search

stanfordnlp / treelstm

Tree-structured Long Short-Term Memory networks (http://arxiv.org/abs/1503.00075)
GNU General Public License v2.0
875 stars 236 forks source link

the result is "nan" in semantic relatedness task #12

Open xxxxxxxxy opened 7 years ago

xxxxxxxxy commented 7 years ago

while in training model,the dev score is nan. And the final score is nan. I opened the prediton file, the result is all nan. So I checked the data, and it is ok. I have no idea about why the result is nan.

xxxxxxxxy commented 7 years ago

using the default parameter to run the code will get 'nan', but changing the model parameter such as lstm will not get nan.

bgstn commented 6 years ago

Same problems here.

bgstn commented 6 years ago

I found something new to the problem. It seems that during the training period, the parameters for the model become nan after several iterations(around 50) in epoch 1. Any suggestions why?

ankitsgh commented 6 years ago

Have any of you sorted the issue? @xxxxxxxxy @bgstn @arunchaganty

knjwhn commented 5 years ago

Same problem,Have anyone solve the issue?Help please.

knjwhn commented 5 years ago

@ankitsgh have you solve this problem?

smilenrhyme commented 4 years ago

+1 facing the same issue.

Anyone found solution for this ? @bgstn @ankitsgh @knjwhn @xxxxxxxxy

Thanks !!

Captain6606 commented 9 months ago

@xxxxxxxxy @ankitsgh @bgstn @knjwhn @smilenrhyme How did you solve it