I have a python 3.7 project on windows that is using this detection model . This works great as a stand-alone solution using Tensorflow_gpu 1.15.2 and CUDA 10.0.
My problem is I am also using a commercial joint modeling solution (wrnchAI) which requires CUDA 10.1.
Is anyone aware of a Tensorflow_gpu 1.15.2 build that has been complied against CUDA 10.1?
I should add that when trying to run the Keras-yolo2 code with CUDA 10.1, I get can get past the .dll errors buy adding the CUDA 10.0 to CUDA 10.1 however, I can't get past the following error.
Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize
I have a python 3.7 project on windows that is using this detection model . This works great as a stand-alone solution using Tensorflow_gpu 1.15.2 and CUDA 10.0.
My problem is I am also using a commercial joint modeling solution (wrnchAI) which requires CUDA 10.1.
Is anyone aware of a Tensorflow_gpu 1.15.2 build that has been complied against CUDA 10.1?
I should add that when trying to run the Keras-yolo2 code with CUDA 10.1, I get can get past the .dll errors buy adding the CUDA 10.0 to CUDA 10.1 however, I can't get past the following error.
Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize