def __init__(self, config, ckeck_point):
self.config = config
self.epochs = config.epochs
self.model_type = config.model
self.ckpt = ckeck_point
self.tensorboard = config.tensorboard
if self.tensorboard:
self.summary_writer = SummaryWriter(self.ckpt.log_dir, 300)
# set model, criterion, optimizer
self.model = Model(config)
summary(self.model, config_file=self.ckpt.config_file)
# set criterion, optimizer
self.criterion = Loss(config)
self.optimizer = make_optimizer(config, self.model)
# load ckpt, model, optimizer
if (self.ckpt.exp_load is not None) or (not config.is_train):
print("Loading model... ")
self.load(self.ckpt)
print(self.ckpt.last_epoch, self.ckpt.global_step)
def summary_text(text, l):
return summary(text, l)
use_noise=args['--use_noise'],
noise_sigma=float(
args['--noise_sigma']))
video_discriminator = build_discriminator(args['--video_discriminator'],
dim_categorical=dim_z_category,
n_channels=n_channels,
use_noise=args['--use_noise'],
noise_sigma=float(
args['--noise_sigma']))
if torch.cuda.is_available():
generator.cuda()
image_discriminator.cuda()
video_discriminator.cuda()
print('The number of parameters for Video disciminator is : {0}'.format(
count_parameters(video_discriminator)))
summary((3, 64, 64, 16), video_discriminator)
trainer = Trainer(image_loader,
video_loader,
int(args['--print_every']),
int(args['--batches']),
args['<log_folder>'],
use_cuda=torch.cuda.is_available(),
use_infogan=args['--use_infogan'],
use_categories=args['--use_categories'])
trainer.train(generator, image_discriminator, video_discriminator)
def summary(self, size=30):
return summary(self.get_markdown(), size)
def summary(self, size=30):
return summary(self.get_markdown(), size)
def summary_text(text, l):
return summary(text, l)