Max Berrendorf

Hi @mhmgad , we currently have [pykeen.pipeline.PipelineResult.save_to_s3](https://pykeen.readthedocs.io/en/stable/api/pykeen.pipeline.PipelineResult.html#pykeen.pipeline.PipelineResult.save_to_s3) to upload the final results to `s3`, but no support for checkpoints yet. We would be happy to see a PR for this....

You could also consider to either make a GitHub release with a release artifact to download the data, or, maybe even better, create a persistent dump of the data in...

> There is a [zenodo release](https://zenodo.org/record/268568), but that doesn't fix the problem. It might, at least for use within PyKEEN, fix the problem, if we change to using the zenodo...

Hi @mortenterhart , ComplEx uses vectors of complex numbers as representation. `torch` does not yet fully support complex tensors (at least not in all places); thus, they are sometimes "flattened",...

Hi @martytom , if you train a model on a dataset that uses inverse relations that has an effect on the model: - the model will have twice as many...

> The Tests workflow has the command tox -e py failing apparently because of OSError: /home/runner/work/pykeen/pykeen/.tox/py/lib/python3.8/site-packages/torch_scatter/_scatter_cpu.so: undefined symbol: _ZN2at4_ops6narrow4callERKNS_6TensorElll . This is due to the update of PyTorch to 2.0,...

> The Documentation workflow is failing because of the README file. However, the changes I made there were minor and I followed the existing pattern (commit: https://github.com/pykeen/pykeen/commit/b3871a033932c9ad82f8aeb4b11b04a42b67e677). The table parts...

> Hi @mberr and @cthoyt, > > I added the results of the experiments, except for the experiment in `kristiadi2019_distmult+literale_glin_yago310.yaml` for which I got an error (see stacktrace below). >...

There is an `evaluation_fallback` parameter of the pipeline, which has been added in https://github.com/pykeen/pykeen/pull/122 and will try to evaluate the model on CPU, if the final evaluation fails for memory...

@AntonisKl if you already want to give it a try, you can take a look at #1261