Moritz Fürneisen

My original plan was to do something modular. So the individual classifiers could also be of fasttext type. Config would be like the following, excluding some stuff for brevity. ```...

`whitelist`/`blacklist` would definitely be more general. I think the single concept case would probably require some additional work. (Regarding metrics mostly) Also single concept would allow for easy sampling i.e....

I think for more than one concept the algorithms should just behave as they would normally, just on that subset of concepts. Actually it may be a good starting point...

A small status update. To figure out good parameter values for the YSO data set from the tutorial repository. Unfortunately the suggest command is really slow. Presumably since I run...

Rebased the code. If you want to try it you can use the following config: ``` [xtransformer-yso-en] name=XTransformer YSO english backend=xtransformer analyzer=simple language=en vocab=yso batch_size=16 truncate_length=256 learning_rate=0.0001 num_train_epochs=1 max_leaf_size=18000 ```...

The default language model is currently defined as https://huggingface.co/distilbert-base-multilingual-cased I opted for the multilingual as default so you don't have to worry that much about model selection. You can try...

The `model_shortcut` parameter only support models from the huggingface hub. Pecos supports local files but this parameter is currently not exposed. If you want I can add it.

For ZBW the main issue in the past has been compatibility of serialized model between versions. Unfortunately this is the hardest part as this can break when dependencies are updated...

@juhoinkinen I added you as a collaborator to my fork. Maybe this helps. I can also try the `git pull origin api-i18n --allow-unrelated-histories` later.

Just wanted to add some reading material for semantic search on dense word vectors: * https://arxiv.org/abs/1908.10084 General Approach for simmillarity on word vectors * https://arxiv.org/abs/2109.04404 Investigates what Layer of the...