---
tags:
- diger
- rq-vae
- generative-recommendation
- semantic-id
- recommendation
pipeline_tag: feature-extraction
---

# DIGER RQ-VAE Checkpoint (instruments)

This repository provides the pretrained RQ-VAE checkpoint for **instruments** used in the paper:

**DIGER: Differentiable Semantic IDs for Generative Recommendation**

## Paper

This artifact is associated with the DIGER paper page:

https://huggingface.co/papers/2601.19711

## Files

- `best_collision_model.pth`: released checkpoint from the DIGER RQ-VAE pretraining pipeline.

## Usage

Download with `huggingface_hub`:

```python
from huggingface_hub import hf_hub_download

ckpt_path = hf_hub_download(
    repo_id="junchenfu/diger-rqvae-instruments",
    filename="best_collision_model.pth",
)
```

The checkpoint can be loaded with the RQ-VAE implementation in the DIGER GitHub repository:

https://github.com/junchen-fu/DIGER

Please refer to the repository README for the full training configuration and commands.

## Embeddings

LLaMA embeddings used by DIGER should be generated following the procedure described in:

https://github.com/honghuibao2000/letter

## Dataset Note

The underlying recommendation datasets are publicly available from their original sources. Processed interaction data are not hosted in this model repository; they can be regenerated following the DIGER codebase.

## Citation

If you use this checkpoint, please cite the DIGER paper.