def train_with_ctr(self):
cfg = self.cfg
if cfg.dataset == 'codraw':
self.model.ctr.E.load_state_dict(torch.load('models/codraw_1.0_e.pt'))
elif cfg.dataset == 'iclevr':
self.model.ctr.E.load_state_dict(torch.load('models/iclevr_1.0_e.pt'))
iteration_counter = 0
for epoch in tqdm(range(self.cfg.epochs), ascii=True):
if cfg.dataset == 'codraw':
self.dataset.shuffle()
for batch in self.dataloader:
if iteration_counter >= 0 and iteration_counter % self.cfg.save_rate == 0:
torch.cuda.empty_cache()
evaluator = Evaluator.factory(self.cfg, self.visualizer,
self.logger)
res = evaluator.evaluate(iteration_counter)
print('\nIter %d:' % (iteration_counter))
print(res)
self.logger.write_res(iteration_counter, res)
del evaluator
iteration_counter += 1
self.model.train_batch_with_ctr(batch,
epoch,
iteration_counter,
self.visualizer,
self.logger)
def train(self):
iteration_counter = 0
for epoch in tqdm(range(self.cfg.epochs), ascii=True):
if cfg.dataset == 'codraw':
self.dataset.shuffle()
for batch in self.dataloader:
if iteration_counter >= 0 and iteration_counter % self.cfg.save_rate == 0:
torch.cuda.empty_cache()
evaluator = Evaluator.factory(self.cfg, self.visualizer,
self.logger)
res = evaluator.evaluate(iteration_counter)
print('\nIter %d:' % (iteration_counter))
print(res)
self.logger.write_res(iteration_counter, res)
del evaluator
iteration_counter += 1
self.model.train_batch(batch,
epoch,
iteration_counter,
self.visualizer,
self.logger)
def train(self):
iteration_counter = 0
#print("Total number of training data: {}".format(len(self.dataset)))
num_batches = len(self.dataloader)
total_iterations = num_batches * self.cfg.epochs
current_batch_time = 0 # Record the time it takes to process one batch
for epoch in range(self.cfg.epochs):
if cfg.dataset in ['codraw', 'codrawDialog', 'gandraw', "gandraw_clean", "gandraw_64", "gandraw_64_DA"]:
self.dataset.shuffle()
for batch in self.dataloader:
#iteration_counter += 1
current_batch_start = time.time()
self.model.train_batch(batch,
epoch,
iteration_counter,
self.visualizer,
self.logger,
total_iters=total_iterations,
current_batch_t=current_batch_time
)
current_batch_time = time.time() - current_batch_start
print("batch_time is: {}".format(current_batch_time))
if iteration_counter >= 0 and iteration_counter % self.cfg.save_rate == 0:
print("Run Evaluation")
torch.cuda.empty_cache()
#evaluator = TellerEvaluator(self.cfg, self.visualizer, self.logger)
evaluator = Evaluator.factory(self.cfg, self.visualizer,
self.logger)
evaluator.evaluate(iteration_counter, self.model)
del evaluator
iteration_counter += 1
def train(self):
iteration_counter = 0
for epoch in range(self.cfg.epochs):
if cfg.dataset == 'codraw':
self.dataset.shuffle()
for batch in self.dataloader:
if iteration_counter >= 0 and iteration_counter % self.cfg.save_rate == 0:
torch.cuda.empty_cache()
evaluator = Evaluator.factory(self.cfg, self.visualizer,
self.logger)
evaluator.evaluate(iteration_counter)
del evaluator
iteration_counter += 1
self.model.train_batch(batch, epoch, iteration_counter,
self.visualizer, self.logger)
def train(self):
iteration_counter = 0 # Last Iteration
best_saved_iteration = 0 #highest iteration that we have a saved model
best_scene_sim_score = 0
#print("Total number of training data: {}".format(len(self.dataset)))
num_batches = len(self.dataloader)
total_iterations = num_batches * self.cfg.epochs
current_batch_time = 0 # Record the time it takes to process one batch
#print("total iteration is: {}".format(total_iterations))
visualize_images = []
for epoch in range(self.cfg.epochs):
if cfg.dataset in [
'codraw', 'gandraw', "gandraw_64", "gandraw_64_DA"
]:
self.dataset.shuffle()
for batch in self.dataloader:
if cfg.gan_type == "recurrent_gan":
self.model.train_batch(batch, epoch, iteration_counter,
self.visualizer, self.logger)
else:
current_batch_start = time.time()
self.model.train_batch(batch,
epoch,
iteration_counter,
self.visualizer,
self.logger,
total_iters=total_iterations,
current_batch_t=current_batch_time)
current_batch_time = time.time() - current_batch_start
print("batch_time is: {}".format(current_batch_time))
if iteration_counter >= 0 and iteration_counter % self.cfg.save_rate == 0:
torch.cuda.empty_cache()
evaluator = Evaluator.factory(
self.cfg,
self.visualizer,
self.logger,
visualize_images=visualize_images)
metrics_report = evaluator.evaluate(iteration_counter)
print(
"evaluation results for iter: {} on validation data: \n"
.format(iteration_counter))
for key, value in metrics_report.items():
print("{metric_name}: {metric_value}; \n".format(
metric_name=key, metric_value=value))
#udpate the best scene_sim_score
if metrics_report['scene_sim_score'] > best_scene_sim_score:
best_scene_sim_score = metrics_report[
'scene_sim_score']
best_saved_iteration = iteration_counter
del evaluator
iteration_counter += 1
# if iteration_counter > 1:
#Evaluate on the test data
torch.cuda.empty_cache()
evaluator = Evaluator.factory(self.cfg,
self.visualizer,
self.logger,
visualize_images=visualize_images)
metrics_report = evaluator.evaluate(best_saved_iteration,
use_test=True)
print("best iteration is: {}".format(best_saved_iteration))
print("evaluation results for iter: {} on test data: \n".format(
best_saved_iteration))
for key, value in metrics_report.items():
print("{metric_name}: {metric_value}; \n".format(
metric_name=key, metric_value=value))
del evaluator
from geneva.data.datasets import DATASETS
from geneva.evaluation.evaluate import Evaluator
from geneva.utils.config import keys, parse_config
from geneva.utils.visualize import VisdomPlotter
from geneva.models.models import MODELS
from geneva.data import gandraw_dataset
import time
if __name__ == '__main__':
config_file = "example_args/gandraw_args.json"
with open(config_file, 'r') as f:
cfg = json.load(f)
cfg = easydict.EasyDict(cfg)
best_iteration = 1500 #Manually define
#Initialize the evaluator
visualizer = VisdomPlotter(env_name=cfg.exp_name, server=cfg.vis_server)
logger = None
evaluator = Evaluator.factory(cfg, visualizer, logger, visualize_images=[])
metric_report = evaluator.evaluate(best_iteration, use_test=True)
print("evaluation results for iter: {} on test data: \n".format(
best_iteration))
for key, value in metric_report.items():
print("{metric_name}: {metric_value}; \n".format(metric_name=key,
metric_value=value))