add: downloading data utils

examples/behavior_1k/download.slurm

@@ -0,0 +1,27 @@
+#!/bin/bash
+#SBATCH -J b1k-download
+#SBATCH -p hopper-cpu
+#SBATCH --qos=high
+#SBATCH -c 64                    # CPUs per conversion (tune as needed)
+#SBATCH -t 20:00:00              # Time per conversion
+#SBATCH -D /admin/home/francesco_capuano/lerobot
+#SBATCH -o /admin/home/francesco_capuano/lerobot/examples/behavior_1k/logs/%x-%A.out
+#SBATCH -e /admin/home/francesco_capuano/lerobot/examples/behavior_1k/logs/%x-%A.err
+
+set -euo pipefail
+set -x
+export PYTHONUNBUFFERED=1
+export OMP_NUM_THREADS=${SLURM_CPUS_PER_TASK:-1}
+
+# activate your env if needed
+source "$HOME/.bashrc" 2>/dev/null || true
+if ! command -v conda >/dev/null 2>&1; then
+  source "$HOME/miniconda3/etc/profile.d/conda.sh" 2>/dev/null || true
+  source "$HOME/anaconda3/etc/profile.d/conda.sh" 2>/dev/null || true
+fi
+conda activate lerobot
+
+python examples/behavior_1k/download_data.py \
+  --repo-id "behavior-1k/2025-challenge-demos" \
+  --local-dir "/fsx/francesco_capuano/behavior1k-2025-v21" \
+  --max-workers 64

examples/behavior_1k/download_data.py

@@ -1,8 +1,26 @@
+import shutil
+
 from huggingface_hub import snapshot_download
 
 if __name__ == "__main__":
+    import argparse
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--repo-id", type=str, required=True)
+    parser.add_argument("--max-workers", type=int, default=8)
+    parser.add_argument("--local-dir", type=str, required=True)
+    parser.add_argument("--force-download", action="store_true")
+
+    args = parser.parse_args()
+
+    if args.force_download:
+        shutil.rmtree(args.local_dir, ignore_errors=True)
+
     snapshot_download(
-        repo_id="behavior-1k/2025-challenge-demos",
+        repo_id=args.repo_id,
         repo_type="dataset",
-        local_dir="/fsx/francesco_capuano/behavior1k-2025-v21",
+        force_download=args.force_download,
+        max_workers=args.max_workers,
+        local_dir=args.local_dir,
+        ignore_patterns=["annotations/*"],  # NOTE(fracapuano): Dropping textual annotations right now
     )