Skip to content
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.

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

Support for 32 bits architecture #32315 #32627

Closed
steffanjensen opened this issue Sep 18, 2019 · 13 comments
Closed

Support for 32 bits architecture #32315 #32627

steffanjensen opened this issue Sep 18, 2019 · 13 comments
Assignees
Labels
type:build/install Build and install issues type:feature Feature requests

Comments

@steffanjensen
Copy link

steffanjensen commented Sep 18, 2019

Several users are still using Python32 bits and they cannot install TensorFlow. For them, pip install TensorFlow fails as no wheel matches the tags expected by their environment (to debug, pip debug --verbose shows only tags that don't math the filenames of our wheels).

There is some requests to support 32 bits, see for example #31431

This is not going to be easy as we need to also compile the C++ codebase in 32 bits mode and that would cause issues with code written assuming types have a certain bit width.

There is no change in the user visible API, just a new set of wheels to support more users.

Opening this to reference in all similar issues.

After closing my thread, i just reopen it. Google employees trying to tell me this is not the roadmap for the tool. The day i am going to care about a Google employee opinion is the day hell froze.

If you close this issue I will keep posting daily, until issue is fixed.

@ewilderj
Copy link
Contributor

Hi there @reliefs

TensorFlow is an open source project: while the people paid to work on it by Google have limited bandwidth, this doesn't mean other people can't join in and help. If you want to improve the situation, please consider joining the SIG Build project in which many community collaborators are working to build TF for many architectures.

Repeatedly pinging the same issue for something we're aware of makes things worse. Your tone in this issue and in #32315 is unhelpful: there are real people working on this project who care. Please be civil and constructive, for the sake of everyone who works hard and contributes, and abide by our Code of Conduct.

Thank you

@steffanjensen
Copy link
Author

Hi there @reliefs

TensorFlow is an open source project: while the people paid to work on it by Google have limited bandwidth, this doesn't mean other people can't join in and help. If you want to improve the situation, please consider joining the SIG Build project in which many community collaborators are working to build TF for many architectures.

Repeatedly pinging the same issue for something we're aware of makes things worse. Your tone in this issue and in #32315 is unhelpful: there are real people working on this project who care. Please be civil and constructive, for the sake of everyone who works hard and contributes, and abide by our Code of Conduct.

Thank you

Thanks you for your reply and understanding. People just ask me daily on how to fix this issue, it really bothers me and take time out of my work day..

Have a great weekend.

@ewilderj
Copy link
Contributor

I'm interested in why you are getting asked this daily? Excuse my ignorance of your job/other projects, but I'm very curious as to which users and use cases mean you get this repeated question? It might give us clues or data that will help.

@steffanjensen
Copy link
Author

steffanjensen commented Sep 19, 2019

Face detection is the new thing around script kitties.

But I have many people in last thread and other who use my work, TensorFlow is a good framework, if people want to use it in their package, and they should eliminate all 32bit servers and PCs, it's really crazy.

Not all people have as fast pcs as us.

@oanush oanush self-assigned this Sep 19, 2019
@oanush oanush added the type:build/install Build and install issues label Sep 19, 2019
@oanush oanush assigned ymodak and unassigned oanush Sep 19, 2019
@alexzruiz
Copy link

alexzruiz commented Sep 19, 2019

When im try to install from raspberry (ubuntu 18) says error:

Collecting tensorflow==2.0.0rc0 (from -r requirements.txt (line 3))
Could not find a version that satisfies the requirement tensorflow==2.0.0rc0 (from -r requirements.txt (line 3)) (from versions: )
No matching distribution found for tensorflow==2.0.0rc0 (from -r requirements.txt (line 3))

What happen?

@pandrey-fr
Copy link
Contributor

@alexzruiz
No offense, but your issue has nothing to do with the current thread. That being said, it seems that there is a typo in the requirements file you are using: it should be tensorflow==2.0.0-rc0.

@steffanjensen
Copy link
Author

I will build a work around to fix this for my users, i feel bad for the rest of the community.. Good luck :/

@tensorflow-bot
Copy link

tensorflow-bot bot commented Oct 1, 2019

Are you satisfied with the resolution of your issue?
Yes
No

@mihaimaruseac
Copy link
Collaborator

If possible, please release the work around back to the community, in true spirit of open source development

@weizhuang-93
Copy link

would love to see another wheels for 32bit system too.. eg. Hololens 1, although hololens 2 is x64 but i am sure a lot other produkt will still be in x32

@weizhuang-93
Copy link

https://github.com/fo40225/tensorflow-windows-wheel/tree/master/1.8.0/py36/CPU/sse2 is a possible functional wheel for 32bit system..

@steffanjensen steffanjensen reopened this Nov 14, 2019
@ymodak ymodak added the type:feature Feature requests label Jun 1, 2020
@ymodak ymodak removed their assignment Jun 1, 2020
@kumariko kumariko self-assigned this Oct 26, 2021
@google-ml-butler
Copy link

Are you satisfied with the resolution of your issue?
Yes
No

@mihaimaruseac
Copy link
Collaborator

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
type:build/install Build and install issues type:feature Feature requests
Projects
None yet
Development

No branches or pull requests

9 participants