AttributeError: 'ConfigDict' object has no attribute 'pipeline'

[quantumstudent]

您好，非常感谢您做的杰出工作。现在我想将您提供的分割模型应用于我野外无人机拍摄影像的语义分割（地物分类任务中）获取mask，想使用loveda的dg_loveda_rural2urban_1024x1024.py中的模型，并在https://huggingface.co/Cusyoung/CrossEarth/tree/main上下载补全了权重文件R2U.pth。但现在我调用此模型去单独预测某个图像时总是报错“Traceback (most recent call last):
File "G:\PYZ_Image_Matching\CrossEarth-main\CrossEarth_infer.py", line 97, in
segmentor = CrossEarthSegmentor(
File "G:\PYZ_Image_Matching\CrossEarth-main\CrossEarth_infer.py", line 43, in init
self.inferencer = MMSegInferencer(
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmseg\apis\mmseg_inferencer.py", line 81, in init
super().init(
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmengine\infer\infer.py", line 181, in init
self.pipeline = self._init_pipeline(cfg)
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmseg\apis\mmseg_inferencer.py", line 359, in _init_pipeline
pipeline_cfg = cfg.test_dataloader.dataset.pipeline
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmengine\config\config.py", line 113, in getattr
raise AttributeError(f"'{self.class.name}' object has no "
AttributeError: 'ConfigDict' object has no attribute 'pipeline'
”，能请教您一下这是怎么回事，以及我该如何修改代码？
Hello, thank you very much for your outstanding work. Now I want to apply the segmentation model you provided to the semantic segmentation (land cover classification task) of the images taken by my field drone to obtain masks. I want to use the model in the dg_loveda_rural2urban_1024x1024.py of loveda and download the completed weight file R2U.pth from https://huggingface.co/Cusyoung/CrossEarth/tree/main. However, when I call this model to predict a single image, I always get an error: "Traceback (most recent call last): File "G:\PYZ_Image_Matching\CrossEarth-main\CrossEarth_infer.py", line 97, in
segmentor = CrossEarthSegmentor(
File "G:\PYZ_Image_Matching\CrossEarth-main\CrossEarth_infer.py", line 43, in init
self.inferencer = MMSegInferencer(
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmseg\apis\mmseg_inferencer.py", line 81, in init
super().init(
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmengine\infer\infer.py", line 181, in init
self.pipeline = self._init_pipeline(cfg)
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmseg\apis\mmseg_inferencer.py", line 359, in _init_pipeline
pipeline_cfg = cfg.test_dataloader.dataset.pipeline
File "C:\Users\Pmiracle.conda\envs\py39\lib\site-packages\mmengine\config\config.py", line 113, in getattr
raise AttributeError(f"'{self.class.name}' object has no "
AttributeError: 'ConfigDict' object has no attribute 'pipeline'
Could I ask you what's going on here and how I should modify the code?

[quantumstudent]

My code is as follows: `import os
import os.path as osp
ROOT_DIR = osp.abspath(osp.dirname(file))
os.chdir(ROOT_DIR)
import sys
sys.path.append(ROOT_DIR)
from typing import Union, List
import CrossEarth
import rs_dataset
import mmcv
import torch
from mmengine.config import Config
from mmseg.apis import init_model, inference_model, show_result_pyplot, MMSegInferencer

class CrossEarthSegmentor:

def __init__(self,
             config_file: str,
             checkpoint_file: str,
             device: str = "cuda:0"):
    self.config_file = config_file
    self.checkpoint_file = checkpoint_file
    self.device = device

    print(f"Initializing model from config: {config_file}, checkpoint: {checkpoint_file}")

    self.model = init_model(
        config_file,
        checkpoint_file,
        device=device
    )
    self.inferencer = MMSegInferencer(
        model=self.config_file,
        weights=self.checkpoint_file,
        device=self.device
    )

    self.model.eval()

@torch.no_grad()

def predict(self,
            images: Union[str, List[str]],
            save_vis: bool = False,
            vis_out_dir: str = None,
            opacity: float = 0.5):

    if isinstance(images, str):
        images = [images]

    if save_vis and vis_out_dir:
        os.makedirs(vis_out_dir, exist_ok=True)

    results = []

    for img_path in images:

        result = self.inferencer(img_path)

        results.append(result)

        if save_vis and vis_out_dir:
            out_file = os.path.join(vis_out_dir, os.path.basename(img_path))

            self.inferencer.show_result(
                img_path,
                result,
                out_file=out_file,
                opacity=opacity
            )
            print(f"[Saved visualization] {out_file}")

    return results

config_file = "configs/CrossEarth_dinov2/CrossEarth_dinov2_mask2former_1024x1024_bs4x2.py"
checkpoint_file = "./checkpoints/R2U.pth"

segmentor = CrossEarthSegmentor(
config_file=config_file,
checkpoint_file=checkpoint_file,
device="cuda:0"
)

results = segmentor.predict(
"mypath/0001.JPG",
save_vis=True,
vis_out_dir="output_vis"
)

print("Done.")

`