ACT — v6 blue89 180ep, GelSight 320x240
- Dataset: Kaz55/dg5f_ur5e_v6_blue89_180ep_gs320 — 180 episodes / 208,109 frames
- GelSight: 320x240
- RealSense: 640x480
- Policy: ACT, chunk_size=60, n_action_steps=60
- Training: 100,000 steps (~3.84 epochs), batch 8, seed 1000
Inputs
observation.state (26) + RealSense x2 + GelSight x2
observation.velocity and observation.effort exist in the dataset but are
deliberately excluded — feature auto-derivation would otherwise feed them to
the policy and add a second difference between runs.
Sweep
| GelSight | model | loss |
|---|---|---|
| 500x375 | act-v6-blue89-180ep-gs500-ac60 | 0.132 |
| 320x240 | act-v6-blue89-180ep-gs320-ac60 | 0.134 |
Caveat
Different dataset from the blue_180ep sweep (186 ep / 186,152 frames) — losses
are not comparable across the two. On the related combined sweep, training loss
was identical at every GelSight resolution including no-GelSight-at-all, so
treat loss as a sanity check rather than evidence about tactile resolution; that
needs on-robot evaluation.
- Downloads last month
- 14