GithubHelp home page GithubHelp logo

Comments (5)

ZhenYangIACAS avatar ZhenYangIACAS commented on May 28, 2024
  • There is no difference between the train.src_path and generator.src_path. Actually, they are not used at the same time. The former is used to pre-train the model and the latter is used to train the generator during the gan training.
  • The dis_positive_data and the dis_negative_data is generated automatically. In our experiments, it will includes 5000 sentence for training the discriminator during gan training.
  • the parameter λ is not defined in the config file, just set in the model.py
  • We stop the training when the model achieves no improvement on the development sets. The process for the evaluating is the same with the testing. We didn't test the PPL score for evaluation.

from nmt_gan.

zoharai avatar zoharai commented on May 28, 2024

Hi @ZhenYangIACAS ,
Thanks for the answer!
You said that the dis_positive_data and the dis_negative_data are generated automatically, but we need to generate it for the first iteration of the first epoch right?
Another question about the evaluation - did you use evaulate.sh for evaluation?
When did you evaluate bleu score on the dev set? every epoch?
Did you stop training when the model achieves no improvement for the tenth evaluation on the dev set or first evaluation or another number?

Thanks!

from nmt_gan.

borgr avatar borgr commented on May 28, 2024

After the preprocessing you start training with GANs, do you use early stopping there as well (with patience 10?) or do you have another way to decide when to stop training? (similar to @zoharai's last question)

from nmt_gan.

ZhenYangIACAS avatar ZhenYangIACAS commented on May 28, 2024

@zoharai @borgr During the GAN training, you do not need to generate the dis_positive_data and dis_negative_data even for the first epoch. You can find the logic behind in the gan_train.py. For the evaluation, we stop training when the model achieves no improvement for the tenth evaluation on the dev set, which has been clarified in our paper. And we evaluate the model at regular intervals.

from nmt_gan.

xieexiaotuzi avatar xieexiaotuzi commented on May 28, 2024

@zoharai @ZhenYangIACAS Sorry to bother both of you. I am still confused about the evaluation. I can't find it where it is used. Could help me ?

from nmt_gan.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.