lerobot_policy_turbovla_so101
Copyright 2026 ravediamond

This product includes software originally developed as part of
turbovla-lerobot-plugin (https://github.com/ez1540/turbovla-lerobot-plugin),
Copyright 2026 The TurboVLA-LeRobot contributors, licensed under the
Apache License, Version 2.0.

The module structure (configuration/modeling/processor split), the
PreTrainedConfig/PreTrainedPolicy integration, the bidirectional
vision-language fusion layers, and the ACT-style action decoder are
adapted from that project. See README.md for the list of changes made
in this fork.

TurboVLA is itself based on research published as:
  Xie, H., Yao, C., Wu, X., et al. "TurboVLA: Real-Time Vision-Language-Action
  Model at 32 Hz on an RTX 4090 with <1 GB VRAM." arXiv:2607.27205 (2026).
  Upstream research code: https://github.com/H-EmbodVis/TurboVLA
