vame.train_model() GPU was not found...proceeding with CPU

[lpereiras17]

I am using two NVIDIA GeForce GTX 1080 Ti GPU's for network training. I was able to sucessfully train and anaylze a DeepLabCut network using CUDA version 10.1.105 and NVIDIA studio studio driver version 552.22 for Windows 10 64-bit (DeepLabCut recognized GPUs and was able to train without relying on CPU). When I run vame.train_model(config) I get a warning message of "a GPU was not found...processing with CPU (slow!)".

I made sure to 1) run in VAME enviroment: conda install cudnn=8 (in case it wasn't finding cudart64.110.dll file) and 2) conda install -c conda-forge cudnn==8.4.1.50 cudatoolkit==11.7.0 (to solve the same problem as before). I had a problem with DLC identifying my GPU and that seemed to solve it but it's not working for VAME. Do you have any advice or know why this is happening?

Also! This is something else but even though I have institutional access to VAME's Nature paper through Michigan State University, I'm unable to see the suppplementary figures attached (only able ot see supplementary videos attached). For example I can not see supplementary Figure 12 only supplementary movie 12. Do you know why this is so?