Max Berrendorf

> in step 3. it should be scores[num_items:] ri8? No. With the code above, you use the node ids `[0, ..., num_users)` for user nodes, and `[num_users, ..., num_users +...

Hi @noy-sternlicht , we would be happy to accept PRs to add this model. We already have some support for using transformer based text encoders, cf. [`pykeen.nn.representation.TextRepresentation`](https://pykeen.readthedocs.io/en/stable/api/pykeen.nn.representation.TextRepresentation.html) and [this tutorial...

Hi @lrey-civetta , would you mind sharing a full traceback of the error? This makes it easier to understand from where this error is coming. Also, can you confirm that...

From your config, it looks like you are using automatic batch size selection/maximization for training, too ```python training_kwargs={ "batch_size": None, ... }, ``` While in theory this shouldn't cause OOM's...

Hm, okay right away I do not see where this is coming from. If you want to dig deeper into it, you could try to get more information about the...

Hi @lrey-civetta @LizzAlice , you may want to try again with the fixed brought to `master` by https://github.com/pykeen/pykeen/commit/2b4b77683461d1e2dd4ec93f125c477d8ead1f51

This should be the relevant part: https://pykeen.readthedocs.io/en/stable/api/pykeen.nn.init.PretrainedInitializer.html#pykeen.nn.init.PretrainedInitializer

Hi @sbonner0 , first of all, thanks for starting with a proposal. :rocket: I like the idea of having (positive) triple weights, too, since it allows users to provide any...

> Hey @mberr - thanks for the suggestion to add weights into the training instances. I guess that the appropriate weights could then be returned from the data loader via...

> Hey @mberr sorry for the delay in this but I have some free time and would love to try and make some progress on this from my end finally!...