Instructions to use XiangyunHuang/Route2Step with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use XiangyunHuang/Route2Step with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("XiangyunHuang/Route2Step", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 1,085 Bytes
c645b86 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 | ---
library_name: transformers
tags:
- vision-language-navigation
- qwen3-vl
- safetensors
---
# Route2Step
This repository contains the released Route2Step checkpoints.
## Qwen3-VL-4B release
`MAG-Qwen3-vl-4b/` is the Qwen3-VL-4B MAG checkpoint for local action execution. In our reported evaluation, it reaches an SR of 62.
The folder contains the inference-required model shards, configuration, tokenizer, and image/video preprocessing files. Training-only files are omitted.
## Available checkpoints
- `MIA/`: released MIA checkpoint.
- `MAG/`: released MAG checkpoint.
- `MAG-Qwen3-vl-4b/`: Qwen3-VL-4B MAG release.
Load the Qwen3-VL-4B checkpoint with a recent `transformers` version supporting Qwen3-VL:
```python
from transformers import AutoProcessor, Qwen3VLForConditionalGeneration
model_id = "XiangyunHuang/Route2Step"
model = Qwen3VLForConditionalGeneration.from_pretrained(
model_id,
subfolder="MAG-Qwen3-vl-4b",
torch_dtype="auto",
device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id, subfolder="MAG-Qwen3-vl-4b")
```
|