---
datasets:
- ShuaiYang03/VLA_Instruction_Tuning
base_model:
- nvidia/Eagle2-2B
pipeline_tag: robotics
---

# Model Card for InstructVLA-SimplerEnv

**Weights**

* `checkpoints/step-240000-epoch-05-loss=0.0910.pt`: Complete model checkpoint for direct evaluation. This checkpoint has better performance compared with multi-node version.

> If you want to try stage-2 VLA-IT training with this checkpoint, please use `https://github.com/InternRobotics/InstructVLA/blob/main/scripts/unload_lora.py` to get adapter and action_expert.pt.

**Evaluation Results**

* `results_step-240000-epoch-05-loss=0.0910_1-3`: Results on **SimplerEnv**.