Instructions to use jclinton1/act-gem-stack-blocks-three-200k-val with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use jclinton1/act-gem-stack-blocks-three-200k-val with LeRobot:
- Notebooks
- Google Colab
- Kaggle
metadata
library_name: lerobot
tags:
- robotics
- lerobot
- act
- gem-stack-blocks-three
datasets:
- jclinton1/gem_stack_blocks_three_20260502_194746
ACT Gem Stack Blocks Three 200k
ACT policy trained for 200,000 steps on jclinton1/gem_stack_blocks_three_20260502_194746.
Validation used the final 10% of dataset episodes, episodes 278-307.
Final validation metrics at step 200000:
val_loss: 0.3886val_l1_loss: 0.3885val_kld_loss: 0.0000
Feature rename map used during training:
{
"observation.images.head": "observation.images.image",
"observation.images.wrist": "observation.images.image2"
}