Instructions to use hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning", device_map="auto")

PEFT
How to use hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning with PEFT:
```
Task type is invalid.
```
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning

SGLang

How to use hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning with Docker Model Runner:
```
docker model run hf.co/hoangphann/LoRA-Qwen2-VL-2B-Instruct-captioning
```

LoRA-Qwen2-VL-2B-Instruct-captioning / run_config.json

hoangphann

Upload folder using huggingface_hub

a481070 verified 9 months ago

Raw History Blame Contribute Delete

943 Bytes

	{
	"created_at": "2026-01-03 01:00:58",
	"preset": "A",
	"dataset_id": "bbdontcry/vietnamese-image-captioning",
	"model_id": "Qwen/Qwen2-VL-2B-Instruct",
	"quantization": null,
	"min_pixels": 200704,
	"max_pixels": 802816,
	"system_message": "Bạn là một Vision-Language Model hỗ trợ người khiếm thị bằng cách tạo mô tả ảnh bằng tiếng Việt cho trang sách giáo khoa. Chỉ mô tả những gì thấy trong ảnh, không suy đoán.Khi OCR, trích toàn bộ chữ nhìn thấy, theo thứ tự trên xuống dưới, trái sang phải",
	"prompt_detail": "Hãy thuyết minh chi tiết trang SGK trong ảnh theo luồng đọc từ trên xuống dưới. Khi gặp chữ trong ảnh, hãy đưa vào đúng ngữ cảnh và trích nguyên văn trong ngoặc kép. Không suy luận.",
	"gen": {
	"max_new_tokens_detail": 256,
	"temperature": 0.2,
	"top_p": 0.9,
	"do_sample": false
	}
	}