[FEAT] <Brief Description>

Nobody has claimed this yet.

### 💡 Is your feature request related to a problem?
First of all, thank you for this project. It has helped me a lot. However, I noticed that there is currently no related content or examples for distillation and pretraining with the SAM series models. This makes it difficult to apply the framework when working with SAM-based architectures.
A clear and concise description of the problem you're trying to solve.

### 🧰 Describe the solution you'd like
I hope the project could provide examples or documentation for distillation and pretraining using SAM series models (such as SAM, SAM2, etc.). Ideally, this could include:

Example training pipelines for SAM-based models

Guidance on how to integrate SAM models into the framework

Example configs or scripts for distillation or pretraining

This would make it much easier for users who want to apply the framework to SAM-related tasks.
What should the feature do? How would it work?

### 🛠 Alternatives you've considered
Currently, the only option is to integrate SAM models manually by defining MyModelWrapper. However, this approach requires significant effort, and without official examples, it may lead to potential issues during implementation.

### 📝 Additional context
Adding official examples or documentation for SAM-series models would greatly improve usability for researchers working on segmentation and foundation models, and would help more users adopt the project.
Add any other context, designs, or screenshots here.

Start by reviewing the existing MyModelWrapper integration, since the issue identifies it as the current manual path for SAM models. Define the scope for SAM/SAM2 support, including training pipelines, integration guidance, and distillation or pretraining configs and scripts; done means those examples or documentation are available for users.

Written by the indexing model from the issue text.