Caffe for NLP

659 views
Skip to first unread message

Anuj Gupta

unread,
Sep 1, 2015, 7:08:13 AM9/1/15
to Caffe Users
Hi Everyone

I am a newbie here. Just yesterday installed caffe and ran a LeNet on MNIST

Question: I am working on a NLP application want to understand how can I try Caffe for my problem.
Can someone point me to some simple working examples of Caffe and NLP ?

Thanks
Anuj

Cyndi

unread,
Sep 2, 2015, 3:18:03 AM9/2/15
to Caffe Users
You can perhaps get some ideas by looking at code from Donahue et al (http://jeffdonahue.com/lrcn/).  There is code to train a language model with an LSTM (look at train_language model.sh in the Jeff Donahue's reucrrent branch under /examples/coco_caption).  Caffe has an "embed" layer which is important for NLP applications since it allows you to easily embed a one-hot vector.  Hopefully, that small bit of info helps :-)

Yi Liu

unread,
Sep 2, 2015, 2:49:57 PM9/2/15
to Caffe Users
CaffeNLP is another possible resource.

https://github.com/Russell91/NLPCaffe
Reply all
Reply to author
Forward
0 new messages