---
base_model: unsloth/Qwen2-VL-2B-Instruct-unsloth-bnb-4bit
library_name: peft
---

# Qwen2-VL-2B-RobotArm, Leveraging [Qwen/Qwen2-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct) in Robotic Field

### Model Description
This project focuses on fine-tuning visual language models to predict the position and orientation of a robotic arm for reaching a target object based on a human command and an image from a camera view. The system interprets natural language instructions and visual context to generate actionable pose outputs.

<p align="center">
    <img src="Robot_Arm_Diagram.png" width="95%"/>
<p>
  
## Usage
Inference with unsloth

```python
from unsloth import FastVisionModel
from PIL import Image
from transformers import TextStreamer
model, processor = FastVisionModel.from_pretrained(
        'SakalYin\Qwen2-VL-2B-RobotArm-Camera',
        load_in_4bit=True,   
        local_files_only=True,
        attn_implementation="flash_attention_2",
    )
def resize_image(image_input, size=None):
    """Load image from path or matrices and resize them"""
    size = size if size else (854,480)
    if isinstance(image_input, str):  
        image = Image.open(image_input)
    else:  
        image = image_input
    size = size if size else (854,480)
    image = image.resize(size).convert('RGB') 
    return image
image_path = "https://i.imgur.com/vAleq1e.png"
size = (854, 480) # Recommeded Size
system_message = "You are a Visual Language Model Trained to output robot arm end-effectors parameters. Base on the user requests, locate the appropriate object in the image and you must return the position and orientation to reach it in xml format."
prompt = "Give me a cup.<camera>0.5,-0.75,1.8</camera>"
image = resize_image(image_path, size=size)
messages = [
    {
        "role": "system",
        "content": [{"type": "text", "text": system_message if system_message else "You are a helpful assistant."}],
    },
    {"role": "user", "content": [
        {"type": "image"},
        {"type": "text", "text": prompt}]
    }
]
input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
inputs = processor(
    image,
    input_text,
    add_special_tokens=False,
    return_tensors="pt",
).to("cuda")
# Generate text without using the streamer
text_streamer = TextStreamer(processor, skip_prompt=True)
generated_tokens = model.generate(
    **inputs,
    streamer=text_streamer,  
    max_new_tokens=128, 
    use_cache=True, 
    temperature=1.5, 
    min_p=0.1
)
# Decode output properly
output_text = processor.decode(generated_tokens[0], skip_special_tokens=True).strip()
```

**Output**: \<obj\>plastic_cup\<\/obj\> \<pose\>0.625,-0.785,0.548\<\/pose\> \<orient\>0.35478,0.71881,0.30505,0.5141\<\/orient\>\<|im_end|\>

### Results
Result compared to other model in the [Collection](https://huggingface.co/collections/SakalYin/robotarm-eff-single-action-681daa76a9286b53a51b401d)

<table>
<thead>
  <tr>
    <th>Model</th>
    <th>Euclidean</th>
    <th>X MSE</th>
    <th>Y MSE</th>
    <th>Z MSE</th>
    <th>Roll Error</th>
    <th>Pitch Error</th>
    <th>Yaw Error</th>
  </tr>
</thead>
<tbody align="center">
  <tr>
    <td>Qwen2VL-2B  <sup><sup>(Trained with Camera Location)</sup></sup></td>
    <td>0.1089</td>
    <td>0.0505</td>
    <td>0.0575</td>
    <td>0.0363</td>
    <td>6.8334</td>
    <td>5.4204</td>
    <td>6.7619</td>
  </tr>
  <tr>
    <td>Qwen2VL-2B</td>
    <td>0.3865</td>
    <td>0.1239</sup></sup></td>
    <td>0.3411</td>
    <td>0.0000</td>
    <td>2.1462</td>
    <td>0.9029</td>
    <td>1.1926</td>
  </tr>
  <tr>
    <td>Qwen2VL-7B</td>
    <td>0.0509</td>
    <td>0.0305</td>
    <td>0.0320</td>
    <td>0.0008</td>
    <td>0.4148</td>
    <td>0.1066</td>
    <td>0.1734</td>
  </tr>
  <tr>
    <td>LlaVA-NeXT 7B</td>
    <td>0.0480</td>
    <td>0.0306</td>
    <td>0.0296</td>
    <td>0.0000</td>
    <td>0.0000</td>
    <td>0.0000</td>
    <td>0.0000</td>
  </tr>
  <tr>
    <td>Llama Vision 11B</td>
    <td>-</td>
    <td>-</td>
    <td>-</td>
    <td>-</td>
    <td>-</td>
    <td>-</td>
    <td>-</td>
  </tr>
</tbody>
</table>