hybrid-framesamp-simplesubgoal-policy – AI Model by bingaochen | AlphaNeural AI

You can deploy this model and start earning money today!

bingaochen/
hybrid-framesamp-simplesubgoal-policy

like

0

vision-language-action

memory-augmented

Views

No views yet

hybrid-framesamp-simplesubgoal-policy

Memory-augmented VLA fine-tuned from the π₀.₅ backbone. Combines a FrameSamp visual-memory pathway with simple-subgoal symbolic conditioning.

Backbone: π₀.₅ (openpi)
Action space: joint_angle (8-dim)
License: MIT

Use

Serve as an openpi policy server:

bash

python scripts/serve_policy.py \
    --port=18001 \
    policy:checkpoint \
        --policy.dir=/path/to/this/ckpt \
        --policy.config=mme_vla_suite

The on-disk layout (after unzip 79999.zip):

_CHECKPOINT_METADATA
assets/
params/