def execute():
from tutils import trans_args, trans_init, load_yaml, dump_yaml
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("--config", default="ablation.yaml")
args = trans_args(parser)
logger, config = trans_init(args)
ablation_trainer = AblationTrainer(logger, config)
ablation_trainer.run()
def template(_file_name):
from tutils import trans_args, trans_init, load_yaml, dump_yaml
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("--config", default="ablation.yaml")
args = trans_args(parser)
logger, config = trans_init(args)
if config['ablation']['is']:
# Check opts to do ablation
ablation_trainer = AblationTrainer(logger, config)
ablation_trainer.run_train()
ablation_trainer.run_test()
def usage():
from tutils import trans_args, trans_init, print_dict
args = trans_args()
print(args)
logger, config = trans_init(args)
print(" ---------------------------------------------------------")
print_dict(config)
metriclogger = MetricLogger(logger=None)
for data in metriclogger.log_every(range(20),
print_freq=2,
header="[trans]"):
i = 0
metriclogger.update(**{"ind": i, "data": data})
results = metriclogger.return_final_dict()
**config['training'])
if args.pretrain:
model.load()
model.cuda()
else:
trainer.fit(model, dataset)
def test(self, logger, config, args):
model = Learner(config, logger)
epoch = args.epoch
pth = tfilename(config['runs_dir'], f"model_epoch_{epoch}.pth")
model.load(pth)
model.cuda()
tester_train = Tester(logger, config, mode="Train")
tester_test = Tester(logger, config, mode="Test1+2")
logger.info(f"Dataset Training")
tester_train.test(model)
logger.info(f"Dataset Test 1+2")
tester_test.test(model)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
args = trans_args(parser)
logger, config = trans_init(args)
save_script(config['base']['runs_dir'], __file__)
function_manager = FunctionManager()
# getattr(function_manager, args.func)(logger, config, args)
function_manager.run_function(logger, config, args)