在AI studio别人的公开项目的基础上，改动配置文件来使用自建数据集进行训练，发生‘metaclass conflict‘错误

[X-MingYang]

问题确认 Search before asking

• [X] 我已经搜索过问题，但是没有找到解答。I have searched the question and found no related answer.

请提出你的问题 Please ask your question

原项目链接：https://aistudio.baidu.com/projectdetail/2089732?channelType=0&channel=0 本人数据集按照官方文档进行搭建： https://github.com/PaddlePaddle/PaddleSeg/blob/release/2.9/docs/data/transform/transform_cn.md 对于配置文件只改动了【num_classes】、【dataset_root】、【train_path】、【val_path】。发生报错如图所示 ##本人改动过的ocrnet_dlrsd.yml base: './configs/base/dlrsd.yml'

batch_size: 2 iters: 160000

model: type: OCRNet backbone: type: HRNet_W18 pretrained: https://bj.bcebos.com/paddleseg/dygraph/hrnet_w18_ssld.tar.gz num_classes: 2 backbone_indices: [0]

optimizer: type: sgd

lr_scheduler: type: PolynomialDecay learning_rate: 0.01 power: 0.9

loss: types: - type: CrossEntropyLoss - type: CrossEntropyLoss coef: [1, 0.4]

##本人改动过的dlrsd.yml batch_size: 2 iters: 80000

train_dataset: type: Dataset dataset_root: home/aistudio/data/labeldata/labeldata train_path: home/aistudio/data/labeldata/labeldata/train.txt num_classes: 20 transforms: - type: ResizeStepScaling min_scale_factor: 0.5 max_scale_factor: 2.0 scale_step_size: 0.25 - type: RandomPaddingCrop crop_size: [640, 640] - type: RandomHorizontalFlip - type: Normalize mode: train

val_dataset: type: Dataset dataset_root: home/aistudio/data/labeldata/labeldata val_path: home/aistudio/data/labeldata/labeldata/val.txt num_classes: 20 transforms: - type: Normalize mode: val

optimizer: type: sgd momentum: 0.9 weight_decay: 4.0e-5

lr_scheduler: type: PolynomialDecay learning_rate: 0.01 end_lr: 0 power: 0.9

loss: types: - type: CrossEntropyLoss coef: [1]

原项目中的对应配置文件如下： ##原项目中的ocrnet_dlrsd.yml base: './configs/base/dlrsd.yml'

batch_size: 2 iters: 160000

model: type: OCRNet backbone: type: HRNet_W18 pretrained: https://bj.bcebos.com/paddleseg/dygraph/hrnet_w18_ssld.tar.gz num_classes: 18 backbone_indices: [0]

optimizer: type: sgd

lr_scheduler: type: PolynomialDecay learning_rate: 0.01 power: 0.9

loss: types: - type: CrossEntropyLoss - type: CrossEntropyLoss coef: [1, 0.4]

##原项目中的dlrsd.yml batch_size: 2 iters: 80000

train_dataset: type: Dataset dataset_root: dataset/dlrsd train_path: dataset/dlrsd/train.txt num_classes: 18 transforms: - type: ResizeStepScaling min_scale_factor: 0.5 max_scale_factor: 2.0 scale_step_size: 0.25 - type: RandomPaddingCrop crop_size: [256, 256] - type: RandomHorizontalFlip - type: Normalize mode: train

val_dataset: type: Dataset dataset_root: dataset/dlrsd val_path: dataset/dlrsd/val.txt num_classes: 18 transforms: - type: Normalize - type: Resize target_size: [256, 256] mode: val

optimizer: type: sgd momentum: 0.9 weight_decay: 4.0e-5

lr_scheduler: type: PolynomialDecay learning_rate: 0.01 end_lr: 0 power: 0.9

loss: types: - type: CrossEntropyLoss coef: [1] 万望各位大佬指点一二！

[shiyutang]

你好，请尝试删除"meta=ABCMeta"后查看是否会持续产生这个错误，“meta=ABCMeta”用于指定继承类为抽象类而非实现类，根据代码判断这是一个风格比较混合，在forward中有具体实现的类，因此可以尝试去除继承ABCMeta。