| `shelves/` | `shelves-v2` | `52VCU-S42` | 10.069027 |
+| `containers/` | `containers-v2` | `ODAV4-S42` | 347.932620 |
+
+| Selected inference case | Raw tensor path | Test split item |
+|:--|:--|:--|
+| Windows | `raw/windows/model.safetensors` | index 5, `810_wr1fr_1` |
+| Cuboids | `raw/cuboids/model.safetensors` | index 3, `669_cube_1001_1285_1263` |
+| Shelves | `raw/shelves/model.safetensors` | index 4, `box_h620_w500_d220.0_sh1.0_sv2.0` |
+| Containers | `raw/containers/model.safetensors` | index 1, `spoegcr3gv` |
+
+The values above are the per-run validation monitor used for top-k checkpoint
+selection. They are not the three-seed test results reported in the paper.
+Containers is a separate low-data experiment; `ODAV4-S42` is released because
+it is the checkpoint referenced by the original in-domain, OOD, and video
+evaluation scripts.
+
+## Download and evaluate
+
+From the code repository, download one category plus release manifests:
+
+```bash
+python reproduce.py download --category windows
+python reproduce.py smoke --category windows --device cuda
+```
+
+With the separate raw evaluation meshes, trajectories, and fixed splits:
+
+```bash
+python reproduce.py evaluate \
+  --category windows \
+  --eval-root /absolute/path/to/evaluation-data \
+  --episodes 0
+```
+
+Replace `windows` with `cuboids`, `shelves`, or `containers`. Numeric evaluation
+is metrics-only by default. The train-ready dataset is train-only and cannot be
+used as the evaluation root.
+
+To reproduce one selected project-page visualization with the exact raw tensor
+variant, use the tagged `inference-v1` code release. It replays the archived
+`GT_Cond -> Pred_Cond` RNG order and reports the prediction-conditioned result:
+
+```bash
+python reproduce.py download \
+  --category windows \
+  --weight-variant raw \
+  --models-only
+
+python reproduce.py inference \
+  --category windows \
+  --eval-root /absolute/path/to/raw-evaluation-data \
+  --save-artifacts
+```
+
+The full checkpoint/config/test-index matrix and SHA-256 regression hashes are
+in the code repository's `docs/INFERENCE.md` and
+`configs/inference/seed42_selected_episodes.json`.
+
+## Format and integrity
+
+- Root `<category>/model.safetensors` files contain complete EMA policy states.
+- `raw/<category>/model.safetensors` files contain the complete raw policy
+  states used by the archived selected-visualization tests.
+- Both variants include the action and point-cloud normalizer tensors.
+- Optimizer state, full Python/Dill training checkpoints, W&B metadata, and
+  machine-local paths are intentionally excluded.
+- `metrics.json` records the source run, epoch, step, selection metric, source
+  checkpoint digest, release digest, and exact tensor-roundtrip validation.
+- `manifest.json` and `SHA256SUMS` provide repository-wide integrity metadata.
+
+The original training checkpoints must be treated as trusted pickle files. All
+files in this repository use tensor-only safetensors; users never need to load
+the source `.ckpt` files.
+
+## Intended use and limitations
+
+These checkpoints generate ordered 6-DoF coverage-trajectory chunks from a
+5,120-point observation and recent execution history. They are research
+artifacts evaluated on the corresponding geometry categories; they are not a
+certified motion-planning or robot-safety system. Validate collision handling,
+kinematic feasibility, workcell constraints, and emergency behavior before any
+physical deployment.
+
+Only the selected seed-42 checkpoint is currently released for each category.
+The paper's three-seed mean and standard deviation cannot be regenerated until
+the other checkpoints or their per-seed result JSON files are published.
+
+## License
+
+No standalone repository or model-weight license has been selected yet.
+Third-party components remain subject to their original terms; see the notices
+in the code repository.
+
+## Citation
+
+```bibtex
+@misc{chen2026_3dcovdiffusion,
+  title  = {{3D-CovDiffusion}: 3D-Aware Diffusion Policy for Coverage Path Planning},
+  author = {Chen, Chenyuan and Ding, Haoran and Ding, Ran and Liu, Tianyu
+            and He, Zewen and Duan, Anqing and Nakamura, Yoshihiko},
+  year   = {2026},
+  note   = {Manuscript}
+}
+```

raw/SHA256SUMS

@@ -0,0 +1,14 @@
+de65562f7a1fdad52dd85ef2ac129b331c6da107db3f1b28b6c52bd9832b057c  README.md
+b379357ea61882e4e0f47e3ba7f2c6a06e46d2b4f5141d867af3b99fa60f43c8  containers/config.yaml
+9346ac77e468dcbfeccdf544ec2375be8c0b8d499ad23fab19a6c3606cb7915d  containers/metrics.json
+2d95ed639db8789456fd6b4bfb203a66d2f97ffac06c6ac4f616967bb6675d78  containers/model.safetensors
+fd34ec223d2133779c75d30b7280af6e1e469d3c35bd72995f768415dbf6bbd3  cuboids/config.yaml
+4ec352f3ae9d52e73d6e06be73ab71eec55e441331ca5de784c7a76a42b5af5b  cuboids/metrics.json
+0fc51c588b79e6952758535d5b4b5ca8baed7b08ca76b185ef9776e768ce9110  cuboids/model.safetensors
+f30a27989b859b61b3f19c9d8e5c22f3966a72d1b36faa9aa637f67195c98c66  manifest.json
+814000c0e59f859343a35cfccd33de218799c5c1d87f285c0c5bc5f65972861f  shelves/config.yaml
+07144bb25cc4061c1878a2a4eee1180ec80de7273e0dd5571c5c7ab8d70dd99c  shelves/metrics.json
+2051bd6fc6e2bc297f54f5ef97f9dd7d3f0515fa595aab78cd61d5a15d169380  shelves/model.safetensors
+f438f51db30e0dbc91eef89723cc9a36820e46fd29244d1d4d6711187f0eff5d  windows/config.yaml
+449fb2f2eb63bfd61d1524962ebb3288e8576e0be1463178c7e515a9a244f119  windows/metrics.json
+a0c381fbffbe2e12c83a1a40bc8a2591717f12579561a21e146ce4b1b18156ae  windows/model.safetensors

raw/containers/config.yaml

@@ -0,0 +1,76 @@
+format_version: 3dcov-inference-config-v2
+task_name: CovDiffusion
+io_type: CovDiffusion
+dataset:
+- containers-v2
+action_dim: 24
+horizon: 16
+n_action_steps: 100
+n_obs_steps: 1
+encoder_output_dim: 256
+shape_meta:
+  obs:
+    point_cloud:
+      shape:
+      - 5120
+      - 3
+    low_dim:
+      shape:
+      - 24
+  action:
+    shape:
+    - 24
+diffusion:
+  model_type: dp3
+  diffusion_step_embed_dim: 128
+  down_dims:
+  - 512
+  - 1024
+  - 2048
+  kernel_size: 5
+  n_groups: 8
+  condition_type: film
+  use_down_condition: true
+  use_mid_condition: true
+  use_up_condition: true
+  num_inference_steps: 10
+  obs_as_global_cond: true
+model:
+  backbone: dp3
+  affinetrans: false
+  hidden_size:
+  - 1024
+  - 1024
+  pretrained: false
+  pretrained_custom: null
+  load_strict: false
+noise_scheduler:
+  _target_: diffusers.schedulers.scheduling_ddim.DDIMScheduler
+  num_train_timesteps: 100
+  beta_start: 0.0001
+  beta_end: 0.02
+  beta_schedule: squaredcos_cap_v2
+  clip_sample: true
+  set_alpha_to_one: true
+  steps_offset: 0
+  prediction_type: sample
+partial_observation_enabled: false
+partial_observation_method: fixed_camera
+partial_observation_ratio: 0.3
+ablation_prev_traj: normal
+random_traj_seed: 123
+pc_points: 5120
+traj_points: 1000
+lambda_points: 4
+overlapping: 1
+asymm_overlapping: false
+normalization: per-dataset
+extra_data:
+- orientnorm
+weight_orient: 0.25
+load_extra_data:
+- stroke_masks
+traj_with_equally_spaced_points: true
+equal_spaced_points_distance: 0.05
+equal_in_3d_space: false
+augmentations: []

raw/containers/metrics.json

@@ -0,0 +1,31 @@
+{
+  "category": "containers",
+  "dataset": "containers-v2",
+  "epoch": 500,
+  "format_version": "3dcov-checkpoint-metadata-v1",
+  "global_step": 141783,
+  "release": {
+    "exact_tensor_roundtrip": true,
+    "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+    "model_sha256": "2d95ed639db8789456fd6b4bfb203a66d2f97ffac06c6ac4f616967bb6675d78",
+    "normalizer_fields": [
+      "action",
+      "point_cloud"
+    ],
+    "tensor_bytes": 1020966632,
+    "tensor_count": 182
+  },
+  "run_id": "ODAV4-S42",
+  "seed": 42,
+  "selection_metric": {
+    "mode": "min",
+    "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+    "scope": "seed-42 validation/top-k selection",
+    "value": 347.93262
+  },
+  "source": {
+    "checkpoint_basename": "epoch=500-step=141783-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=347p9326.ckpt",
+    "sha256": "7c66fbf944afa0182e2863b5a163beb09a5696c29333143cf8f25594c5ffbfb8"
+  },
+  "weight_variant": "raw"
+}

raw/containers/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d95ed639db8789456fd6b4bfb203a66d2f97ffac06c6ac4f616967bb6675d78
+size 1020987472

raw/cuboids/config.yaml

@@ -0,0 +1,68 @@
+format_version: 3dcov-inference-config-v2
+task_name: CovDiffusion
+io_type: CovDiffusion
+dataset:
+- cuboids-v2
+action_dim: 24
+horizon: 16
+n_action_steps: 100
+n_obs_steps: 1
+encoder_output_dim: 256
+diffusion:
+  model_type: dp3
+  diffusion_step_embed_dim: 128
+  down_dims:
+  - 512
+  - 1024
+  - 2048
+  kernel_size: 5
+  n_groups: 8
+  condition_type: film
+  use_down_condition: true
+  use_mid_condition: true
+  use_up_condition: true
+  num_inference_steps: 10
+  obs_as_global_cond: true
+model:
+  backbone: dp3
+  affinetrans: false
+  hidden_size:
+  - 1024
+  - 1024
+  pretrained: false
+  pretrained_custom: null
+  load_strict: false
+noise_scheduler:
+  _target_: diffusers.schedulers.scheduling_ddim.DDIMScheduler
+  num_train_timesteps: 100
+  beta_start: 0.0001
+  beta_end: 0.02
+  beta_schedule: squaredcos_cap_v2
+  clip_sample: true
+  set_alpha_to_one: true
+  steps_offset: 0
+  prediction_type: sample
+pc_points: 5120
+traj_points: 2000
+lambda_points: 4
+overlapping: 1
+asymm_overlapping: false
+normalization: per-dataset
+extra_data:
+- orientnorm
+weight_orient: 0.25
+load_extra_data:
+- stroke_masks
+traj_with_equally_spaced_points: true
+equal_spaced_points_distance: 0.05
+equal_in_3d_space: false
+augmentations: []
+shape_meta:
+  obs:
+    point_cloud:
+      shape:
+      - 5120
+      - 3
+  action:
+    shape:
+    - 24

raw/cuboids/metrics.json

@@ -0,0 +1,31 @@
+{
+  "category": "cuboids",
+  "dataset": "cuboids-v2",
+  "epoch": 40,
+  "format_version": "3dcov-checkpoint-metadata-v1",
+  "global_step": 38007,
+  "release": {
+    "exact_tensor_roundtrip": true,
+    "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+    "model_sha256": "0fc51c588b79e6952758535d5b4b5ca8baed7b08ca76b185ef9776e768ce9110",
+    "normalizer_fields": [
+      "action",
+      "point_cloud"
+    ],
+    "tensor_bytes": 1020966632,
+    "tensor_count": 182
+  },
+  "run_id": "X1PD1-S42",
+  "seed": 42,
+  "selection_metric": {
+    "mode": "min",
+    "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+    "scope": "seed-42 validation/top-k selection",
+    "value": 6.612324
+  },
+  "source": {
+    "checkpoint_basename": "epoch=40-step=38007-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=6p6123.ckpt",
+    "sha256": "9758331dda8cbc0b1f13365b6c0e4e105dd2436ed7d254884bdc5da00ae11dee"
+  },
+  "weight_variant": "raw"
+}

raw/cuboids/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fc51c588b79e6952758535d5b4b5ca8baed7b08ca76b185ef9776e768ce9110
+size 1020987464

raw/manifest.json

@@ -0,0 +1,186 @@
+{
+  "categories": {
+    "containers": {
+      "category": "containers",
+      "dataset": "containers-v2",
+      "epoch": 500,
+      "format_version": "3dcov-checkpoint-metadata-v1",
+      "global_step": 141783,
+      "release": {
+        "exact_tensor_roundtrip": true,
+        "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+        "model_sha256": "2d95ed639db8789456fd6b4bfb203a66d2f97ffac06c6ac4f616967bb6675d78",
+        "normalizer_fields": [
+          "action",
+          "point_cloud"
+        ],
+        "tensor_bytes": 1020966632,
+        "tensor_count": 182
+      },
+      "run_id": "ODAV4-S42",
+      "seed": 42,
+      "selection_metric": {
+        "mode": "min",
+        "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+        "scope": "seed-42 validation/top-k selection",
+        "value": 347.93262
+      },
+      "source": {
+        "checkpoint_basename": "epoch=500-step=141783-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=347p9326.ckpt",
+        "sha256": "7c66fbf944afa0182e2863b5a163beb09a5696c29333143cf8f25594c5ffbfb8"
+      },
+      "weight_variant": "raw"
+    },
+    "cuboids": {
+      "category": "cuboids",
+      "dataset": "cuboids-v2",
+      "epoch": 40,
+      "format_version": "3dcov-checkpoint-metadata-v1",
+      "global_step": 38007,
+      "release": {
+        "exact_tensor_roundtrip": true,
+        "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+        "model_sha256": "0fc51c588b79e6952758535d5b4b5ca8baed7b08ca76b185ef9776e768ce9110",
+        "normalizer_fields": [
+          "action",
+          "point_cloud"
+        ],
+        "tensor_bytes": 1020966632,
+        "tensor_count": 182
+      },
+      "run_id": "X1PD1-S42",
+      "seed": 42,
+      "selection_metric": {
+        "mode": "min",
+        "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+        "scope": "seed-42 validation/top-k selection",
+        "value": 6.612324
+      },
+      "source": {
+        "checkpoint_basename": "epoch=40-step=38007-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=6p6123.ckpt",
+        "sha256": "9758331dda8cbc0b1f13365b6c0e4e105dd2436ed7d254884bdc5da00ae11dee"
+      },
+      "weight_variant": "raw"
+    },
+    "shelves": {
+      "category": "shelves",
+      "dataset": "shelves-v2",
+      "epoch": 45,
+      "format_version": "3dcov-checkpoint-metadata-v1",
+      "global_step": 40204,
+      "release": {
+        "exact_tensor_roundtrip": true,
+        "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+        "model_sha256": "2051bd6fc6e2bc297f54f5ef97f9dd7d3f0515fa595aab78cd61d5a15d169380",
+        "normalizer_fields": [
+          "action",
+          "point_cloud"
+        ],
+        "tensor_bytes": 1020966632,
+        "tensor_count": 182
+      },
+      "run_id": "52VCU-S42",
+      "seed": 42,
+      "selection_metric": {
+        "mode": "min",
+        "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+        "scope": "seed-42 validation/top-k selection",
+        "value": 10.069027
+      },
+      "source": {
+        "checkpoint_basename": "epoch=45-step=40204-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=10p0690.ckpt",
+        "sha256": "7a68066d135ab835314779a789305547ecf18de355113b21f917171e4b8358fb"
+      },
+      "weight_variant": "raw"
+    },
+    "windows": {
+      "category": "windows",
+      "dataset": "windows-v2",
+      "epoch": 165,
+      "format_version": "3dcov-checkpoint-metadata-v1",
+      "global_step": 48472,
+      "release": {
+        "exact_tensor_roundtrip": true,
+        "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+        "model_sha256": "a0c381fbffbe2e12c83a1a40bc8a2591717f12579561a21e146ce4b1b18156ae",
+        "normalizer_fields": [
+          "action",
+          "point_cloud"
+        ],
+        "tensor_bytes": 1020966632,
+        "tensor_count": 182
+      },
+      "run_id": "TML4Q-S42",
+      "seed": 42,
+      "selection_metric": {
+        "mode": "min",
+        "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+        "scope": "seed-42 validation/top-k selection",
+        "value": 10.410878
+      },
+      "source": {
+        "checkpoint_basename": "epoch=165-step=48472-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=10p4109.ckpt",
+        "sha256": "bcb0db0b7a5f0c2a4c7bf0472785b5588c4288d627ec212379a37c40b64f14e3"
+      },
+      "weight_variant": "raw"
+    }
+  },
+  "files": {
+    "README.md": {
+      "bytes": 5200,
+      "sha256": "de65562f7a1fdad52dd85ef2ac129b331c6da107db3f1b28b6c52bd9832b057c"
+    },
+    "containers/config.yaml": {
+      "bytes": 1503,
+      "sha256": "b379357ea61882e4e0f47e3ba7f2c6a06e46d2b4f5141d867af3b99fa60f43c8"
+    },
+    "containers/metrics.json": {
+      "bytes": 963,
+      "sha256": "9346ac77e468dcbfeccdf544ec2375be8c0b8d499ad23fab19a6c3606cb7915d"
+    },
+    "containers/model.safetensors": {
+      "bytes": 1020987472,
+      "sha256": "2d95ed639db8789456fd6b4bfb203a66d2f97ffac06c6ac4f616967bb6675d78"
+    },
+    "cuboids/config.yaml": {
+      "bytes": 1307,
+      "sha256": "fd34ec223d2133779c75d30b7280af6e1e469d3c35bd72995f768415dbf6bbd3"
+    },
+    "cuboids/metrics.json": {
+      "bytes": 950,
+      "sha256": "4ec352f3ae9d52e73d6e06be73ab71eec55e441331ca5de784c7a76a42b5af5b"
+    },
+    "cuboids/model.safetensors": {
+      "bytes": 1020987464,
+      "sha256": "0fc51c588b79e6952758535d5b4b5ca8baed7b08ca76b185ef9776e768ce9110"
+    },
+    "shelves/config.yaml": {
+      "bytes": 1307,
+      "sha256": "814000c0e59f859343a35cfccd33de218799c5c1d87f285c0c5bc5f65972861f"
+    },
+    "shelves/metrics.json": {
+      "bytes": 952,
+      "sha256": "07144bb25cc4061c1878a2a4eee1180ec80de7273e0dd5571c5c7ab8d70dd99c"
+    },
+    "shelves/model.safetensors": {
+      "bytes": 1020987464,
+      "sha256": "2051bd6fc6e2bc297f54f5ef97f9dd7d3f0515fa595aab78cd61d5a15d169380"
+    },
+    "windows/config.yaml": {
+      "bytes": 1498,
+      "sha256": "f438f51db30e0dbc91eef89723cc9a36820e46fd29244d1d4d6711187f0eff5d"
+    },
+    "windows/metrics.json": {
+      "bytes": 954,
+      "sha256": "449fb2f2eb63bfd61d1524962ebb3288e8576e0be1463178c7e515a9a244f119"
+    },
+    "windows/model.safetensors": {
+      "bytes": 1020987464,
+      "sha256": "a0c381fbffbe2e12c83a1a40bc8a2591717f12579561a21e146ce4b1b18156ae"
+    }
+  },
+  "format_version": "3dcov-model-release-v1",
+  "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+  "repository": "ChenyuanC/3D-CovDiffusion",
+  "weight_variant": "raw"
+}

raw/shelves/config.yaml

@@ -0,0 +1,68 @@
+format_version: 3dcov-inference-config-v2
+task_name: CovDiffusion
+io_type: CovDiffusion
+dataset:
+- shelves-v2
+action_dim: 24
+horizon: 16
+n_action_steps: 100
+n_obs_steps: 1
+encoder_output_dim: 256
+diffusion:
+  model_type: dp3
+  diffusion_step_embed_dim: 128
+  down_dims:
+  - 512
+  - 1024
+  - 2048
+  kernel_size: 5
+  n_groups: 8
+  condition_type: film
+  use_down_condition: true
+  use_mid_condition: true
+  use_up_condition: true
+  num_inference_steps: 10
+  obs_as_global_cond: true
+model:
+  backbone: dp3
+  affinetrans: false
+  hidden_size:
+  - 1024
+  - 1024
+  pretrained: false
+  pretrained_custom: null
+  load_strict: false
+noise_scheduler:
+  _target_: diffusers.schedulers.scheduling_ddim.DDIMScheduler
+  num_train_timesteps: 100
+  beta_start: 0.0001
+  beta_end: 0.02
+  beta_schedule: squaredcos_cap_v2
+  clip_sample: true
+  set_alpha_to_one: true
+  steps_offset: 0
+  prediction_type: sample
+pc_points: 5120
+traj_points: 3000
+lambda_points: 4
+overlapping: 1
+asymm_overlapping: false
+normalization: per-dataset
+extra_data:
+- orientnorm
+weight_orient: 0.25
+load_extra_data:
+- stroke_masks
+traj_with_equally_spaced_points: true
+equal_spaced_points_distance: 0.05
+equal_in_3d_space: false
+augmentations: []
+shape_meta:
+  obs:
+    point_cloud:
+      shape:
+      - 5120
+      - 3
+  action:
+    shape:
+    - 24

raw/shelves/metrics.json

@@ -0,0 +1,31 @@
+{
+  "category": "shelves",
+  "dataset": "shelves-v2",
+  "epoch": 45,
+  "format_version": "3dcov-checkpoint-metadata-v1",
+  "global_step": 40204,
+  "release": {
+    "exact_tensor_roundtrip": true,
+    "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+    "model_sha256": "2051bd6fc6e2bc297f54f5ef97f9dd7d3f0515fa595aab78cd61d5a15d169380",
+    "normalizer_fields": [
+      "action",
+      "point_cloud"
+    ],
+    "tensor_bytes": 1020966632,
+    "tensor_count": 182
+  },
+  "run_id": "52VCU-S42",
+  "seed": 42,
+  "selection_metric": {
+    "mode": "min",
+    "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+    "scope": "seed-42 validation/top-k selection",
+    "value": 10.069027
+  },
+  "source": {
+    "checkpoint_basename": "epoch=45-step=40204-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=10p0690.ckpt",
+    "sha256": "7a68066d135ab835314779a789305547ecf18de355113b21f917171e4b8358fb"
+  },
+  "weight_variant": "raw"
+}

raw/shelves/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2051bd6fc6e2bc297f54f5ef97f9dd7d3f0515fa595aab78cd61d5a15d169380
+size 1020987464

raw/windows/config.yaml

@@ -0,0 +1,76 @@
+format_version: 3dcov-inference-config-v2
+task_name: CovDiffusion
+io_type: CovDiffusion
+dataset:
+- windows-v2
+action_dim: 24
+horizon: 16
+n_action_steps: 100
+n_obs_steps: 1
+encoder_output_dim: 256
+shape_meta:
+  obs:
+    point_cloud:
+      shape:
+      - 5120
+      - 3
+    low_dim:
+      shape:
+      - 24
+  action:
+    shape:
+    - 24
+diffusion:
+  model_type: dp3
+  diffusion_step_embed_dim: 128
+  down_dims:
+  - 512
+  - 1024
+  - 2048
+  kernel_size: 5
+  n_groups: 8
+  condition_type: film
+  use_down_condition: true
+  use_mid_condition: true
+  use_up_condition: true
+  num_inference_steps: 10
+  obs_as_global_cond: true
+model:
+  backbone: dp3
+  affinetrans: false
+  hidden_size:
+  - 1024
+  - 1024
+  pretrained: false
+  pretrained_custom: null
+  load_strict: false
+noise_scheduler:
+  _target_: diffusers.schedulers.scheduling_ddim.DDIMScheduler
+  num_train_timesteps: 100
+  beta_start: 0.0001
+  beta_end: 0.02
+  beta_schedule: squaredcos_cap_v2
+  clip_sample: true
+  set_alpha_to_one: true
+  steps_offset: 0
+  prediction_type: sample
+partial_observation_enabled: false
+partial_observation_method: fixed_camera
+partial_observation_ratio: 0.3
+ablation_prev_traj: normal
+random_traj_seed: 123
+pc_points: 5120
+traj_points: 10
+lambda_points: 4
+overlapping: 1
+asymm_overlapping: false
+normalization: per-dataset
+extra_data:
+- orientnorm
+weight_orient: 0.25
+load_extra_data:
+- stroke_masks
+traj_with_equally_spaced_points: true
+equal_spaced_points_distance: 0.05
+equal_in_3d_space: false
+augmentations: []

raw/windows/metrics.json

@@ -0,0 +1,31 @@
+{
+  "category": "windows",
+  "dataset": "windows-v2",
+  "epoch": 165,
+  "format_version": "3dcov-checkpoint-metadata-v1",
+  "global_step": 48472,
+  "release": {
+    "exact_tensor_roundtrip": true,
+    "git_revision": "10403c5d7365d765522f279d892b581d1d4083e5",
+    "model_sha256": "a0c381fbffbe2e12c83a1a40bc8a2591717f12579561a21e146ce4b1b18156ae",
+    "normalizer_fields": [
+      "action",
+      "point_cloud"
+    ],
+    "tensor_bytes": 1020966632,
+    "tensor_count": 182
+  },
+  "run_id": "TML4Q-S42",
+  "seed": 42,
+  "selection_metric": {
+    "mode": "min",
+    "name": "pred_cond_Pred_Cond_mean_point-wise_chamfer_distance",
+    "scope": "seed-42 validation/top-k selection",
+    "value": 10.410878
+  },
+  "source": {
+    "checkpoint_basename": "epoch=165-step=48472-pred_cond_Pred_Cond_mean_point-wise_chamfer_distance=10p4109.ckpt",
+    "sha256": "bcb0db0b7a5f0c2a4c7bf0472785b5588c4288d627ec212379a37c40b64f14e3"
+  },
+  "weight_variant": "raw"
+}

raw/windows/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0c381fbffbe2e12c83a1a40bc8a2591717f12579561a21e146ce4b1b18156ae
+size 1020987464