Llama-Factory 微调 Qwen2.5-VL：数据集制作与部署流程

环境配置

1. Ubuntu 24
2. NVIDIA 3090 (24G)
3. CUDA 12.9

一、数据集制作

1. Label-studio 制作数据集

这是从零开始制作数据集的方法。 安装完 label-studio 后，输入指令启动：

label-studio start

进入浏览器界面创建项目（Create Project），引入图片后选择图像描述数据集制作（Image Captioning）。

2. 利用 Qwen2.5-VL 半自动制作数据集

利用 Qwen 的图像描述能力进行预生成，再人工复核修改，可减少人力成本。 脚本示例如下：

import torch
from modelscope import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
import time
import os
from pathlib import Path
import json

def process_single_image(model, processor, image_path, prompt):
    messages = [{"role": "user", "content": [{"type": "image", "image": image_path}, {"type": "text", "text": prompt}]}]
    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")
    time_start = time.time()
    generated_ids = model.generate(**inputs, max_new_tokens=256, do_sample=False)
    time_end = time.time()
    print(f"Inference time for : s")
    generated_ids_trimmed = [out_ids[(in_ids):]  in_ids, out_ids  (inputs.input_ids, generated_ids)]
    output_text = processor.batch_decode(generated_ids_trimmed, skip_special_tokens=, clean_up_tokenization_spaces=)
     output_text[]

 ():
    image_extensions = {, , , , , }
    image_files = []
     file  Path(image_folder).iterdir():
         file.suffix.lower()  image_extensions:
            image_files.append(file)
    image_files.sort()
      image_files:
        ()
        
    ()
    results = []
     image_file  image_files:
        ()
        :
            result = process_single_image(model, processor, (image_file), prompt)
            ()
            results.append({: image_file.name, : (image_file), : result})
         Exception  e:
            ()
            results.append({: image_file.name, : (image_file), : , : })
     output_file:
         (output_file, , encoding=)  f:
             item  results:
                json_line = {: item[], : item[]}
                f.write(json.dumps(json_line, ensure_ascii=) + )
        ()
     results

 __name__ == :
    model = Qwen2_5_VLForConditionalGeneration.from_pretrained(, torch_dtype=, device_map=)
    min_pixels =  *  * 
    max_pixels =  *  * 
    processor = AutoProcessor.from_pretrained(, min_pixels=min_pixels, max_pixels=max_pixels)
    image_folder = 
    prompt = 
    output_file = 
    results = process_images_in_folder(model, processor, image_folder, prompt, output_file)
    ()