Pytorch 1.3 ?

[cedias]

Hi,
The 1.3 version of pytorch (https://github.com/pytorch/pytorch/releases) have been released and I was wondering if you were planning to add support ?

Thanks

[laurenyu]

Thank you for your interest! I've reached out to the corresponding team for their input.

[anankira]

Thank you for your interest ! Yes, we plan to add support for Pytorch 1.3. However, we currently do not have a timeline to share.

[ruodingt]

It would be really nice to have the support for pytorch1.3

[akartsky]

Pytorch 1.3.1 containers are released.
Docker files will be merged to this repo in a few days.

[domino14]

Hi @akartsky and others. Does AWS update their pre-built Docker images frequently? I am still using this line in my Dockerfile:

FROM 520713654638.dkr.ecr.us-east-1.amazonaws.com/sagemaker-pytorch:1.1.0-gpu-py3

Replacing 1.1.0 with anything above does not work - those images don't seem to exist.

[laurenyu]

versions 1.2.0+ follow a different format:

763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.3.1-gpu-py3

https://docs.aws.amazon.com/sagemaker/latest/dg/pre-built-containers-frameworks-deep-learning.html

[siavashdarkvision]

@laurenyu where is a list of these magic ECR repositories? The link that you mention does not list them.

[laurenyu]

@siavashdarkvision looks like the page has been updated, but it does link to: https://github.com/aws/deep-learning-containers/blob/master/available_images.md

[siavashdarkvision]

@laurenyu if I extend a recent version of pytorch-training image like 1.7.1 using this, will I be able to do hyper-parameter tuning?

[laurenyu]

@siavashdarkvision it should work!

(if you run into problems, please open a new issue since hyperparameter tuning is a separate topic and this issue has been closed for over a year)