README.md

metadata

license: apache-2.0
library_name: transformers
base_model: 2toINF/X-VLA-Pt
tags:
  - robotics
  - vision-language-action
  - lerobot
  - x-vla
  - openarm

X-VLA OpenArm POS 2-Camera

Fine-tuned from 2toINF/X-VLA-Pt on asterisms45/my_openarm_dataset_test.

Training setup:

• dataset format: LeRobot v3.0
• action/proprio: *.pos only
• effective action dim: 8
• cameras:
  • observation.images.cam_right_hand
  • observation.images.cam_ego
• excluded camera:
  • observation.images.cam_left_hand
• training steps: 20000

This checkpoint was trained in X-VLA with action_mode=auto, real_action_dim=8, and max_action_dim=8.

Loading

from transformers import AutoModel, AutoProcessor

model = AutoModel.from_pretrained(
    "asterisms45/X-VLA-openarm-pos-2cam",
    trust_remote_code=True,
)
processor = AutoProcessor.from_pretrained(
    "asterisms45/X-VLA-openarm-pos-2cam",
    trust_remote_code=True,
)