Release CANOPY training stack and models on Hugging Face

Hi @AlibabaResearch 🤗

I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work through Hugging Face Daily Papers: https://arxiv.org/abs/2609.01245
The paper page lets people discuss about your paper and lets them find artifacts about it (your models for instance), you can also claim the paper as yours which will show up on your public profile at HF, add Github and project page URLs.

I see that you plan to release the complete training stack for CANOPY. That's great! Would you also consider releasing the trained models (e.g., the Qwen3-14B policy and Qwen3.5-9B models) on https://huggingface.co/models? Hosting them on Hugging Face would give them much better visibility and discoverability, and we can link them directly to the paper page.

If you're open to it, here's a guide on uploading models: https://huggingface.co/docs/hub/models-uploading. We also have the [PyTorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) class that makes it easy to push and load custom PyTorch models.

Let me know if you're interested or need any guidance — happy to help!

Kind regards,
Niels