Release missing LeVo checkpoints on Hugging Face

[NielsRogge]

Hi @tencent-ailab 🤗

I'm Niels and work as part of the open-source team at Hugging Face. I discovered your work through Hugging Face's daily papers, the paper page is here: https://huggingface.co/papers/2506.07520.
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.

It's great to see a pre-trained model released on Hugging Face already. I noticed that the Hugging Face link mentioned in your README and on the paper page links to the tencent/SongGeneration/tree/main/ckpt/songgeneration_base_zh directory,
rather than to https://huggingface.co/waytan22/SongGeneration. It may be worth updating this link.

Additionally, I saw that you plan to release the following two checkpoints on Hugging Face as well. Is there an ETA on those, and could I help in any way?

• SongGeneration-base(zh&en)
• SongGeneration-full(zh&en)

It'd be great to make those checkpoints available on the 🤗 hub, to improve their discoverability/visibility.
We can add tags so that people find them when filtering https://huggingface.co/models.

See here for a guide: https://huggingface.co/docs/hub/models-uploading.

In this case, we could leverage the PyTorchModelHubMixin class which adds from_pretrained and push_to_hub to any custom nn.Module. Alternatively, one can leverages the hf_hub_download one-liner to download a checkpoint from the hub.

We encourage researchers to push each model checkpoint to a separate model repository, so that things like download stats also work. We can then also link the checkpoints to the paper page.

Let me know if you're interested/need any help regarding this!

Kind regards,

Niels

[tencent-ailab]

Thank you for your suggestion!​​

Due to the large number of checkpoints in our model, using PyTorchModelHubMixin for packaging would require excessive effort. For this reason, we’ve decided to store the checkpoints in the current format for now.

Because of this, we can only host all models in a single repository.

Additionally, the ​​SongGeneration-base (zh&en)​​ and ​​SongGeneration-full (zh&en)​​ models are currently under training. Once completed, they will be open-sourced in this repository.

[NielsRogge]

Ok thanks for your reply!

I've opened a PR to add download metrics for you: huggingface/huggingface.js#1543

Let me know if you prefer another extension to track downloads for