libero_seen_pretrain
Collection
Seen-suite (libero_90) pretraining for LIBERO-Goal few-shot: SmolVLA action BC + action-free latent dynamics (IDM/FDM). • 5 items • Updated
Latent dynamics (IDM/FDM) · libero_90 · seed 1000 · shuffled-future control
Action-free visual dynamics pretrain: frozen SmolVLA vision features → IDM → FDM.
Trained on RGB from nvidia/libero_90 only (no action / state / reward labels).
Control run: trained with shuffle_future=True — the future RGB frame is sampled from a different episode, so temporal dynamics are broken. Same architecture / steps / data volume as the real IDM; use as a negative control for video-latent fine-tuning (VIDEO_CKPT=.../idm_fdm.pt from this repo).
Files:
idm_fdm.pt— model weights + config dictidm_fdm.json— training hyperparameters
Used as frozen teacher for video-latent few-shot SmolVLA fine-tuning
(--video.checkpoint).