README: 1,100 / 1,100 demos; the 41 recovered demos; which corpus each model trained on

README.md

@@ -26,20 +26,20 @@ Private repo. The read-only access token is in the `Checkpoints` section of the
 
 | task | arms | demos | success | frames | cameras | action dim | GB | prompt |
 |---|---|---|---|---|---|---|---|---|
-| `bimanual_assembly` | 2 | 96 | 96 | 69,793 | back, handcam_rgb_left, handcam_rgb_right | 46 | 15.9 | Grasp the tray with the left hand and the peg with the right hand, then insert the peg into the hole. |
-| `bimanual_hanoi` | 2 | 86 | 86 | 92,675 | back, handcam_rgb_left, handcam_rgb_right | 46 | 22.2 | Execute the final two moves of the three-level Tower of Hanoi: move the medium disk from the middle peg to the right peg with the right hand, then move the small disk from the left peg to the right peg with the left hand. |
+| `bimanual_assembly` | 2 | 100 | 100 | 72,642 | back, handcam_rgb_left, handcam_rgb_right | 46 | 16.5 | Grasp the tray with the left hand and the peg with the right hand, then insert the peg into the hole. |
+| `bimanual_hanoi` | 2 | 100 | 100 | 109,105 | back, handcam_rgb_left, handcam_rgb_right | 46 | 26.3 | Execute the final two moves of the three-level Tower of Hanoi: move the medium disk from the middle peg to the right peg with the right hand, then move the small disk from the left peg to the right peg with the left hand. |
 | `bimanual_microwave_cook` | 2 | 100 | 100 | 85,441 | back, handcam_rgb_left, handcam_rgb_right | 46 | 20.7 | Open the microwave door, place the food inside the microwave, close the door, and press the start button. |
-| `bimanual_photograph` | 2 | 99 | 99 | 39,974 | back, handcam_rgb_left, handcam_rgb_right | 46 | 8.2 | Grasp the camera with the left hand, align it with the logo, and press the shutter button with the right hand. |
+| `bimanual_photograph` | 2 | 100 | 100 | 40,255 | back, handcam_rgb_left, handcam_rgb_right | 46 | 8.3 | Grasp the camera with the left hand, align it with the logo, and press the shutter button with the right hand. |
 | `bimanual_unlock_ipad` | 2 | 100 | 100 | 53,144 | back, handcam_rgb_left, handcam_rgb_right | 46 | 12.1 | Grasp the iPad and enter the password 123 to unlock the device. |
 | `click_mouse` | 1 | 100 | 100 | 40,296 | front, handcam_rgb | 23 | 5.5 | Move the mouse to the purple mouse pad and click the left mouse button. |
-| `fold_glasses` | 1 | 92 | 92 | 52,577 | front, handcam_rgb | 23 | 8.0 | Fold the glasses and place them into the case. |
+| `fold_glasses` | 1 | 100 | 100 | 57,900 | front, handcam_rgb | 23 | 8.8 | Fold the glasses and place them into the case. |
 | `hammer_nail` | 1 | 100 | 100 | 25,034 | front, handcam_rgb | 23 | 4.3 | Use the hammer to drive the nail into the wooden board. |
 | `pick_bucket` | 1 | 100 | 100 | 46,910 | front, handcam_rgb | 23 | 7.3 | Place the boxed food into the bucket and then lift the bucket. |
-| `pinch_tongs` | 1 | 86 | 86 | 37,071 | front, handcam_rgb | 23 | 5.0 | Grasp the tongs and perform three consecutive open-close motions. |
+| `pinch_tongs` | 1 | 100 | 100 | 43,441 | front, handcam_rgb | 23 | 5.9 | Grasp the tongs and perform three consecutive open-close motions. |
 | `water_plant` | 1 | 100 | 100 | 27,753 | front, handcam_rgb | 23 | 4.3 | Grasp the watering can and apply water to the plant. |
 
-Only demos whose replay reaches the task's success predicate are included. Some tasks therefore have fewer than 100
-demos. A released demo that does not succeed on replay in the current environment is left out.
+All 1,100 released demos are included, 100 per task, and every one reaches the task's success predicate on replay.
+83 of them do not succeed when replayed as released; they were recovered (see "How it was made").
 
 ## H5 schema (per demo, T = `n_frames`, frames at 30 Hz)
 
@@ -75,6 +75,13 @@ Large arrays are gzip-compressed. Read them with `h5py`.
   height is also restored to the env's own rule (0.92 m + table offset). The stand->iPad offset and the recorded
   actions are unchanged, and success is the env's own `info["succeed"]`. With the 58 demos that replay as released,
   that gives 100. The 42 recovered demos show the lit buttons and the unlocked screen; the 58 older captures do not.
+- The other 41 demos that fail on plain replay (pinch_tongs 14, bimanual_hanoi 14, fold_glasses 8, bimanual_assembly 4,
+  bimanual_photograph 1) are marginal contact outcomes: the release stores the initial object poses as float32, and
+  the replay flips with tiny numeric differences. 40 are recovered by a search over the initial object poses: 23 by
+  the plain replay, 17 with a position offset of at most 3.2 mm (search order: plain replay, then random offsets at
+  sigma 1 mm, then 3 mm; the smallest successful offset is kept; orientations and actions unchanged). `hanoi_demo_11_2026-03-13_15-46-49_120095` succeeds
+  on a plain replay on a different CPU (AMD EPYC 7452). All replays run with `OMP_NUM_THREADS=1`; without the
+  thread pin, marginal demos flip. Added 2026-10-28.
 
 ## Download
 
@@ -82,7 +89,7 @@ Large arrays are gzip-compressed. Read them with `h5py`.
 pip install -U huggingface_hub   # the `hf` CLI (huggingface-cli was removed in 1.x)
 hf auth login --token <read-only token, see above>
 hf download chrockey/DexJoCo-preprocessed --repo-type dataset --local-dir DexJoCo-preprocessed --include "water_plant/*"   # one task
-hf download chrockey/DexJoCo-preprocessed --repo-type dataset --local-dir DexJoCo-preprocessed                             # all (~113 GB)
+hf download chrockey/DexJoCo-preprocessed --repo-type dataset --local-dir DexJoCo-preprocessed                             # all (~120 GB)
 ```
 
 or from Python:
@@ -94,5 +101,13 @@ snapshot_download("chrockey/DexJoCo-preprocessed", repo_type="dataset", local_di
 ```
 
 To train PointWAM on it: the GitHub README's `DexJoCo training data` section has the symlinks that put these
-folders where `configs/data/dexjoco/*.yaml` expect them. The multi-task config for this 100-demo ipad corpus is
-`data=dexjoco/mt11_ipadrec`.
+folders where `configs/data/dexjoco/*.yaml` expect them. Which demos each released model trained on:
+
+| corpus | demos | config | model |
+|---|---|---|---|
+| everything here | 1,100 | (none yet) | — |
+| all but `hanoi_demo_11_2026-03-13_15-46-49_120095` | 1,099 | `data=dexjoco/mt11_ipadrec_drec` | PointWAM v1.1.0 / v1.1.1 |
+| all but the 41 above | 1,059 | `data=dexjoco/mt11_ipadrec` | — |
+| all but the 41 and the 42 recovered unlock_ipad demos | 1,017 | `data=dexjoco_all11_mixed_224` | PointWAM v1.0.0 (ICLR) |
+
+The 41 are listed in `experiments/dexjoco-demo-recovery.md` of the GitHub repo.