Skip to content

Instantly share code, notes, and snippets.

@ando-takahiro
Created March 30, 2017 13:34
Show Gist options
  • Save ando-takahiro/08c11e5a87da9304afa3a0437c75fead to your computer and use it in GitHub Desktop.
Save ando-takahiro/08c11e5a87da9304afa3a0437c75fead to your computer and use it in GitHub Desktop.
FROM nvidia/cuda:8.0-cudnn5-devel
RUN apt-get update -y && \
apt-get install -y --no-install-recommends \
python-dev \
python-setuptools \
python-pip && \
pip install --upgrade pip && \
rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*
RUN pip install chainer==1.22.0
@ando-takahiro
Copy link
Author

test result on RazerBlade 2016 with ubuntu 16.04

without GPU

$ time python train_mnist.py
(... snip ...)
     total [#################################################.] 99.17%
20          0.00686444  0.117194              0.997833       0.9789                    617.806       
     total [##################################################] 100.00%

real	10m19.075s
user	38m6.804s
sys	41m37.452s

with GPU

$ time python train_mnist.py -g 0
(... snip ...)
  total [#################################################.] 99.17%
20 0.00838405 0.118239 0.998182 0.9791 39.6067
  total [##################################################] 100.00%

real 0m41.047s
user 0m42.508s
sys 0m2.784s

@ando-takahiro
Copy link
Author

If you got this error CURAND_STATUS_LAUNCH_FAILURE, when using official chainer docker image, above file would help you.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment