cgmhaicenter / exBERT

Apache License 2.0
57 stars 15 forks source link

Enquiring about the "val_p" in Pretraining.py #14

Open ghost opened 2 years ago

ghost commented 2 years ago

Hi, I am trying to pre-train the exBERT model with a really small vocabulary file from the stack exchange data dump. While doing that, I am receiving the ValueError: negative dimensions are not allowed.

I do understand where the problem came from, since my file is small, my value of the "len(temp[0]) 2 " is smaller than 1000. As a result, when the code "temp_tl = np.zeros(len(temp[0])2-val_p*2)" is implemented, it would produce the negative error.

May I know what that val_p is for, as well as what I can do to solve this issue?

image