Closed contribcode closed 3 years ago
Hi, thanks for using the library! About your questions:
model_parameters
function. Apologies if you already know this.Thank you @kmkurn for your fast reply, you covered my questions. Thank you for the library too.
Hello,
I have two questions regarding the usage of the library, and since they are not issues, I just open one issue. If there is a more appropriate place for asking questions about the library, let me know and I will transfer them there.
Question 1
Firstly, I want to use the CRF layer for a binary task. The number of tags of the CRF layer will be one or two? I ask this because in many cases, for binary tasks, libraries' functions accept parameters just for the positive class.
Question 2
My second question is about weights. I want to use a CRF layer on top of a BERT model. My implementation is the following
I define CRF layer out of the model definition, because it returns the log likelihood. subsequently, I pass the weights to the optimizer. Is this the rigth way to update to update the weights for the model, i.e. the model along with the CRF layer?