allow to pass unresolved loss function to ERModel's __init__

[mberr]

So then it's fine to finish this PR without adding a new model, but it would be important to write some narrative documentation explaining how one might use this interaction to construct a new model

This should be the place to put it? To slightly modify the example there and create an "anonymous" model instead of declaring a new class:

from pykeen.models import ERModel
from pykeen.nn import EmbeddingSpecification
model = ERModel(
    triples_factory=...,
    loss=..., # TODO: we should support passing a string here, and use the loss-resolver to resolve it
    interaction="transformer",
    entity_representations=EmbeddingSpecification(embedding_dim=64),
    relation_representations=EmbeddingSpecification(embedding_dim=64),
)

Originally posted by @mberr in #714 (comment)

[cthoyt]

Yeah, that's a great place for it!

[mberr]

We could also consider allowing to pass an integer as {entity,relation}_representations and create and internally create an EmbeddingSpecification with default settings from it. This might make it easier to use of ERModel via CLI. 🤔

[cthoyt]

You mean like if you pass an integer it just assumes you want a vanilla embedding spec with that as the dim? Or a sequence of integers would make a sequence of embeddings? Or a sequence of tuples of ints would make fun shaped ones?

[mberr]

You mean like if you pass an integer it just assumes you want a vanilla embedding spec with that as the dim? Or a sequence of integers would make a sequence of embeddings? Or a sequence of tuples of ints would make fun shaped ones?

Something like this. A shortcut which allows you to create something without having to import EmbeddingSpecification if you are not interested in all the possible configurations.