NVIDIA and CUDA errors while running the code on CPU

[SaraAlthubaiti]

Hi,

I'm trying to follow the steps in the README.md, everything goes well till step 3 while trying to run the code on CPU:

$docker run -it cancerdatascience/scnn:1.0 /bin/bash
root@97d439b58033:/# cd /root/scnn
root@97d439b58033:/# python model_train.py

I got this error at the end (after Filling queue with 495 images before starting to train. This will take a few minutes.):

NVIDIA: no NVIDIA devices found
E tensorflow/stream_executor/cuda/cuda_driver.cc:509] failed call to cuInit: CUDA_ERROR_UNKNOWN
I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:145] kernel driver does not appear to be running on this host (8518fd91b0ac): /proc/driver/nvidia/version does not exist
Killed

Any help to solve this issue, please?

[willeppy]

Same issue here, running on Macbook pro so no GPU access. I assume the TensorFlow code is trying to force GPU usage but not sure since cannot see code yet (issue #16 ).

Is there a flag we can use to force CPU for TensorFlow when running this file?

[chengjianhong]

@SaraAlthubaiti
@willeppy
@cooperlab
Hi, Same issue here!!! Have you solved it yet?