Include encoded SS targets for 3dfront in standard restoration

README.md

@@ -12,7 +12,7 @@ pretty_name: PartProbe3D compact reproduction data
 # PartProbe3D reproduction data
 
 This release supports SS, generative SLAT segmentation, and 32³ SLAT shape training.
-The default transfer is **166.32 GB**, including both training datasets and object shape targets.
+The default transfer is **197.57 GB**, including observations, SS targets, and available shape targets.
 Default downloads exclude optional and deprecated folders.
 
 <!-- partprobe3d-required-files:start -->
@@ -71,7 +71,7 @@ Add `--keep-archives` only when you need temporary transfer files.
 
 | Dataset | SS/segmentation training views | Shape training views | Monitoring views | Transfer, GB |
 | --- | ---: | ---: | ---: | ---: |
-| 3dfront | 39,930 | 39,930 | 16 | 91.76 |
+| 3dfront | 39,930 | 39,930 | 16 | 123.01 |
 | sage | 17,456 | 17,453 | 16 | 74.56 |
 | foundationpose | 42,256 | Unavailable | 16 | 61.76 |
 
@@ -191,3 +191,28 @@ Apply the loader patch described in the bundle documentation before evaluating i
 python -m preprocess.huggingface download --repo cubec/PartProbe3D-preprocessed --dataset evaluation
 ```
 <!-- partprobe3d-frozen-evaluation:end -->
+
+<!-- partprobe3d-ss-targets:start -->
+## Encoded SS targets
+
+The standard download includes encoded SS targets for the datasets listed below.
+These targets remove GPU encoding from the restore procedure.
+They retain the original cache manifests, latent tensors, and monitoring targets.
+Their native latent shape is `8×16×16×16`.
+Model weights remain separate dependencies.
+
+| Dataset | Target views | Transfer, GB |
+| --- | ---: | ---: |
+| 3dfront | 49,996 | 31.249 |
+
+Use the current project checkout and download the selected datasets:
+
+```bash
+python -m preprocess.huggingface download --repo cubec/PartProbe3D-preprocessed
+```
+
+Publication continues for the remaining datasets. The listed datasets already include SS targets.
+The downloader checks archive hashes before extraction.
+The training loader checks dataset and frozen codec identities.
+The image encoders compute features during training.
+<!-- partprobe3d-ss-targets:end -->

training/index.json

@@ -10,16 +10,29 @@
     "3dfront": {
       "required": [
         "training/observations/3dfront",
-        "training/shape/32/3dfront"
+        "training/shape/32/3dfront",
+        "training/ss/3dfront"
       ],
-      "bytes": 91759687680,
+      "bytes": 123008819200,
       "source_manifest_sha256": "93fd87c7bc4e59a08b401fe83c36c02591441406f5b0535fd9e2f135b5952446",
       "resolution": 64,
       "source_views": 49996,
       "views": 39946,
       "monitor_views": 16,
       "training_views": 39930,
-      "shape_excluded_views": []
+      "shape_excluded_views": [],
+      "ss_targets": {
+        "component": "training/ss/3dfront",
+        "bytes": 31249131520,
+        "views": 49996,
+        "latent_shape": [
+          8,
+          16,
+          16,
+          16
+        ],
+        "target_scope": "masked_objects_complete_occupancy"
+      }
     },
     "sage": {
       "required": [
@@ -101,5 +114,6 @@
       "evaluation"
     ],
     "required_for_evaluation": true
-  }
+  },
+  "restore_git_revision": "32d92a1e1b9e99eafaac9dda13c43e676a7aeeb3"
 }