Release In-Place TTT models on Hugging Face

Hi @GuhFeng 🤗

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 as yours got featured: https://huggingface.co/papers/2604.06169.
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 saw in your GitHub repository that the codebase for "In-Place Test-Time Training" is open-sourced, and you provide a script (`scripts/merge_dcp_to_hf.py`) to convert your checkpoints into Hugging Face format. This is fantastic!

Your paper mentions extensive experiments where In-Place TTT enables a 4B-parameter model to achieve superior performance and consistently outperforms TTT-related approaches when pretrained from scratch, also mentioning Qwen3-8B and LLaMA-3.1-8B. Would you be interested in hosting the specific pre-trained or continually trained In-Place TTT models from your experiments on https://huggingface.co/models? This would make it easier for the community to directly access and utilize the models you evaluated.

Hosting on Hugging Face will give you more visibility/enable better discoverability. We can add tags in the model cards so that people find the models easier, link them to the paper page, etc.

If you're down, leaving a guide [here](https://huggingface.co/docs/hub/models-uploading). If it's a custom PyTorch model, you can use the [PyTorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin)
class which adds `from_pretrained` and `push_to_hub` to the model which lets you to upload the model and people to download and use models right away.
You already have a conversion script, which is a great start! 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.

You can also build a demo for your model on [Spaces](https://huggingface.co/spaces), we can provide you a ZeroGPU [grant](https://huggingface.co/docs/hub/en/spaces-gpus#community-gpu-grants), which gives you A100 GPUs for free.

Let me know if you're interested/need any guidance :)

Kind regards,

Niels

Start with scripts/merge_dcp_to_hf.py and the Hugging Face model-uploading guide linked in the issue. Identify the pre-trained or continually trained checkpoints from the reported experiments, convert them to Hugging Face format, and upload each checkpoint to its own model repository. Done means the selected models are hosted and discoverable on Hugging Face.

Written by the indexing model from the issue text.