/SeqGAN-vs-MLE-vs-PG-BLEU-vs-ScheduleSampling

Primary LanguagePython

SeqGAN-vs-MLE-vs-PG-BLEU-vs-ScheduledSampling-PyTorch

A implementation of SeqGAN, MLE, PG-BLEU and Scheduled Sampling in PyTorch. We compare the methods on synthetic dataset, and real dataset consisting of Barack Obama speeches.

Tested with:

  • PyTorch v1 Stable
  • Python 3.6
  • CUDA at least 8.0 (For GPU)

Origin

The idea is from paper [SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient] (https://arxiv.org/pdf/1609.05473.pdf)

The code is written in PyTorch with SeqGAN is based on the implementation from https://github.com/ZiJianZhao/SeqGAN-PyTorch

Runing

$ python main.py