pi05_real_pb_from_left

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

Task

  • Task: Push Block
  • Training data: From-left mode only
  • Dataset: real_push_block_from_left
  • 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 = 0.0064
  • Step 4000: loss = 0.0037
  • Step 4999

Loss Curve

Step Loss
0 0.0946
500 0.0150
1000 0.0126
1500 0.0105
2000 0.0083
2500 0.0069
3000 0.0064
3500 0.0053
4000 0.0037
4500 0.0034

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