Use ADT tar archives instead of expanded ADT files

README.md

@@ -58,21 +58,41 @@ Install the Hugging Face CLI if needed:
 pip install -U "huggingface_hub[cli]"
 ```
 
-Download the dataset while preserving repository paths:
+Download the dataset while preserving repository paths. The expanded ADT frame tree is not stored directly in the repository; download `aria_digital_twin/ADT_tars/*.tar` and restore it locally.
 
 ```bash
 DATA_ROOT=/mnt/nvme6/jieneng/data/spatialencoder_full
 mkdir -p "$DATA_ROOT"
 
-huggingface-cli download qicq1c/spatialencoder_full \
-  --repo-type dataset \
+hf download qicq1c/spatialencoder_full \
+  --type dataset \
   --local-dir "$DATA_ROOT" \
-  --local-dir-use-symlinks False
+  --exclude "aria_digital_twin/ADT/**"
 ```
 
-The full dataset is about 2 TiB, so make sure the target filesystem has enough space before starting.
+The ADT tar archives are about 405 GiB. After extraction, `aria_digital_twin/ADT/` is about 406 GiB. Keep both the archives and expanded ADT tree if you want the download to be resumable and reproducible.
 
-## 2. Expand Packed Add-Ons If Present
+## 2. Restore ADT From Tar Archives
+
+Restore the ADT frame tree from the downloaded tar archives:
+
+```bash
+cd "$DATA_ROOT"
+mkdir -p aria_digital_twin/ADT
+
+for tar_file in aria_digital_twin/ADT_tars/*.tar; do
+  tar --skip-old-files -xf "$tar_file" -C aria_digital_twin/ADT
+done
+```
+
+This recreates paths such as:
+
+```text
+aria_digital_twin/ADT/Apartment_release_clean_seq131_M1292/depth_frames/...
+aria_digital_twin/ADT/Apartment_release_clean_seq131_M1292/rgb_frames/...
+```
+
+## 3. Expand Packed Add-Ons If Present
 
 If the download contains archive files such as `pickle.zip`, `hyperism-train-json.zip`, or `hyperism-val-json.zip`, unzip them at the data root:
 
@@ -100,7 +120,7 @@ fi
 
 Skip this step for files that are already expanded in the final tree.
 
-## 3. Check Required Files
+## 4. Check Required Files
 
 Run these checks before training:
 
@@ -129,7 +149,7 @@ Expected minimum result:
 - `BoxFromMotion/dataset/{CA-1M,hyperism,ADT}.json` exist.
 - Hyperism and ADT frame paths referenced by the json files exist under `BOX_DATA_PATH`.
 
-## 4. Set Training Environment
+## 5. Set Training Environment
 
 From the SpatialEncoder code checkout:
 
@@ -159,7 +179,7 @@ wget -P "$BOX_WEIGHTS_PATH" \
   https://huggingface.co/facebook/sam3.1/resolve/main/sam3.1_multiplex.pt
 ```
 
-## 5. Dataset Smoke Tests
+## 6. Dataset Smoke Tests
 
 The merged training config samples datasets according to:
 
@@ -205,7 +225,7 @@ Train Epoch: [0][   0/...] ... Losses/train_all_loss: ...
 
 The first batch can be slow because workers are filling caches. Later steps should have near-zero `Data Time`.
 
-## 6. Mixed Training
+## 7. Mixed Training
 
 Once all three single-dataset smoke tests print loss, launch the mixed run: