4

From TensorFlow Download and Setup under

Docker installation I see:

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

I know 2. & 3. are with source code and I am using 2. for now.

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

TLDR:

First of all - thanks for Docker images! They are the easiest and cleanest way to start with TF.

Few aside things about images

  • there is no PIL
  • there is no nano (but there is vi) and apt-get cannot find it. yes i probable can configure repos for it, but why not out of the box
Guy Coder
  • 24,501
  • 8
  • 71
  • 136
rgr
  • 543
  • 2
  • 6
  • 13
  • 1
    Well - TF GG is telling "ask your questions at SO". SO "says" go to your GG! There should be end of this loop. – rgr Jan 09 '16 at 15:07
  • 1
    You are not wrong and did ask at the right place. It was just the format of how you asked that needed updating, so I did that. StackOverflow is not a discussion board and tends to hold the questions and answers to a standard. People come here because they can get quick answers to specific questions, and your question was specific. However people who answer questions like to have all of the info needed for them in the question. I had to hunt down the source of the info and add the link because it was missing the info about CPU and GPU. No harm, I had the same problems when I started. – Guy Coder Jan 09 '16 at 15:32
  • Can you edit the question and link back to the Google Group question. I can't seem to find the question there. – Guy Coder Jan 09 '16 at 15:37

1 Answers1

9

There are four images:

  1. b.gcr.io/tensorflow/tensorflow: TensorFlow CPU binary image.
  2. b.gcr.io/tensorflow/tensorflow:latest-devel: CPU Binary image plus source code.
  3. b.gcr.io/tensorflow/tensorflow:latest-gpu: TensorFlow GPU binary image.
  4. gcr.io/tensorflow/tensorflow:latest-devel-gpu: GPU Binary image plus source code.

And the two properties of concern are:
1. CPU or GPU
2. no source or plus source

CPU or GPU: CPU

For a first time user it is highly recommended to avoid the GPU version as they can be any where from difficult to impossible to use. The reason is that not all machines have an NVidia graphic chip that meet the requirements. You should first get TensorFlow working to understand it then move onto using the GPU version if you want/need.

From TensorFlow Build Instructions

Optional: Install CUDA (GPUs on Linux)

In order to build or run TensorFlow with GPU support, both Cuda Toolkit 7.0 and CUDNN 6.5 V2 from NVIDIA need to be installed.

TensorFlow GPU support requires having a GPU card 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

The docker images will work without needing the source. You should only want or need the source if you need to rebuild TensorFlow for some reason such as adding a new OP.

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

Guy Coder
  • 24,501
  • 8
  • 71
  • 136
  • Thanks it means very first **b.gcr.io/tensorflow/tensorflow-full** is gone now. However it is still available to pull. docker pull b.gcr.io/tensorflow/tensorflow-full Using default tag: latest latest: Pulling from tensorflow/tensorflow-full Digest: sha256:82c4f725adad2196877dfe70d849024ec6e3940081a05a0555cee99c9b213b2d Status: Image is up to date for b.gcr.io/tensorflow/tensorflow-full:latest – rgr Jan 09 '16 at 15:19
  • I installed tensorflow-full (as you said it's still available to pull). However, it seems tensorflow-full doesn't contain alexnet and imagenet. So I guess our best option is to use tensorflow instead (since tensorflow-full is gone)? – Ruby May 28 '16 at 15:12
  • Does the GPU binary works also in CPU-only environments, or is GPU required? – AkiRoss Sep 21 '16 at 10:11
  • @AkiRoss I do not know; I have not tried it. You should ask that as a separate question. – Guy Coder Sep 21 '16 at 11:40
  • Ok, I tried and apparently it falls back to CPU if GPU is missing, but I don't know if some optimizations are missing. I commented here because it appeared pertinent. – AkiRoss Sep 21 '16 at 11:42
  • @AkiRoss I basically put TensorFlow aside until it works on Windows with the GPU. At present I plan to use [Bash on Ubuntu on Windows](https://msdn.microsoft.com/en-us/commandline/wsl/about) but am waiting for that to work with the [GPU](https://wpdev.uservoice.com/forums/266908-command-prompt-console-bash-on-ubuntu-on-windo/suggestions/13250010-ubuntu-shell-access-to-system-gpu-ability-to-run) – Guy Coder Sep 21 '16 at 12:08