Installation not working
sravya8 opened this issue · comments
I am trying to install this package as:
pip install --no-cache-dir -U https://aws-s3-plugin.s3.us-west-2.amazonaws.com/binaries/0.0.1/1c3e69e/awsio-0.0.1-cp38-cp38-manylinux1_x86_64.whl
(Grabbed the wheel location from the docker file)
And when I run the test:
bash tests/smoke_tests/import_awsio.sh
I see the following error:
from awsio._version import version
ModuleNotFoundError: No module named 'awsio._version'
which platform are you running it on? Can you provide more details?
I am trying to install it on an ec2 instance. May be there are more prereqs needed? Do you recommend using that wheel? Or is there an officially supported wheel?
I recommend using following containers for no-hassle use. You can "docker pull" any of the following container, log into it and start using the plugin directly
763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training:1.9.0-cpu-py38-ubuntu20.04-v1.1
763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training:1.9.0-gpu-py38-cu111-ubuntu20.04-v1.1
763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training:1.8.1-cpu-py36-ubuntu18.04-v1.6
763104351884.dkr.ecr.us-west-2.amazonaws.com/pytorch-training:1.8.1-gpu-py36-cu111-ubuntu18.04-v1.7
How can we use this without your docker image? This plugin would be very useful for us doing local training or who have our own clusters.
@sravya8
Try cd tests/smoke_tests/
, then run bash import_awsio.sh
I got the error as well when running in the project root dir, and python imports the local awsio/ package directly instead of the one you installed previously.
Hi all, , I've updated the installation guide for the plugin: https://github.com/aws/amazon-s3-plugin-for-pytorch#installation
However, we're deprecating support for the S3 plugin, and I suggest to try the new torchdata
S3 IO datapipes. The S3 IO is upstreamed into torchdata
package. We're supporting the new torchdata
package, and continuously improving user experience and performance of the S3 IO datapipes.