说我GPU和软件版本不兼容，那和哪个版本兼容？GTX950M

[e0r]

I1231 23:51:03.401109 12376 analysis_predictor.cc:1660] ======= optimize end =======
I1231 23:51:03.402117 12376 naive_executor.cc:164] --- skip [feed], feed -> x
I1231 23:51:03.405109 12376 naive_executor.cc:164] --- skip [softmax_11.tmp_0], fetch -> fetch
W1231 23:51:03.443147 12376 gpu_resources.cc:96] The GPU architecture in your current machine is Maxwell, which is not compatible with Paddle installation with arch: 70 75 80 86 , it is recommended to install the corresponding wheel package according to the installation information on the official Paddle website.
W1231 23:51:03.443147 12376 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 5.0, Driver API Version: 12.2, Runtime API Version: 11.2
W1231 23:51:03.459131 12376 gpu_resources.cc:149] device: 0, cuDNN Version: 8.2.
predict img: C:\Users\Administrator\Desktop\Screenshot_2023_1111_104306.png
The detection visualized image saved in ./output//Screenshot_2023_1111_104306.png

装了CUDA 11.2.2 CUDNN 8.2.1.32 TensorRT 8.0.3.4
OCR就是github下载的zip，用VS2022编译的C++版
paddle_inference.zip是这个：
https://paddle-inference-lib.bj.bcebos.com/2.5.2/cxx_c/Windows/GPU/x86-64_cuda11.2_cudnn8.2.1_trt8.0.1.6_mkl_avx_vs2019/paddle_inference.zip

[e0r]

CMAKE里面勾了GPU，编译出来还是用CPU，识别一张手机截图要十几秒，我就把工程里的use_gpu全替换成true，结果不兼容了

[e0r]

试试看232版本的paddle，我的问题和你类似，解决了 python -m pip install paddlepaddle-gpu==2.3.2.post112 -f https://www.paddlepaddle.org.cn/whl/windows/mkl/avx/stable.html

好的我试试