Downloads · 30 days
0
sanity15k_annealing
lair-nyu/sanity15k_annealing
sanity15k_annealing is a machine learning model from lair-nyu. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
pi0.5 sanity check (4-task pnp subset, 15k steps, correctly-scaled cosine decay, batch 128, 2xH200). Re-run of the earlier yoriclpi05easypnpv2sanity15k check with lrschedule.decaysteps fixed to match numtrainsteps=150…
Access
Public
Updated Aug 28, 2026
Repo size
12.4 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
Other12.4 GB · 100%
At a glance
- Access
- Public
- Created
- Aug 28, 2026
- Updated
- Aug 28, 2026
- SHA
- 3af4f9c4
Model card
From the Hugging Face model README
sanity15k_annealing
pi0.5 sanity check (4-task pnp subset, 15k steps, correctly-scaled cosine
decay, batch 128, 2xH200). Re-run of the earlier yor_icl_pi05_easy_pnp_v2_sanity15k
check with lr_schedule.decay_steps fixed to match num_train_steps=15_000
(the earlier run left decay_steps=50_000, so its cosine decay only
completed ~30% of its span).
- openpi TrainConfig:
sanity15k_annealing - Checkpoint step: 14999
- Contents:
params/(deployable policy weights) +assets/(norm stats etc.) only --train_state/(optimizer state, needed only to resume this exact training run) is NOT included. - Trained with openpi (pi0.5 backbone) on the icl-dataset.
- Created
- Aug 28, 2026
- Updated
- Aug 28, 2026