Model card: latents, exclude data/ from checkpoint download

README.md

@@ -22,6 +22,11 @@ Code, environments and full instructions: the Roboactualizer repository.
 Each folder holds the checkpoint, the `dataset_stats.json` used for action/state normalization (keep it next to the
 checkpoint; evaluation looks for it there) and the training `config.yaml`. `SHA256SUMS` lists file hashes.
 
+`data/` holds the precomputed V-JEPA 2.1 (300M) training latents with their `latent_norm_stats.json`:
+`data/int2_clip4/libero/` (LIBERO, clip 4, frame interval 2, 208 GB) and `data/int4_clip4/robotwin/` (RoboTwin 2.0
+50-demo subset, clip 4, frame interval 4, 518 GB). Download them from the code repository root with
+`--include "data/int2_clip4/libero/*" --local-dir .` (or the RoboTwin path).
+
 All checkpoints use the 300M V-JEPA 2.1 encoder:
 `https://dl.fbaipublicfiles.com/vjepa2/vjepa2_1_vitl_dist_vitG_384.pt`, saved as
 `checkpoints/vjepa2/vjepa2_1_vit_large_384.pt` in the code repository.
@@ -29,7 +34,7 @@ All checkpoints use the 300M V-JEPA 2.1 encoder:
 ## Usage
 
 ```bash
-huggingface-cli download db12312607/Roboactualizer --local-dir checkpoints/roboactualizer
+huggingface-cli download db12312607/Roboactualizer --exclude "data/*" --local-dir checkpoints/roboactualizer
 
 # from the code repository root
 python experiments/libero/run_libero_manager.py --config-name eval_libero_dits \