search

CheriseZhu / GIL

The implementation of our NeurIPS 2020 paper "Graph Geometry Interaction Learning" (GIL)
44 stars 6 forks source link

environment problems #2

Open pkuyzy opened 3 years ago

pkuyzy commented 3 years ago

Hi, when I try to run your code, several problems rise due to the environment.

  1. the package sage, if installed by pip, will only has a version sage==0.0.0. is that the right package to use?
  2. if working with geoopt==0.0.1, there is no module 'geoopt.manifolds.poincare.math', only geoopt==0.1.x has the module, however, it does not support torch==1.1.0
  3. could you please share other version of the packages, such as torch-sparse and torch-cluster?

Thanks very much!

CheriseZhu commented 3 years ago

Sorry to be slow replying, it has been busy with something.

  1. We update the version of sage with 9.0. And you can install it via https://doc.sagemath.org/html/en/installation/conda.html.
  2. The version of geoopt == 0.0.1 and torch == 1.1.0 were used in the paper as shown in Readme. If it still does not work in your environment, the latest version is recommended, geoopt == 0.3.1.
  3. The version of torch-sparse == 0.4.0 and torch-cluster == 1.4.2.
pkuyzy commented 3 years ago

Actually, I have already run the code with geoopt==0.1.0, but thanks anyway for the response.

Sorry to be slow replying, it has been busy with something.

  1. We update the version of sage with 9.0. And you can install it via https://doc.sagemath.org/html/en/installation/conda.html.
  2. The version of geoopt == 0.0.1 and torch == 1.1.0 were used in the paper as shown in Readme. If it still does not work in your environment, the latest version is recommended, geoopt == 0.3.1.
  3. The version of torch-sparse == 0.4.0 and torch-cluster == 1.4.2.