GPU not working. Does the docker image still contain the correct files?

Hi Bill, I'm trying to use deeplasmid on a set of K. pneumo genomes. I would like to use the GPU to make my predictions; however the script always uses 16 CPUs. This is happening on two separate workstations, equipped with different CPUs, GPUs and OS version. I checked the GPU driver installation on both machines and tested the GPU capability of your docker image (billandreo/deeplasmid.tf.gpu2) by running a very simple tensorflow command:

docker run --gpus all -it --rm billandreo/deeplasmid.tf.gpu2 python -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"

The command above uses GPU (I checked using nvidia-smi) and double checked the tensorflow version with:

docker run --gpus all -it --rm billandreo/deeplasmid.tf.gpu2 python -c "import tensorflow as tf; print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU')))"

I also double-checked the GPU usage by running nvidia-smi in the same container as the deeplasmid analysis, it was still stuck at 0%, while 16 CPUs were in use.

Last check: analyzing 1 genome with the tf.gpu2 image takes the same time as with the cpu-only image.

I am not good at digging in someone else's code. However, I tried to find the line where any of the scripts tries to call gpus and opts for cpus. I was unable to find any line containing the term "gpu" or "nvidia". Also, only one line imports tensorflow, but it's commented out.

docker run -it billandreo/deeplasmid.tf.gpu2 grep "tensorflow" DL_Model.py

this results in: #import tensorflow as tf

What am I doing wrong? Am I using the correct image? Was the gpu2 image modified?

Thanks in advance

Stefano

wandreopoulos / deeplasmid

GPU not working. Does the docker image still contain the correct files? #11