Hugging Face's logo

AlexWortega
/
tinyvla

Robotics
LeRobot
Safetensors
vla
libero
dual-rate
Model card Files
xet
 Community
tinyvla / tinyvla_b200 /scripts

Commit History

  1. VLA closed-loop driver (fixed head camera + auto-gain): falls <1s all behaviors — BC covariate shift on unstable system

    AlexWortega
    Verified signature
    57af0e0

  2. closed-loop MicroDuck harness: obs.rs 61D builder validated by ONNX teacher (stands 8s @10Hz replan)

    AlexWortega
    Verified signature
    5a27506

  1. CRITICAL fix: torch.compile'd expert saved as expert._orig_mod.* -> from_pretrained silently dropped it (random flow head at eval). Save the uncompiled module

    AlexWortega
    Verified signature
    f0fcec5

  2. lekiwi zero-shot eval harness (hub streaming, 2x2 demo/descriptor grid)

    AlexWortega
    Verified signature
    a6a7adb

  1. eval_microduck: zero-shot eval+render harness (v4 mode: demos from same behavior, duck descriptor)

    AlexWortega
    Verified signature
    4dabbd2

  2. train_fast: non-persistent hub workers (recycle each epoch to reset leaks)

    AlexWortega
    Verified signature
    d990e88

  1. 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

    AlexWortega
    Verified signature
    b152c62

  2. train_fast: v4 all-channels via hub stream (support passthrough, morph_text tokenization, guarded shard morph block)

    AlexWortega
    Verified signature
    d9b76b2

  1. fix DDP hang: spawn dataloader workers (fork-after-CUDA/NCCL starved torchrun heartbeat), picklable Collate, fewer workers per rank

    AlexWortega
    Verified signature
    4ac1f5f

  2. fix: yaml 1.1 parses '2250.0e12' as string -> float() cast in MFU calc (crashed first log step)

    AlexWortega
    Verified signature
    fd625f9

  3. train_fast: torchrun DDP (per-rank streaming, rank0 logging/saves, global throughput)

    AlexWortega
    Verified signature
    0febc93

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

    AlexWortega
    Verified signature
    51ce5b3