Device id must be less than GPU count, but received id is: 3. GPU count is: 1

[xndm-tech]

🔎 Search before asking

• I have searched the PaddleOCR Docs and found no similar bug report.
• I have searched the PaddleOCR Issues and found no similar bug report.
• I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

Traceback (most recent call last):
File "/home/xxxxxxx/subtitle/video-subtitle-extractor/backend/test_device_id_bug.py", line 4, in
ocr = PaddleOCR(#device = 'GPU',
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddleocr/_pipelines/ocr.py", line 144, in init
super().init(**base_params)
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddleocr/_pipelines/base.py", line 73, in init
self.paddlex_pipeline = self._create_paddlex_pipeline()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddleocr/_pipelines/base.py", line 108, in _create_paddlex_pipeline
return create_pipeline(config=self._merged_paddlex_config, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/pipelines/init.py", line 165, in create_pipeline
pipeline = BasePipeline.get(pipeline_name)(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/utils/deps.py", line 195, in _wrapper
return old_init_func(self, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/pipelines/_parallel.py", line 103, in init
self._pipeline = self._create_internal_pipeline(config, self.device)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/pipelines/_parallel.py", line 158, in _create_internal_pipeline
return self._pipeline_cls(
^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/pipelines/ocr/pipeline.py", line 114, in init
self.text_det_model = self.create_model(
^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/pipelines/base.py", line 99, in create_model
model = create_predictor(
^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/models/init.py", line 77, in create_predictor
return BasePredictor.get(model_name)(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/models/text_detection/predictor.py", line 57, in init
self.pre_tfs, self.infer, self.post_op = self._build()
^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/models/text_detection/predictor.py", line 77, in _build
infer = self.create_static_infer()
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/models/base/predictor/base_predictor.py", line 240, in create_static_infer
return PaddleInfer(self.model_dir, self.MODEL_FILE_PREFIX, self._pp_option)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/models/common/static_infer.py", line 274, in init
self.predictor = self._create()
^^^^^^^^^^^^^^
File "/opt/anaconda3/envs/subtitle/lib/python3.12/site-packages/paddlex/inference/models/common/static_infer.py", line 437, in _create
predictor = paddle.inference.create_predictor(config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: (InvalidArgument) Device id must be less than GPU count, but received id is: 3. GPU count is: 1.
[Hint: Expected id < GetGPUDeviceCount(), but received id:3 >= GetGPUDeviceCount():1.] (at ../paddle/phi/backends/gpu/cuda/cuda_info.cc:248)

问题还是源于GPUtil，但修复后需要确认下面的程序是否能跑通。

🏃‍♂️ Environment (运行环境)

python代码，如果文件名为test.py,则运行CUDA_VISIBLE_DEVICES=5 python backend/test_device_id_bug.py 会报错
from paddleocr import PaddleOCR
import cv2
ocr = PaddleOCR(#device = 'GPU',
# 设置文本检测模型路径
text_detection_model_name='PP-OCRv5_server_det',
# 设置每张图文本框批处理数量
text_recognition_batch_size=6,
# 设置文本识别模型路径
text_recognition_model_name='PP-OCRv5_server_rec',
use_doc_orientation_classify = False,
use_doc_unwarping = False,
use_textline_orientation = False,
text_det_unclip_ratio=1.2,
text_det_limit_side_len=1080,
text_det_limit_type='max',
text_rec_score_thresh = 0.5,
#lang='ch'
)

读取图片

img_path = "test_src.jpg"
image = cv2.imread(img_path)

执行 OCR 识别

result = ocr.predict(image)

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

可参考上面内容
这里面涉及几个点：1、运行时设置CUDA_VISIBLE_DEVICES=5，也可以设置为其他值，对于8卡机，小于8大于0即可 ，2、报错一律为received id is: 3， 即这里拿到的id跟机器本身GPU的显存状态有关，3. GPU count is：1