Training_Dependency_Based_Model

[LeonHammerla]

The config jsonnet for the bert-base model (bert_base_dep.jsonnet) has the following lines:

"model": { "type": "transformer_srl_span", "embedding_dropout": 0.1, "bert_model": "bert-base-cased", },

which should be transformer_srl_dependency??? When changing to dependency running and training of the model results in an ERROR

When trying to train a model based on the bert_tiny_dep.jsonnet (which has the correct model-type) training also results in an ERROR, stating:

TypeError: forward() got an unexpected keyword argument 'sentence_end'

[LeonHammerla]

possible fix: add **kwargs to forward method in model.py of transformer_srl_dependency model?