pi05_real_pb_mixed

Fine-tuned pi0.5 VLA model for real robot manipulation.

Task

  • Task: Push Block
  • Training data: Mixed modes (from-left + from-right)
  • Dataset: real_push_block_mixed
  • Robot: Franka Panda (7-DOF)
  • Cameras: Base RGB + Wrist RGB (256x256)

Training Configuration

Parameter Value
Base model pi0.5 (PaliGemma 2B + Gemma 2B action expert)
Total parameters ~3.35B
Action dimension 32
Action horizon 10
Batch size 16
Training steps 5,000
Learning rate Cosine decay: warmup=500, peak=5e-5, end=5e-6
Optimizer AdamW (gradient clip norm=1.0)
GPUs 8x NVIDIA A100
Normalization Quantile normalization

Checkpoints

  • Step 3000: loss = 0.0071
  • Step 4000: loss = 0.0047
  • Step 4999

Loss Curve

Step Loss
0 0.0818
500 0.0155
1000 0.0120
1500 0.0110
2000 0.0084
2500 0.0075
3000 0.0071
3500 0.0060
4000 0.0047
4500 0.0042

Part of Mode Editing Research

This checkpoint is part of the "Don't Filter Your Data, Edit Your Policy" project (CoRL 2026), investigating post-hoc behavior mode editing for robot policies using Classifier-Guided Distillation (CG-Distill).

Downloads last month

-

Downloads are not tracked for this model. How to track
Video Preview
loading