PP-StructureV3产线CPU推理异常(InvalidArgument) The size of inputs must be equal to 2.

[hipitt]

Checklist:

• 查找历史相关issue寻求解答
• 翻阅FAQ
• 翻阅PaddleX 文档
• 确认bug是否在新版本里还未修复

描述问题

执行测试代码：

from paddlex import create_pipeline

pipeline = create_pipeline(pipeline="PP-StructureV3")

output = pipeline.predict(
    "f:/22.png",
    use_doc_orientation_classify=False,
    use_doc_unwarping=False,
    use_common_ocr=True,
    use_seal_recognition=True,
    use_table_recognition=True,
)

for res in output:
    res.print()
    res.save_to_img("./output")
    res.save_to_json("./output")
    res.save_to_xlsx("./output")
    res.save_to_html("./output")
    res.save_to_markdown("./output")

异常：
Traceback (most recent call last):
File "F:\code\PaddleX\api_examples\test1.py", line 3, in
pipeline = create_pipeline(pipeline="PP-StructureV3")
File "F:\code\PaddleX\paddlex\inference\pipelines_init_.py", line 155, in create_pipeline
pipeline = BasePipeline.get(pipeline_name)(
File "F:\code\PaddleX\paddlex\inference\pipelines\layout_parsing\pipeline_v2.py", line 59, in init
self.inintial_predictor(config)
File "F:\code\PaddleX\paddlex\inference\pipelines\layout_parsing\pipeline_v2.py", line 154, in inintial_predictor
self.formula_recognition_pipeline = self.create_pipeline(
File "F:\code\PaddleX\paddlex\inference\pipelines\base.py", line 113, in create_pipeline
pipeline = create_pipeline(
File "F:\code\PaddleX\paddlex\inference\pipelines_init_.py", line 155, in create_pipeline
pipeline = BasePipeline.get(pipeline_name)(
File "F:\code\PaddleX\paddlex\inference\pipelines\formula_recognition\pipeline.py", line 103, in init
self.formula_recognition_model = self.create_model(formula_recognition_config)
File "F:\code\PaddleX\paddlex\inference\pipelines\base.py", line 86, in create_model
model = create_predictor(
File "F:\code\PaddleX\paddlex\inference\models_init_.py", line 112, in create_predictor
return BasicPredictor.get(model_name)(
File "F:\code\PaddleX\paddlex\inference\models\formula_recognition\predictor.py", line 67, in init
self.pre_tfs, self.infer, self.post_op = self._build()
File "F:\code\PaddleX\paddlex\inference\models\formula_recognition\predictor.py", line 96, in _build
infer = StaticInfer(
File "F:\code\PaddleX\paddlex\inference\models\common\static_infer.py", line 125, in init
self._update_option(option)
File "F:\code\PaddleX\paddlex\inference\models\common\static_infer.py", line 131, in _update_option
self._reset()
File "F:\code\PaddleX\paddlex\inference\models\common\static_infer.py", line 148, in _reset
) = self._create()
File "F:\code\PaddleX\paddlex\inference\models\common\static_infer.py", line 299, in _create
predictor = create_predictor(config)
ValueError: (InvalidArgument) The size of inputs must be equal to 2.
[Hint: Expected input_size == 2, but received input_size:1 != 2:2.] (at paddle\fluid\pir\dialect\operator\ir\pd_op3.cc:13372)

Process finished with exit code 1

环境

1. 请提供您使用的PaddlePaddle和PaddleX的版本号
   3.0-RC
2. 请提供您使用的操作系统信息，如Linux/Windows/MacOS
   windows 11
3. 请问您使用的Python版本是？
   3.9
4. 请问您使用的CUDA/cuDNN的版本号是？
   CPU推理

[Albus-Tim-Brown]

使用gpu版本推理时也会遇到同样的ValueError: (InvalidArgument) The size of inputs must be equal to 2.
[Hint: Expected input_size == 2, but received input_size:1 != 2:2.] (at paddle\fluid\pir\dialect\operator\ir\pd_op3.cc:13372)
PaddlePaddle为3.0.0，PaddleX为3.0rc，使用Windows 11，Python版本为3.9，CUDA版本为12.6，cudnn版本为9.8.0

[changdazhou]

建议尝试使用11.8的CUDA和配套的PaddlePaddle哈

[Albus-Tim-Brown]

建议尝试使用11.8的CUDA和配套的PaddlePaddle哈

但他描述的问题中，貌似CPU推理也有问题。此外我也去试了一下部署CPU版本的PaddlePaddle，推理时也遇到这个ValueError: (InvalidArgument) The size of inputs must be equal to 2.
[Hint: Expected input_size == 2, but received input_size:1 != 2:2.] (at paddle\fluid\pir\dialect\operator\ir\pd_op3.cc:13372)

[changdazhou]

辛苦把联系方式发送到我的邮箱哈：changda1650@163.com

[metaimagine]

楼主有解决么，gpu也有类似问题