How to run Caffe on AWS GPU instance via Docker

[Traun Leyden]

I managed to get Caffe running on a GPU instance.  I'm running it under Docker because it makes it easy to spin up instances without having to wrangle with installing all of the dependencies.  (the dependencies, along with a pre-compiled caffe, are baked into the Docker image.).

Here is a blog post that describes how to get up and running:

Running Caffe on AWS GPU Instance via Docker

and it has links to the Docker image and Dockerfile.

Using a g2.2xlarge AWS instance (without cudnn) here's how long it took to train the MNIST LeNet example:

real 3m51.072s

user 2m57.399s

sys 0m58.011s

It's faster than on my laptop using CPU, but I'd be curious to hear how it compares to people running it on their workstations in non-cudnn GPU mode.

[revspooner]

cool, thanks for writing that up.  I got the following results using the AMI mentioned in this thread:

./examples/mnist/train_lenet.sh on g2.2xlarge

GPU CUDA 6.0 = 239 secs

CPU = 1075 secs

GPU CUDA 6.5 w/CuDNN g++-4.6 = 47 secs

CPU g++-4.6 = 1052secs

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users...@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/6822bced-77bd-444f-8bbd-30d3f074503e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.