Instructions to use allenai/MolmoAct2-FAST-Tokenizer with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use allenai/MolmoAct2-FAST-Tokenizer with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("allenai/MolmoAct2-FAST-Tokenizer", dtype="auto") - Notebooks
- Google Colab
- Kaggle
Diff with Pi FAST
#1
by ifarhatTII - opened
Hello, thank you for the great release
what does it differ from the FAST tokenizer of pi ?
Hi! Thanks for the question!
The main motivation is to make the whole model fully end-to-end open-source, so we trained our own action tokenizer entirely on open-source datasets. The tokenizer architecture and training pipeline are largely the same as the FAST tokenizer published by pi. One thing to notice is that we train with action chunk sizes matched to the control frequency of each dataset. More details could be found in our technical report.
hqfang changed discussion status to closed