domains / families / media / openvla-oft

openvla-oft

separate

Fine-tunes and evaluates OpenVLA-OFT and OpenVLA-OFT+ policies for robot action generation with continuous action heads, LoRA adaptation, and FiLM conditioning on LIBERO simulation and ALOHA real-world setups. Use when reproducing OpenVLA-OFT paper results, training custom VLA action heads (L1 or diffusion), deploying server-client inference for ALOHA, or debugging normalization, LoRA merge, and cross-GPU issues.

When to use

• The target task is robot action generation with visual and language conditioning
• LoRA-based adaptation of `openvla/openvla-7b` is preferred
• You need official LIBERO or ALOHA workflows from the OpenVLA-OFT paper
• You want continuous action heads (L1 regression or diffusion) instead of tokenized actions
• You need a different VLA architecture (use `fine-tuning-serving-openpi` for pi0/pi0.5 models)
• You need the NVIDIA Cosmos Policy stack (use `evaluating-cosmos-policy`)
• You need general LLM fine-tuning without robot action heads

When not to use

• different VLA architecture (pi0/pi0.5 models)
• NVIDIA Cosmos Policy stack
• general LLM fine-tuning without robot action heads

Skill page

Source