Downloads · 30 days
0
pi0fast-piper-bimanual-lora-r16-step1000
Kavin60606/pi0fast-piper-bimanual-lora-r16-step1000
pi0fast-piper-bimanual-lora-r16-step1000 is a machine learning model from Kavin60606. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Fine-tuned from gs://openpi-assets/checkpoints/pi0fastbase on HD-EPIC egocentric pick-and-place clips for AgileX Piper-H bimanual robot (14-dim state+action).
Access
Public
Updated May 18, 2026
Repo size
8.6 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
Other8.6 GB · 100%
At a glance
- Access
- Public
- Created
- May 18, 2026
- Updated
- May 18, 2026
- SHA
- 53073fcc
Model card
From the Hugging Face model README
pi0-FAST Piper-H Bimanual LoRA (r=16) — Step 1000
Fine-tuned from gs://openpi-assets/checkpoints/pi0_fast_base on HD-EPIC egocentric pick-and-place clips for AgileX Piper-H bimanual robot (14-dim state+action).
Training config
- Model: pi0-FAST (PaliGemma 3B + SigLIP + FAST action tokenizer)
- LoRA rank: 16
- Action dim: 14, action horizon: 10
- Batch size: 32
- LR schedule: cosine, peak 2e-4, warmup 200, decay 5000, final 1e-5
- Optimizer: AdamW, grad clip 1.0
- Steps trained: 1000 (out of 5000 planned)
Files
params/— orbax-format LoRA params + base mergedtrain_state/— full train state (optimizer)assets/— norm stats + tokenizer assetstrain.log— full training logloss_curve.png— training loss + grad normconfig.txt— TrainConfig dump
- Created
- May 18, 2026
- Updated
- May 18, 2026