BERT Baseline - AllenNLP

[xz941]

1. Is the data loader defined correctly? (The field of the label)
2. The architecture of model: bert-embedding -> lstm encoder -> sofmax classifier.
3. Do we need CLS token?
4. How will the training process influence the parameter? will it change parameter in bert embedding?
5. How can we monitor the training process of Allennlp.
6. Any details about the model? padding(sentence with different length), lstm parameter, loss function, initialization

[anikl9705]

As suggested by Pratyay, we can all focus on one library from now.