之前曾经可以执行，后来有部分升级到2.0
GPU执行就报错，该怎么处理？
Loading model...
W0901 14:20:46.985733 8492 device_context.cc:404] Please NOTE: device: 0, GPU Compute Capability: 7.5, Driver API Version: 11.1, Runtime API Version: 10.2
W0901 14:20:46.996703 8492 device_context.cc:422] device: 0, cuDNN Version: 7.3.
W0901 14:20:48.163267 8492 device_context.h:361] WARNING: device: 0. The installed Paddle is compiled with CUDNN 7.6, but CUDNN version in your machine is 7.3, which may cause serious incompatible bug. Please recompile or reinstall Paddle with compatible CUDNN version.
2021-09-01 14:20:49 [INFO] Model[YOLOv3] loaded.
Model loaded.
。。。。。。
File "site-packages\paddlex\cv\nets\resnet.py", line 443, in c1_stage
File "site-packages\paddlex\cv\nets\resnet.py", line 464, in __call__
File "site-packages\paddlex\cv\nets\detection\yolo_v3.py", line 502, in build_net
File "site-packages\paddlex\cv\models\ppyolo.py", line 171, in build_net
File "site-packages\paddlex\cv\models\load_model.py", line 66, in load_model
File "paddlexui\pms\model_tasks\tasks.py", line 246, in _call_paddlex_export_infer
File "multiprocessing\process.py", line 99, in run
File "multiprocessing\process.py", line 297, in _bootstrap
File "multiprocessing\spawn.py", line 118, in _main
File "multiprocessing\spawn.py", line 105, in spawn_main
File "site-packages\PyInstaller\loader\rthooks\pyi_rth_multiprocessing.py", line 43, in _freeze_support
File "paddlexui\PaddleXCore.py", line 39, in main
File "paddlexui\PaddleX.py", line 18, in
ExternalError: Cudnn error, CUDNN_STATUS_EXECUTION_FAILED(at C:/home/workspace/Paddle_release1/paddle/fluid/operators/conv_cudnn_op.cu:348)
[operator < conv2d > error]
下面这两个接口都可以执行，但报错第二个提示貌似更全，
升级前可以用，但是究竟该怎么处理呢？
model = pdx.deploy.Predictor('inference_model', use_gpu=False) # use_gpu=True 2.97s
或
model = pdx.load_model('inference_model')
确认下，安装的paddle版本的cuda和cudnn版本，和你本地安装cuda和cudnn版本是否相符
paddlex的这个1.X--->2.X 升级 感觉有些乱。。。 我现在不知道 该怎么弄了。。。。
主要是：
1、我电脑上 装了2个版本的 paddlex： 1.7 的 和 2.0 的， 后来1.7 、 2.0 也都分别生成过模型 。。。
paddle版本的cuda和cudnn 版本不一样，
2、本地的 paddlepaddle-gpu 也 pip 升级修改过多次
3、pc本地装的CUDA cudnn 11.4 10.2 9.0 都装了
以前1.7下 还可以用，这一路升级后，一路折腾，有桶了马蜂的感觉。。。
都2.几了，用新的吧
The installed Paddle is compiled with CUDNN 7.6, but CUDNN version in your machine is 7.3, which may cause serious incompatible bug. Please recompile or reinstall Paddle with compatible CUDNN version.
看报错信息，是你的cudnn版本的原因，去安装一个7.6版本的cudnn应该就可以了。