class SeqGAN(GenerativeAdversarialNet):
r"""SeqGAN is a generative adversarial network consisting of a generator and a discriminator.
Modeling the data generator as a stochastic policy in reinforcement learning (RL),
SeqGAN bypasses the generator differentiation problem by directly performing gradient policy update.
The RL reward signal comes from the GAN discriminator judged on a complete sequence,
and is passed back to the intermediate state-action steps using Monte Carlo search.
"""
def __init__(self, config, dataset):
super(SeqGAN, self).__init__(config, dataset)
self.generator = SeqGANGenerator(config, dataset)
self.discriminator = SeqGANDiscriminator(config, dataset)
self.pad_idx = dataset.padding_token_idx
def calculate_g_train_loss(self, corpus, epoch_idx):
return self.generator.calculate_loss(corpus)
def calculate_d_train_loss(self, real_data, fake_data, epoch_idx):
return self.discriminator.calculate_loss(real_data, fake_data)
def calculate_g_adversarial_loss(self, epoch_idx):
self.discriminator.eval()
loss = self.generator.adversarial_loss(self.discriminator.forward)
self.discriminator.train()
return loss
def calculate_nll_test(self, corpus, epoch_idx):
return self.generator.calculate_loss(corpus, nll_test=True)
def generate(self, eval_data):
return self.generator.generate(eval_data)
def sample(self, sample_num):
samples = self.generator.sample(sample_num)
return samples
def __init__(self, config, dataset):
super(SeqGAN, self).__init__(config, dataset)
self.generator = SeqGANGenerator(config, dataset)
self.discriminator = SeqGANDiscriminator(config, dataset)
self.pad_idx = dataset.padding_token_idx
def __init__(self, config, dataset):
super(SeqGAN, self).__init__(config, dataset)
self.generator = SeqGANGenerator(config, dataset)
self.discriminator = SeqGANDiscriminator(config, dataset)