Eliyar Eziz

predict 接口就可以传多个 sentence，本来就是个`[sample]`

Please try to access it with vpn. It works just fine.

新版本应该已修复，在测试看看～

> I'm thinking out aloud here... since ultimately this is a transformer model I presume the sequence needs to be aligned. > > Would I structure the data this way?...

Just make a label dict then convert label to numeric sequece. eg: ```python dic = { "O": 0, "B-TERM1": 1, "B-TERM2": 2, "B-TERM3": 3, "B-TERM4": 4, } ``` This will...

You need to convert those tokens to digits.

Yes, Since you don't have 26 tags in one sentence and all you want the model to do is separate two entity's translation. The ii) way is much better and...

Please try it without the CRF layer, let check out is that an issue related to CRF Layer.

@echan00 What is the evaluation result when you evaluate on the training set and the validation set? Maybe this is because of not enough dataset?

I mean the result of ``` model.new_evaluate(train_final_x, train_final_y, debug_info=False) model.new_evaluate(valid_final_x, valid_final_y, debug_info=False) ```