Best IT Recipes

TensorFlow - which Docker image to use?

From Downloading and Configuring TensorFlow under

Install docker I see:

  • b.gcr.io/tensorflow/tensorflow latest 4ac133eed955 653.1 MB
  • b.gcr.io/tensorflow/ tensorflow latest-devel 6a90f0a0e005 2.111 GB
  • b.gcr.io/tensorflow/ tenorflow-full latest edc3d721078b 2.284 GB

I know that 2. and 3. have source code, and I'm using 2. for now.

What is the difference between 2. and 3.? Which one is recommended for "normal" use?

TL; DR:

First of all - thanks for the Docker images! This is the easiest and cleanest way to get started with TF.

A bit about image stuff

  • no pil
  • there is no nano (but there is vi), and apt-get cannot find it. Yes, I can set repositions for it, but why not out of the box.
+3
tensorflow
Jan 09 '16 at 14:32
source share
1 answer

There are four images:

  • b.gcr.io/tensorflow/tensorflow: TensorFlow CPU binary image.
  • b.gcr.io/tensorflow/tensorflow:latest-devel: CPU binary image plus source code.
  • b.gcr.io/tensorflow/tensorflow:latest-gpu: TensorFlow GPU binary image.
  • gcr.io/tensorflow/tensorflow:latest-devel-gpu: a binary image with a GPU and source code.

And both of the concerns are the following:
1. CPU or GPU
2. no source or plus source

CPU or GPU : CPU

For the first time, the user is strongly advised to avoid the GPU version, as they can be any where from hard-to-reach to inability to use. The reason is that not all machines have an NVidia graphics chip that meets the requirements. You must first get TensorFlow to work in order to understand it, and then switch to using the GPU version if you want / need.

From TensorFlow Setup Instructions

Optional: install CUDA (GPUs on Linux)

To create or run TensorFlow with GPU support, both Cuda You must install Toolkit 7.0 and CUDNN 6.5 V2 from NVIDIA.

TensorFlow GPU support requires a graphics processor board with NVidia Compute Capability> = 3.5 . Supported cards include, but are not limited to:

NVidia Titan
NVidia Titan X
NVidia K20
NVidia K40

no source or plus source : no source

Docker images will work without using a source. You should only want or need a source if you need to rebuild TensorFlow for any reason, for example add a new OP .

The standard recommendation for someone new to using TensorFlow is to start with a processor version without a source.

+7
Jan 09 '16 at 15:09
source share


More articles:


All Articles