130 lines
4.8 KiB
Markdown
130 lines
4.8 KiB
Markdown
---
|
|
license: apache-2.0
|
|
language:
|
|
- en
|
|
base_model:
|
|
- Qwen/Qwen2-VL-2B-Instruct
|
|
tags:
|
|
- remote-sensing
|
|
datasets:
|
|
- AdaptLLM/remote-sensing-visual-instructions
|
|
---
|
|
# Adapting Multimodal Large Language Models to Domains via Post-Training (EMNLP 2025)
|
|
|
|
This repos contains the **remote sensing MLLM developed from Qwen-2-VL-2B-Instruct** in our paper: [On Domain-Specific Post-Training for Multimodal Large Language Models](https://huggingface.co/papers/2411.19930). The correspoding training dataset is in [remote-sensing-visual-instructions](https://huggingface.co/datasets/AdaptLLM/remote-sensing-visual-instructions).
|
|
|
|
The main project page is: [Adapt-MLLM-to-Domains](https://huggingface.co/AdaptLLM/Adapt-MLLM-to-Domains)
|
|
|
|
## 1. To Chat with AdaMLLM
|
|
|
|
Our model architecture aligns with the base model: Qwen-2-VL-Instruct. We provide a usage example below, and you may refer to the official [Qwen-2-VL-Instruct repository](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct) for more advanced usage instructions.
|
|
|
|
**Note:** For AdaMLLM, always place the image at the beginning of the input instruction in the messages.
|
|
|
|
<details>
|
|
<summary> Click to expand </summary>
|
|
|
|
1. Set up
|
|
```bash
|
|
pip install qwen-vl-utils
|
|
```
|
|
2. Inference
|
|
```python
|
|
from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoProcessor
|
|
from qwen_vl_utils import process_vision_info
|
|
|
|
# default: Load the model on the available device(s)
|
|
model = Qwen2VLForConditionalGeneration.from_pretrained(
|
|
"AdaptLLM/food-Qwen2-VL-2B-Instruct", torch_dtype="auto", device_map="auto"
|
|
)
|
|
|
|
# We recommend enabling flash_attention_2 for better acceleration and memory saving, especially in multi-image and video scenarios.
|
|
# model = Qwen2VLForConditionalGeneration.from_pretrained(
|
|
# "AdaptLLM/food-Qwen2-VL-2B-Instruct",
|
|
# torch_dtype=torch.bfloat16,
|
|
# attn_implementation="flash_attention_2",
|
|
# device_map="auto",
|
|
# )
|
|
|
|
# default processer
|
|
processor = AutoProcessor.from_pretrained("AdaptLLM/remote-sensing-Qwen2-VL-2B-Instruct")
|
|
|
|
# The default range for the number of visual tokens per image in the model is 4-16384. You can set min_pixels and max_pixels according to your needs, such as a token count range of 256-1280, to balance speed and memory usage.
|
|
# min_pixels = 256*28*28
|
|
# max_pixels = 1280*28*28
|
|
# processor = AutoProcessor.from_pretrained("AdaptLLM/remote-sensing-Qwen2-VL-2B-Instruct", min_pixels=min_pixels, max_pixels=max_pixels)
|
|
|
|
# NOTE: For AdaMLLM, always place the image at the beginning of the input instruction in the messages.
|
|
messages = [
|
|
{
|
|
"role": "user",
|
|
"content": [
|
|
{
|
|
"type": "image",
|
|
"image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
|
|
},
|
|
{"type": "text", "text": "Describe this image."},
|
|
],
|
|
}
|
|
]
|
|
|
|
# Preparation for inference
|
|
text = processor.apply_chat_template(
|
|
messages, tokenize=False, add_generation_prompt=True
|
|
)
|
|
image_inputs, video_inputs = process_vision_info(messages)
|
|
inputs = processor(
|
|
text=[text],
|
|
images=image_inputs,
|
|
videos=video_inputs,
|
|
padding=True,
|
|
return_tensors="pt",
|
|
)
|
|
inputs = inputs.to("cuda")
|
|
|
|
# Inference: Generation of the output
|
|
generated_ids = model.generate(**inputs, max_new_tokens=128)
|
|
generated_ids_trimmed = [
|
|
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
|
|
]
|
|
output_text = processor.batch_decode(
|
|
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
|
|
)
|
|
print(output_text)
|
|
```
|
|
|
|
</details>
|
|
|
|
## 2. To Evaluate Any MLLM on Domain-Specific Benchmarks
|
|
|
|
Refer to the [remote-sensing-VQA-benchmark](https://huggingface.co/datasets/AdaptLLM/remote-sensing-VQA-benchmark) to reproduce our results and evaluate many other MLLMs on domain-specific benchmarks.
|
|
|
|
## 3. To Reproduce this Domain-Adapted MLLM
|
|
|
|
See [Post-Train Guide](https://github.com/bigai-ai/QA-Synthesizer/blob/main/docs/Post_Train.md) to adapt MLLMs to domains.
|
|
|
|
|
|
## Citation
|
|
If you find our work helpful, please cite us.
|
|
|
|
[Adapt MLLM to Domains](https://huggingface.co/papers/2411.19930) (EMNLP 2025 Findings)
|
|
```bibtex
|
|
@article{adamllm,
|
|
title={On Domain-Adaptive Post-Training for Multimodal Large Language Models},
|
|
author={Cheng, Daixuan and Huang, Shaohan and Zhu, Ziyu and Zhang, Xintong and Zhao, Wayne Xin and Luan, Zhongzhi and Dai, Bo and Zhang, Zhenliang},
|
|
journal={arXiv preprint arXiv:2411.19930},
|
|
year={2024}
|
|
}
|
|
```
|
|
|
|
[Adapt LLM to Domains](https://huggingface.co/papers/2309.09530) (ICLR 2024)
|
|
```bibtex
|
|
@inproceedings{
|
|
cheng2024adapting,
|
|
title={Adapting Large Language Models via Reading Comprehension},
|
|
author={Daixuan Cheng and Shaohan Huang and Furu Wei},
|
|
booktitle={The Twelfth International Conference on Learning Representations},
|
|
year={2024},
|
|
url={https://openreview.net/forum?id=y886UXPEZ0}
|
|
}
|
|
``` |