🚚💻 Fix TransformerEncoder tokens' device

[mberr]

Fix #858 (comment)

simplified MWE

from pykeen.pipeline import pipeline
from pykeen.datasets import get_dataset
from pykeen.nn.representation import LabelBasedTransformerRepresentation
from pykeen.models import ERModel

dataset = get_dataset(dataset="nations")
entity_representations = LabelBasedTransformerRepresentation.from_triples_factory(
    triples_factory=dataset.training,
)
result = pipeline(
    dataset=dataset,
    model=ERModel,
    model_kwargs=dict(
        interaction="distmult",
        entity_representations=entity_representations,
        relation_representations_kwargs=dict(
            shape=entity_representations.shape,
        ),
    ),
    training_kwargs=dict(
        num_epochs=1,
    ),
)

[mberr]

locally tested successfully on a GPU machine

[thtang]

Could I use the latest code via

git clone https://github.com/pykeen/pykeen.git
cd pykeen
pip install -e .

?

[mberr]

Could I use the latest code via

git clone https://github.com/pykeen/pykeen.git
cd pykeen
pip install -e .

?

Yes, but you need to switch the branch before running pip install:

git clone https://github.com/pykeen/pykeen.git
git checkout fix-label-based-repr
cd pykeen
pip install -e .

[mberr]

@PyKEEN-bot test