Hi team, great work on the HELM architecture and the MiCE routing mechanism. I am currently building a pipeline to adapt the HELM-MiCE model for an agentic trajectory distillation task.
In your README.md under the Evaluation section, you mention:
- "Please doanload the 1B model at the following link" However, there is no actual hyperlink attached to that text, and the 1B weights do not appear to be public on your Hugging Face organization page.
I managed to track down the HELM-MiCE-100M checkpoint over on Zenodo (https://zenodo.org/records/18729608) to use as a local sandbox, but I need the 1B weights to execute the full evaluation.
Could you please update the README with the correct URL, or flip the Hugging Face repo for the 1B model to public if any?
Thanks for releasing the code!
Hi team, great work on the HELM architecture and the MiCE routing mechanism. I am currently building a pipeline to adapt the HELM-MiCE model for an agentic trajectory distillation task.
In your
README.mdunder the Evaluation section, you mention:I managed to track down the
HELM-MiCE-100Mcheckpoint over on Zenodo (https://zenodo.org/records/18729608) to use as a local sandbox, but I need the 1B weights to execute the full evaluation.Could you please update the README with the correct URL, or flip the Hugging Face repo for the 1B model to public if any?
Thanks for releasing the code!