fix: HubEpisodeStream.set_epoch was missing — first worker-recycle epoch boundary crashed the run

streaming: episodes_per_epoch worker lifetime cap (dataloader workers leaked to OOM-kill on multi-hour runs twice)

wds mixture: unitree+nav converters (gnm/vamos), wds source branch in train_fast, AMP autoselect (V100 fp16), embedding expand 16->32, push/fetch/validate tooling, DATA.md

streaming: in-context demo support cache (per-dataset episode ring, LRU), morphology+robot-text per sample

guard package init: keep bench/data paths importable without lerobot

streaming: stats fallback via episode-parquet sampling for repos without meta stats (Tuned-Tasks)

add data __init__

add modules __init__

add modules/embodiment.py

add package __init__

add processor_tinyvla.py

add missing configuration_tinyvla.py (package was incomplete)

tinyvla_b200: B200 streaming training on nvidia/physical-ai (745 samples/s e2e, 7x ViT fast path, hub episode streaming, full instructions)