Paddleocr3.0是否可以用cpu运行？

[Sakura9421]

我是刚入门的小白，paddlepaddle安装的为3.0的cpu版本，pip安装完3.0的paddleocr后，运行paddleocr报错ValueError: invalid literal for int() with base 10: "NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running. This can also be happening if non-NVIDIA GPU is running，想问问大家是否是因为新版本的原因无法使用cpu进行ocr识别呢？

[liuhongen1234567]

您好，paddleocr3.0是可以在CPU上使用的，可以给出具体的报错信息和复现命令吗？这个报错看上去像是您的cuda、cudnn和驱动不匹配。

[Sakura9421]

您好！我的代码是：
from paddleocr import PaddleOCR

创建PaddleOCR实例，指定使用GPU

ocr = PaddleOCR(
device="cpu", # 指定使用GPU
use_doc_orientation_classify=False, # 不使用文档方向分类模型
use_doc_unwarping=False, # 不使用文本图像矫正模型
use_textline_orientation=False, # 不使用文本行方向分类模型
)

指定输入图像路径

image_path = "general_ocr_002.png"

执行OCR检测

result = ocr.predict(image_path)

处理结果

for res in result:
print(res)
res.save_to_img("output") # 保存结果到图像
res.save_to_json("output") # 保存结果到JSON文件

我的电脑是核显没有英伟达显卡，他的报错信息是：
Traceback (most recent call last):
File "d:\paddleOCR3-test\test.py", line 4, in
ocr = PaddleOCR(
^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddleocr_pipelines\ocr.py", line 144, in init
super().init(**base_params)
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddleocr_pipelines\base.py", line 73, in init
self.paddlex_pipeline = self._create_paddlex_pipeline()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddleocr_pipelines\base.py", line 106, in _create_paddlex_pipeline
kwargs = prepare_common_init_args(None, self._common_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddleocr_common_args.py", line 68, in prepare_common_init_args
pp_option = PaddlePredictorOption(model_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddlex\inference\utils\pp_option.py", line 52, in init
self._init_option(**kwargs)
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddlex\inference\utils\pp_option.py", line 87, in _init_option
for k, v in self._get_default_config().items():
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddlex\inference\utils\pp_option.py", line 103, in _get_default_config
device_type, device_ids = parse_device(get_default_device())
^^^^^^^^^^^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddlex\utils\device.py", line 44, in get_default_device
avail_gpus = GPUtil.getAvailable()
^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\GPUtil\GPUtil.py", line 140, in getAvailable
GPUs = getGPUs()
^^^^^^^^^
File "C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\GPUtil\GPUtil.py", line 102, in getGPUs
deviceIds = int(vals[i])
^^^^^^^^^^^^
ValueError: invalid literal for int() with base 10: "NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running. This can also be happening if non-NVIDIA GPU is running
谢谢您，劳驾您看下问题出在哪了

[liuhongen1234567]

您好，这确实这版代码对于cpu环境处理的bug，不过wheel包已经打好了，只能下一版再修了。可以cd 到 这个文件里"C:\Users\15570\anaconda3\envs\paddle33\Lib\site-packages\paddlex\inference\utils\pp_option.py", 找到 line 103 里
device_type, device_ids = parse_device(get_default_device()) 这行代码，把它注释掉，然后替换为 device_type, device_ids ="cpu", None, 具体如图所示，

[Sakura9421]

谢谢您，确实有效！