Closed jnistane3 closed 1 year ago
Hi @jnistane3, it appears that you are training a single task model, right? For single task models, selector_dim
should equal 0. selector_dim
refers to the dimension of the selector vector. For multi-task data, prepare_train
will, for each row in your data, create a selector vector. Each vector contains one dimension per task. For single task, prepare_train
creates an empty selector vector, so the dimension is equal to 0. I will add comments to example.py
and example2.py
to clarify this. Try selector_dim = 0
and let me know how it goes.
Perfect, setting selector_dim to zero solved it for my single-task models. Thanks @rishigurnani!
Running into an issue of matrix multiplication while training the model, any ideas on how to solve this?