---
license: apache-2.0
base_model: Qwen/Qwen2.5-Omni-3B
tags:
  - qwen2.5-omni
  - referring-segmentation
  - audio-visual
  - video-segmentation
  - grpo
  - multimodal
language:
  - en
pipeline_tag: image-text-to-text
library_name: transformers
---

# AVSQwen-Omni-3B

GRPO post-trained Qwen2.5-Omni-3B variant used for small-model ablations in the AVSQwen release (`run_abla/*_3B.sh`). Same training recipe as the 7B flagship, trained 4200 steps.

- **Base model**: [Qwen/Qwen2.5-Omni-3B](https://huggingface.co/Qwen/Qwen2.5-Omni-3B)
- **Parameters**: 3B
- **Training checkpoint**: `checkpoint-4200`
- **Paper / code**: https://github.com/<your-org>/AVSQwen

## Usage

```python
from transformers import Qwen2_5OmniForConditionalGeneration, Qwen2_5OmniProcessor

model = Qwen2_5OmniForConditionalGeneration.from_pretrained(
    "Vegetabot/AVSQwen-Omni-3B",
    torch_dtype="auto",
    device_map="auto",
)
processor = Qwen2_5OmniProcessor.from_pretrained("Vegetabot/AVSQwen-Omni-3B")
```

For the full inference pipeline (frame selector + grounding + SAM2 segmenter),
please refer to `inference/` and `run/*.sh` in the release repo.

## Training

- Framework: [ms-swift](https://github.com/modelscope/ms-swift) with GRPO
- Rewards: `bbox_format_reward`, `sam_keyframe_reward`, `minimal_efficiency_reward`
- Data: `swift_train_bbox_grpo_balanced_full_v4_sam.jsonl` (OmniAVS + 7 referring datasets, balanced mix)
- See `training/train_qwen25omni_3B_full_grpo_FINAL_V2.sh` in the release repo.

## Citation

```bibtex
@article{avsqwen2026,
  title   = {AVSQwen: ...},
  author  = {...},
  year    = {2026}
}
```