paddleocr 3版本使用docker部署时，显示Successfully installed ocr_system-1.0.0后报错ValueError: (InvalidArgument) Type of attribute: round_type is not pir::Int32Attribute. [Hint: Expected attributes.at("round_type").isa<pir::Int32Attribute>() == true, but received attributes.at("round_type").isa<pir::Int32Attribute>():0 != true:1.] (at /paddle/build/paddle/fluid/pir/dialect/operator/ir/pd_op2.cc:9057)

[zxyFalcon]

是版本不兼容问题吗？应该修改哪里？

[zxyFalcon]

完整报错如下：
[2025-07-08 02:58:41,602] [ INFO] - Successfully installed ocr_system-1.0.0
[2025-07-08 02:58:44,962] [ WARNING] - The _initialize method in HubModule will soon be deprecated, you can use the init() to handle the initialization of the object
Traceback (most recent call last):
File "/usr/local/bin/hub", line 8, in
sys.exit(execute())
File "/usr/local/lib/python3.10/dist-packages/paddlehub/commands/utils.py", line 84, in execute
status = 0 if com'_entry'.execute(sys.argv[idx:]) else 1
File "/usr/local/lib/python3.10/dist-packages/paddlehub/commands/serving.py", line 352, in execute
self.start_serving()
File "/usr/local/lib/python3.10/dist-packages/paddlehub/commands/serving.py", line 245, in start_serving
self.start_single_app_with_args()
File "/usr/local/lib/python3.10/dist-packages/paddlehub/commands/serving.py", line 221, in start_single_app_with_args
self.preinstall_modules()
File "/usr/local/lib/python3.10/dist-packages/paddlehub/commands/serving.py", line 152, in preinstall_modules
m = hub.Module(**init_args)
File "/usr/local/lib/python3.10/dist-packages/paddlehub/module/module.py", line 393, in new
module = cls.init_with_name(
File "/usr/local/lib/python3.10/dist-packages/paddlehub/module/module.py", line 509, in init_with_name
user_module._initialize(**kwargs)
File "/usr/local/lib/python3.10/dist-packages/paddlehub/compat/paddle_utils.py", line 221, in runner
return func(*args, **kwargs)
File "/root/.paddlehub/modules/ocr_system/module.py", line 69, in _initialize
self.text_sys = TextSystem(cfg)
File "/PaddleOCR/./tools/infer/predict_system.py", line 58, in init
self.text_classifier = predict_cls.TextClassifier(args)
File "/PaddleOCR/./tools/infer/predict_cls.py", line 61, in init
) = utility.create_predictor(args, "cls", logger)
File "/PaddleOCR/./tools/infer/utility.py", line 416, in create_predictor
predictor = inference.create_predictor(config)
ValueError: (InvalidArgument) Type of attribute: round_type is not pir::Int32Attribute.
[Hint: Expected attributes.at("round_type").isapir::Int32Attribute() == true, but received attributes.at("round_type").isapir::Int32Attribute():0 != true:1.] (at /paddle/build/paddle/fluid/pir/dialect/operator/ir/pd_op2.cc:9057)

[zxyFalcon]

您好，可以提供下具体的文档和复现命令吗？paddle3.0现在采用的是新的推理架构，已经不走 paddlehub 了。初步看是导出的静态图模型和paddle版本不一致导致的

是使用的/deploy/docker/hubserving/cpu下的dockerfile，按/deploy/docker/hubserving下的readme中的步骤执行的，把里面用的版本和模型改了一下

FROM ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddle:3.0.0

RUN pip install --upgrade pip -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple/

RUN mkdir -p /PaddleOCR

COPY . /PaddleOCR

WORKDIR /PaddleOCR

RUN pip install -r requirements.txt -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple/

RUN mkdir -p /PaddleOCR/inference/

ADD https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/PP-OCRv5_mobile_det_infer.tar /PaddleOCR/inference/
RUN tar xf /PaddleOCR/inference/PP-OCRv5_mobile_det_infer.tar -C /PaddleOCR/inference/

ADD https://paddleocr.bj.bcebos.com/dygraph_v2.0/slim/ch_ppocr_mobile_v2.0_cls_slim_infer.tar /PaddleOCR/inference/
RUN tar xf /PaddleOCR/inference/ch_ppocr_mobile_v2.0_cls_slim_infer.tar -C /PaddleOCR/inference/

ADD https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0//PP-OCRv5_mobile_rec_infer.tar /PaddleOCR/inference/
RUN tar xf /PaddleOCR/inference/PP-OCRv5_mobile_rec_infer.tar -C /PaddleOCR/inference/

EXPOSE 8868

CMD ["/bin/bash","-c","hub install deploy/hubserving/ocr_system/ && hub serving start -m ocr_system"]

[liuhongen1234567]

您好，我怀疑是 https://paddleocr.bj.bcebos.com/dygraph_v2.0/slim/ch_ppocr_mobile_v2.0_cls_slim_infer.tar 这个模型太老了，已经无法兼容paddle3.0的问题，排查需要一定时间，可以先试下 paddleocr3.0 推荐的部署方式 https://paddlepaddle.github.io/PaddleOCR/latest/version3.x/deployment/serving.html#1

[zxyFalcon]

好的，谢谢您

[liuhongen1234567]

您好，确实是文本行分类模型太老，需要更换了，可以执行下面的命令下载最新的文本行模型

wget https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/PP-LCNet_x0_25_textline_ori_infer.tar
tar -xf PP-LCNet_x0_25_textline_ori_infer.tar -C /paddle/PaddleOCR/inference

然后把

PaddleOCR/deploy/hubserving/ocr_system/params.py

Line 58 in b34d8fa

cfg.cls_model_dir = "./inference/ch_ppocr_mobile_v2.0_cls_infer/"

这个地方的 cfg.cls_model_dir 改成新下载模型的路径
比如：

    cfg.use_angle_cls = True
    cfg.cls_model_dir = "./inference/PP-LCNet_x0_25_textline_ori_infer/"
    cfg.cls_image_shape = "3, 48, 192"
    cfg.label_list = ["0", "180"]
    cfg.cls_batch_num = 30

最后，到

PaddleOCR/tools/infer/predict_cls.py

Line 43 in b34d8fa

if model_name:

注释掉43到46行，应该就可以正常启动了。

class TextClassifier(object):
    def __init__(self, args):
        if os.path.exists(f"{args.cls_model_dir}/inference.yml"):
            model_config = utility.load_config(f"{args.cls_model_dir}/inference.yml")
            model_name = model_config.get("Global", {}).get("model_name", "")
            # if model_name:
            #     raise ValueError(
            #         f"{model_name} is not supported. Please check if the model is supported by the PaddleOCR wheel."
            #     )

[zxyFalcon]

感谢，太详细了，辛苦您了