This repository has been archived by the owner on Jul 22, 2024. It is now read-only.
Installing torchvision for ppc64le architecture #268
Comments
|
That channel (https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/) is for Watson Machine Learning Community Edition. WML CE 1.7.0 includes Pytorch 1.3.1 and torchvision 0.4.2. You do not have the channel enabled properly. Read the docs above in order to properly do that. Note that there is also an unsupported update channel with some slightly newer packages: Anaconda's We've also recently released Open-CE. Please visit us at Open-CE has recipes for complete AI environments. The recipes support ppc64le and x86_64. They must be built yourself, but scripts to help take care of that. |
|
I tried installing it directly from the tar but then the version of pytorch changes. Why is that happening @jayfurmanek |
|
Oh, with conda you always want to use the If you've read through: and still have issues, please post back. |
I've not read everything perfectly, but I've definitively tried some non trivial amount of work and reading. I've reading this one quite carefully: https://www.ibm.com/support/knowledgecenter/SS5SF7_1.7.0/navigation/wmlce_install.html but with no success. I thought it would work to do: and it would install torch 1.3.1 + torchvision 0.4.2 as it says here https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/ but my code still fails + when I do conda list it keeps torchvision at 0.2.2 for some weird reason. What I need is to install thing directly from https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/ |
when I said directly from the tar I meant I did: which I downloaded from the right page https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/ |
|
would specifing the channel to be this (https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/) directly help for the torchvision version I want/need? I'd be surprised if it would since I already tried the tar file directly with conda. The conda list didn't reflect the change and the pytorch version changed to 1.1.1 I believe. This is not my first attempt that did that. |
|
Hi @brando90, That might illuminate where the issue is. |
then then btw, thanks for the help! I really appreciate it :) |
|
it seems that the version of pytorch is 1.1.1. That usually changes from the one I had 1.3.1 when I try installing torchvision. But what I want is 1.3.1 + torchvision 0.4.2 from this page: https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/ |
This was referenced Sep 24, 2020
|
Btw, I tried prepending the channel to the url I thought was right (as suggested in the pytorch forum https://discuss.pytorch.org/t/force-installing-torchvision/97279/13) but I get this error: |
|
the url I visit is different when I go to the browser. I will try appending this instead: |
|
So I did try installing things now. There was some attempts to resolve conflicts and the print statement was too long to fit into a comment but here is a first chunk: I will try to see if my code works now + if the right things were installed. |
|
It doesn't seem the right things are installed still: |
|
perhaps creating a fresh container and installing everything from that channels from scratch would work (or pytorch + torchvsion). Will try that. |
|
It looks like you have the channel enabled now (that prepend command), which is good. |
|
@jayfurmanek thanks! I am trying that out right now. It's installing stuff. Btw, why didn't the prepending work without the final is there a way to
? Thanks so for your time! :) |
it wouldn't have occured to me to append |
btw, how did you know I had the right prepended channel now? |
You can also Creating new conda environments for each purpose is recommended. You can run into problems in complex environments, For example, dependencies will conflict if you install packages across releases into a single environment.
|
|
@jayfurmanek now I am having an issue installing h5py for ppc64le, how do you recommend I proceed? Btw, thanks for the amazingly useful and detailed responses! :) |
|
btw this is my real issue: https://stackoverflow.com/questions/64049603/how-does-one-install-torchmeta-for-a-ppc64le-architecture-in-pytorch which is installing torchmeta for ppc64le. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
I am in a weird scenario were I am forced to use torch
1.3.1(due to hardware see: https://public.dhe.ibm.com/ibmdl/export/pub/software/server/ibm-ai/conda/#/). I read from the pytorch docs that it's corresponding version of torchvision is0.4.1(https://pypi.org/project/torchvision/):but for some reason I have the wrong version of it:
is there a way to install the right version of torchvision?
What I've tried:
First I tried force installing the right version with conda. Conda couldn't find the version of torchvision that I need:
Then I proceeded to try to install it regardless with pip
got an error too.
Is there anything else to try?
I tried but it failed:
related:
https://github.com/pytorch/pytorch/issues/45228
The text was updated successfully, but these errors were encountered: