def main(args):
env_info = get_environ_info()
places = fluid.CUDAPlace(ParallelEnv().dev_id) \
if env_info['Paddle compiled with cuda'] and env_info['GPUs used'] \
else fluid.CPUPlace()
if args.dataset not in DATASETS:
raise Exception(
'`--dataset` is invalid. it should be one of {}'.format(
str(list(DATASETS.keys()))))
dataset = DATASETS[args.dataset]
with fluid.dygraph.guard(places):
test_transforms = T.Compose([T.Resize(args.input_size), T.Normalize()])
test_dataset = dataset(dataset_root=args.dataset_root,
transforms=test_transforms,
mode='test')
model = manager.MODELS[args.model_name](
num_classes=test_dataset.num_classes)
infer(model,
model_dir=args.model_dir,
test_dataset=test_dataset,
save_dir=args.save_dir)
def main(args):
env_info = get_environ_info()
places = fluid.CUDAPlace(ParallelEnv().dev_id) \
if env_info['place'] == 'cuda' and fluid.is_compiled_with_cuda() \
else fluid.CPUPlace()
if args.dataset not in DATASETS:
raise Exception(
'`--dataset` is invalid. it should be one of {}'.format(
str(list(DATASETS.keys()))))
dataset = DATASETS[args.dataset]
with fluid.dygraph.guard(places):
eval_transforms = T.Compose([T.Resize(args.input_size), T.Normalize()])
eval_dataset = dataset(dataset_root=args.dataset_root,
transforms=eval_transforms,
mode='val')
if args.model_name not in MODELS:
raise Exception(
'`--model_name` is invalid. it should be one of {}'.format(
str(list(MODELS.keys()))))
model = MODELS[args.model_name](num_classes=eval_dataset.num_classes)
evaluate(model,
eval_dataset,
model_dir=args.model_dir,
num_classes=eval_dataset.num_classes)
def main(args):
env_info = get_environ_info()
places = fluid.CUDAPlace(ParallelEnv().dev_id) \
if env_info['place'] == 'cuda' and fluid.is_compiled_with_cuda() \
else fluid.CPUPlace()
if args.dataset not in DATASETS:
raise Exception(
'`--dataset` is invalid. it should be one of {}'.format(
str(list(DATASETS.keys()))))
dataset = DATASETS[args.dataset]
with fluid.dygraph.guard(places):
# Creat dataset reader
train_transforms = T.Compose([
T.RandomHorizontalFlip(0.5),
T.ResizeStepScaling(0.5, 2.0, 0.25),
T.RandomPaddingCrop(args.input_size),
T.RandomDistort(),
T.Normalize(),
])
train_dataset = dataset(dataset_root=args.dataset_root,
transforms=train_transforms,
mode='train')
eval_dataset = None
if args.do_eval:
eval_transforms = T.Compose([T.Normalize()])
eval_dataset = dataset(dataset_root=args.dataset_root,
transforms=eval_transforms,
mode='val')
if args.model_name not in MODELS:
raise Exception(
'`--model_name` is invalid. it should be one of {}'.format(
str(list(MODELS.keys()))))
model = MODELS[args.model_name](num_classes=train_dataset.num_classes)
# Creat optimizer
# todo, may less one than len(loader)
num_steps_each_epoch = len(train_dataset) // (args.batch_size *
ParallelEnv().nranks)
decay_step = args.num_epochs * num_steps_each_epoch
lr_decay = fluid.layers.polynomial_decay(args.learning_rate,
decay_step,
end_learning_rate=0,
power=0.9)
optimizer = fluid.optimizer.Momentum(
lr_decay,
momentum=0.9,
parameter_list=model.parameters(),
regularization=fluid.regularizer.L2Decay(
regularization_coeff=4e-5))
train(model,
train_dataset,
places=places,
eval_dataset=eval_dataset,
optimizer=optimizer,
save_dir=args.save_dir,
num_epochs=args.num_epochs,
batch_size=args.batch_size,
pretrained_model=args.pretrained_model,
resume_model=args.resume_model,
save_interval_epochs=args.save_interval_epochs,
log_steps=args.log_steps,
num_classes=train_dataset.num_classes,
num_workers=args.num_workers,
use_vdl=args.use_vdl)
def main(args):
env_info = get_environ_info()
info = ['{}: {}'.format(k, v) for k, v in env_info.items()]
info = '\n'.join(['\n', format('Environment Information', '-^48s')] +
info + ['-' * 48])
logger.info(info)
places = fluid.CUDAPlace(ParallelEnv().dev_id) \
if env_info['Paddle compiled with cuda'] and env_info['GPUs used'] \
else fluid.CPUPlace()
if args.dataset not in DATASETS:
raise Exception(
'`--dataset` is invalid. it should be one of {}'.format(
str(list(DATASETS.keys()))))
dataset = DATASETS[args.dataset]
with fluid.dygraph.guard(places):
# Creat dataset reader
train_transforms = T.Compose([
T.RandomHorizontalFlip(0.5),
T.ResizeStepScaling(0.5, 2.0, 0.25),
T.RandomPaddingCrop(args.input_size),
T.RandomDistort(),
T.Normalize(),
])
train_dataset = dataset(dataset_root=args.dataset_root,
transforms=train_transforms,
mode='train')
eval_dataset = None
if args.do_eval:
eval_transforms = T.Compose([T.Normalize()])
eval_dataset = dataset(dataset_root=args.dataset_root,
transforms=eval_transforms,
mode='val')
model = manager.MODELS[args.model_name](
num_classes=train_dataset.num_classes,
pretrained_model=args.pretrained_model)
# Creat optimizer
# todo, may less one than len(loader)
num_iters_each_epoch = len(train_dataset) // (args.batch_size *
ParallelEnv().nranks)
lr_decay = fluid.layers.polynomial_decay(args.learning_rate,
args.iters,
end_learning_rate=0,
power=0.9)
optimizer = fluid.optimizer.Momentum(
lr_decay,
momentum=0.9,
parameter_list=model.parameters(),
regularization=fluid.regularizer.L2Decay(
regularization_coeff=4e-5))
train(model,
train_dataset,
places=places,
eval_dataset=eval_dataset,
optimizer=optimizer,
save_dir=args.save_dir,
iters=args.iters,
batch_size=args.batch_size,
resume_model=args.resume_model,
save_interval_iters=args.save_interval_iters,
log_iters=args.log_iters,
num_classes=train_dataset.num_classes,
num_workers=args.num_workers,
use_vdl=args.use_vdl)