antonitsin/cube-grasp-visual
SO-101 β’ Updated β’ 52 episodes β’ 162
How to use antonitsin/pi05_cube_grasp with LeRobot:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
Fine-tuned from lerobot/pi05_base (revision a538eb2732) on
antonitsin/cube-grasp-visual (52 episodes, SO-101 arm, overhead + wrist cameras).
Full fine-tune, 6,000 steps, batch size 16, bfloat16.
Trained and validated with lerobot 0.4.2 with the Οβ.β
transformers patches (pip install "lerobot[pi]==0.4.2").
It does not produce correct actions in lerobot 0.6.x.
observation.images.base_0_rgb: overhead camera (rename from observation.images.overhead)observation.images.left_wrist_0_rgb: wrist camera (rename from observation.images.wrist)observation.state: 6 joint positionsPut the cube in the glassGemma is provided under and subject to the Gemma Terms of Use found at ai.google.dev/gemma/terms.