- PyTorch v0.3
- Python 3.6
- CUDA 9.1 (For GPU)
The idea is from paper SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient. We build on SeqGAN and compare three gradient estimators for sequence generation with GANs: REINFORCE (as in SeqGAN), and state-of-the-art REBAR (https://arxiv.org/pdf/1703.07370.pdf) and RELAX (https://arxiv.org/pdf/1711.00123.pdf).
The code is rewrited in PyTorch with the structure derived from (https://github.com/LantaoYu/SeqGAN)
$ python main.py
After runing this file, the results will be printed on terminal. You can change the parameters in the main.py.
Using CUDA
Pass in the gpu device number for e.g. 0
$ python main.py --cude {GPU_DEVICE_NUMBER}
Enable Visualization
Start the server (probably in a screen or tmux):
python -m visdom.server -port 8097Run with --visualize parameter
$ python main.py --cude {GPU_DEVICE_NUMBER} --visualize
