deepmodeling / deepmd-kit

A deep learning package for many-body potential energy representation and molecular dynamics
https://docs.deepmodeling.com/projects/deepmd/
GNU Lesser General Public License v3.0
1.41k stars 486 forks source link

Does conda install of deepmd support Platform: linux-aarch64? #3932

Open robinzyb opened 4 days ago

robinzyb commented 4 days ago

Summary

no

DeePMD-kit Version

x

Backend and its version

x

Python Version, CUDA Version, GCC Version, LAMMPS Version, etc

No response

Details

I tried to install deepmd-kit using conda conda create -n deepmd deepmd-kit lammps horovod -c conda-forge And the message is

Channels:
 - conda-forge
 - defaults
Platform: linux-aarch64
Collecting package metadata (repodata.json): done
Solving environment: failed

PackagesNotFoundError: The following packages are not available from current channels:

  - lammps
  - horovod
  - deepmd-kit

Current channels:

  - https://conda.anaconda.org/conda-forge
  - defaults

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org

and use the search bar at the top of the page.

I found there is only linux-64 not linux-aarch64. I am not sure if the error is caused by the platform. if so, could we add support on linux-aarch64 Screenshot 2024-06-30 at 14 00 58

njzjz commented 4 days ago

xref: https://github.com/conda-forge/tensorflow-feedstock/issues/136

robinzyb commented 4 days ago

thanks. This is required by newest hpc in CSCS (https://www.cscs.ch/computers/decommissioned/toedi). Looking forward to the update. Our new project will start at this new hpc in October. The worst case would be compiling dpmd from scratch.

robinzyb commented 2 days ago

Hi, aarch64 is also required by the docker image. Does the docker iamge also suffer from the problem of upstream (tensorflow)? I only found x86 docker image for tensorflow in dockerhub (https://hub.docker.com/r/tensorflow/tensorflow/tags?page=2&page_size=&ordering=&name=)

njzjz commented 2 days ago

Hi, aarch64 is also required by the docker image. Does the docker iamge also suffer from the problem of upstream (tensorflow)? I only found x86 docker image for tensorflow in dockerhub (https://hub.docker.com/r/tensorflow/tensorflow/tags?page=2&page_size=&ordering=&name=)

Currently, we build docker images just using the offline installer, maintained at https://github.com/deepmd-kit-recipes/docker-images.

robinzyb commented 1 day ago

Hello, I checked the dockerfile. The offline installer only has linux x86_64.sh. How do you build the offline installer? Do we need to build an arm64 offline installer?

njzjz commented 1 day ago

How do you build the offline installer?

The offline package is built at https://github.com/deepmd-kit-recipes/installer, which is a collection of conda packages.

robinzyb commented 1 day ago

Thank you. It seems everything goes back to the conda install deepmd-kit where we need the aarch64 for tensorflow. should fix the upstream problem first.