Hi, I also tried some combinations(^^;). So far I can construct my environment with the following:
remove tensorflow-gpu, cudatoolkit and cudnn from environment.yml
construct new environment with the edited environment.yml
install tensorflow-gpu 1.12.0 with conda command (this can install relevant cudatoolkit and cudnn together)
Hope this can help. :-)
As you may know, please note that relevant version of cudatoolkit and cudnn would also depend on your GPU. I think GeForece GTX1070/1080, RTX 2060 super should be OK with my solution.
Can someone point me to the right combination of keras, tensorflow and python version that this current code will work on.
Any combination of the above mentioned packages seem to throw an error when I run kitti_evaluate.py
The environment.yml file did not work for me.
Thanks