act_cubcyl_objhold_white_kl1
ACT policy for the SO-ARM101, trained by Φ (Physical Hardware Intelligence), the robotics group at Northeastern University's Silicon Valley campus.
Part of an object-holdout generalization study: train on two of three objects, hold the third out entirely, and measure whether grasp behaviour transfers to an object size never seen in training.
Held out of training: white cube, episodes 80-119 (45 mm).
Training
| Dataset | BrutalCaesar/phi_so101_cubes_cylinder_v1 |
| Train / held-out episodes | 80 / 40 |
| Policy | ACT, 51.6M params |
| Cameras | 3 (wrist, front, top) |
chunk_size / n_action_steps |
50 |
kl_weight |
1 |
| batch size / steps | 8 / 100,000 |
| seed | 1000 |
| Hardware | 1x NVIDIA H200, 2:01:48 |
Held-out evaluation
eval_loss is pure L1 with the CVAE latent set to zero, computed on the 40 held-out episodes
every 10,000 steps. It is not the training objective: in eval mode ACT skips the VAE encoder, so no
KL term is included, which makes these numbers comparable across kl_weight settings.
| step | 10k | 20k | 30k | 40k | 50k | 60k | 70k | 80k | 90k | 100k | |---|---|---|---|---|---|---|---|---|---|---|| | eval_loss | 0.3708 | 0.3608 | 0.3564 | 0.3381 | 0.3425 | 0.3405 | 0.3289 | 0.3223 | 0.3256 | 0.3308 |
Best: 0.3223 at step 080000.
⚠️ The final checkpoint is not the best one. Training loss reached ~0.053 on every run in this study while held-out loss varied by up to 28%. Pick a checkpoint by held-out loss, not by the last step.
Use with LeLab
Parv-09/act_cubcyl_objhold_white_kl1@checkpoints/080000
Only that checkpoint is downloaded, not the whole repo. Any of
020000 040000 060000 080000 100000 is valid.
Caveats
- No hardware success rate yet. These numbers are held-out L1, not scored rollouts.
- The two holdout splits are not comparable to each other: different held-out sets (27,272 vs 18,277 frames), different objects, different episode lengths.
- Camera keys in the source dataset are correct and need no transposition.