8000 Require TensorFlow 1.0.0rc0 instead of 1.0.0a0 by luthfianto · Pull Request #428 · blei-lab/edward · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Require TensorFlow 1.0.0rc0 instead of 1.0.0a0 #428

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
wants to merge 1 commit into from
Closed

Require TensorFlow 1.0.0rc0 instead of 1.0.0a0 #428

wants to merge 1 commit into from

Conversation

luthfianto
Copy link
@luthfianto luthfianto commented Jan 30, 2017 •
edited
Loading

EDIT:
While installing Edward via setup.py/pip on Windows is fine. On Linux, pip complains that TensorFlow 1.0.0rc0 isn't compatible with 1.0.0a0

@mariru
8000
Copy link
Contributor
mariru commented Jan 30, 2017

@rilut I Installed tensorflow-1.0.0rc0 on linux using pip with the instructions here
https://github.com/tensorflow/tensorflow/blob/r1.0/tensorflow/g3doc/get_started/os_setup.md
after that I was able to pip install edward without problems. (Note: I am using python 2.7, no GPUs)

@pwl
Copy link
pwl commented Jan 31, 2017

I have the same issue. When running pip install edward==1.2.1 I'm getting

Collecting edward==1.2.1
  Downloading edward-1.2.1.tar.gz
Collecting tensorflow>=1.0.0a0 (from edward==1.2.1)
  Could not find a version that satisfies the requirement tensorflow>=1.0.0a0 (from edward==1.2.1) (from versions: 0.12.0rc0, 0.12.0rc1, 0.12.0, 0.12.1)
No matching distribution found for tensorflow>=1.0.0a0 (from edward==1.2.1)

This is happens after installing TF 1.0.0rc0 with pip install https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.0.0rc0-cp35-cp35m-linux_x86_64.whl.

@luthfianto
Copy link
Author

I'm starting to think: or is it because pip doesn't recognize tensorflow-gpu as tensorflow?

Is it possible to make pip/setup.py recognize both tensorflow and tensorflow-gpu?

@pwl
Copy link
pwl commented Jan 31, 2017 •
edited
Loading

@rilut yep, that seems to be the case, I had no trouble installing Edward with https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.0.0rc0-cp35-cp35m-linux_x86_64.whl, so the non-gpu version of TF.

@dustinvtran
Copy link
Member

that's strange.. from looking at other packages, prettytensor seems to make a similar remark here. i'm also not sure if install_requires accepts booleans like tensorflow>=1.0.0a0 or tensorflow-gpu>=1.0.0a0`.

@gokceneraslan
Copy link

I think there is just no way to handle it properly. One idea is to detect whether there is a GPU during installation in setup.py:

from subprocess import check_output, CalledProcessError

try:
    num_gpus = len(check_output(['nvidia-smi', '--query-gpu=gpu_name',
                                 '--format=csv']).decode().strip().split('\n'))
    tf = 'tensorflow-gpu' if num_gpus > 1 else 'tensorflow'
except CalledProcessError:
    tf = 'tensorflow'

setup(..., install_requires=[tf, ...])

But it's just an ugly hack.

@dustinvtran dustinvtran mentioned this pull request Jan 31, 2017
Closed
@dustinvtran
Copy link
Member

In the mean time I submitted an issue on TensorFlow (tensorflow/tensorflow#7166).

@dustinvtran dustinvtran mentioned this pull request Feb 1, 2017
Closed
@dustinvtran dustinvtran mentioned this pull request Feb 27, 2017
Merged
@luthfianto luthfianto deleted the patch-1 branch February 28, 2017 16:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
5 participants
@luthfianto @mariru @pwl @dustinvtran @gokceneraslan
0