I would like to know which deep learning framework ( Keras , PyTorch or Tensor Flow) is the best option to implement customized architectures for Natural Language Research.
I have attached a sample rnn based architecture for medical concept normalization.
Dear Kalyan Katikapalli ,
I think Keras is a good choice since it is easy to implement complex architectures and provides a good support. However, see this link (Point #2) :
https://medium.com/@UdacityINDIA/tensorflow-or-pytorch-the-force-is-strong-with-which-one-68226bb7dab4
I have no experience with RNNs in Keras but I think it works fine.
@Kalyan keras is good tool for all sorts of deep learning framework. However, modelling seq2seq learning is still not possible via keras (you can always use code provided by other developers on git). But if you are looking for better performamce with ease to code and flexibility go with pytorch.
If you are starting fresh, PyTorch is a really good option. It is easy to use. If your are using Seq2Seq models, then https://ibm.github.io/pytorch-seq2seq/public/index.html may be of your interest.
@kalyan Tensorflow is a really good option.If you are using Seq2seq models,then you can find many GitHub repository for the same.
Thanks for your answer, Vishnu Raj . Can you suggest me some good resources to learn PyTorch (I means books or online courses). I already searched in internet and find many resources, but I'm not able to decide which one to follow
Kalyan Katikapalli The best starting point I found so far is the official tutorials page: https://pytorch.org/tutorials/ and its associated github repo: https://github.com/pytorch/tutorials. It also includes a Seq2Seq tutorial (https://pytorch.org/tutorials/intermediate/seq2seq_translation_tutorial.html), which you can use as starting point for your problem (after covering the basics).
If you find these answers helpful, don't forget to hit recommend button; it will increase the visibility of helpful answers.
- Badges
- Science topic