141 kB

Ctrl+K

1 contributor

History: 12 commits

AlexWortega

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

57af0e0 verified 6 days ago

bench_model.py
5.8 kB
tinyvla_b200: B200 streaming training on nvidia/physical-ai (745 samples/s e2e, 7x ViT fast path, hub episode streaming, full instructions) 12 days ago
build_shards.py
17.7 kB
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 11 days ago
build_shards_v2.py
10.6 kB
tinyvla_b200: B200 streaming training on nvidia/physical-ai (745 samples/s e2e, 7x ViT fast path, hub episode streaming, full instructions) 12 days ago
convert_all.sh
5.45 kB
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 11 days ago
duck_closedloop.py
4.17 kB
closed-loop MicroDuck harness: obs.rs 61D builder validated by ONNX teacher (stands 8s @10Hz replan) 6 days ago
duck_vla_closedloop.py
5.19 kB
VLA closed-loop driver (fixed head camera + auto-gain): falls <1s all behaviors — BC covariate shift on unstable system 6 days ago
eval_lekiwi_hub.py
6.06 kB
lekiwi zero-shot eval harness (hub streaming, 2x2 demo/descriptor grid) 6 days ago
eval_microduck.py
9.46 kB
eval_microduck: zero-shot eval+render harness (v4 mode: demos from same behavior, duck descriptor) 9 days ago
fetch_wds.py
2.37 kB
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 11 days ago
gnm_to_wds.py
9.97 kB
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 11 days ago
hf_push_wds.py
1.91 kB
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 11 days ago
list_unitree_specs.py
2.16 kB
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 11 days ago
make_stream_specs.py
4.87 kB
tinyvla_b200: B200 streaming training on nvidia/physical-ai (745 samples/s e2e, 7x ViT fast path, hub episode streaming, full instructions) 12 days ago
make_wds_mix.py
2.07 kB
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 11 days ago
pack_wds.py
5.5 kB
tinyvla_b200: B200 streaming training on nvidia/physical-ai (745 samples/s e2e, 7x ViT fast path, hub episode streaming, full instructions) 12 days ago
redo_recon.sh
2.17 kB
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 11 days ago
train_fast.py
23.1 kB
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 6 days ago
unitree_parallel.sh
2.24 kB
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 11 days ago
validate_wds.py
6.56 kB
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 11 days ago
vamos_to_wds.py
6.95 kB
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 11 days ago
wds_direct.py
6.88 kB
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 11 days ago