def main():
args = parse_args()
loader = load_val(config.val_path, args, distributed=False)
model = darknet().eval()
#model = torchvision.models.resnet101( pretrained=True ).eval()
checkpoint_path = os.path.join(config.checkpoint_path,
config.checkpoint_name)
load_checkpoint(model, checkpoint_path)
validate(loader, model, args)
from initializers.spn_initializer import SPNInitializer, ShapeAgnosticLoad
from metrics.base import STNCrossEntropy, STNAccuracy
from networks.svhn import SVHNMultiLineResNetNetwork
from utils.create_gif import make_gif
from utils.create_video import make_video
from utils.datatypes import Size
from operations.debug import *
from operations.ones import *
from operations.disable_shearing import *
from utils.plot_log import LogPlotter
if __name__ == '__main__':
parser = utils.parse_args()
args = parser.parse_args()
if args.send_bboxes and args.ip is None:
parser.print_usage()
raise ValueError("You must specify an upstream ip if you want to send the bboxes of each iteration")
time = datetime.datetime.now().isoformat()
args.log_dir = os.path.join(args.log_dir, "{}_{}".format(time, args.log_name))
args.log_file = os.path.join(args.log_dir, 'log')
image_size = Size(width=200, height=200)
source_shape = (args.batch_size, 1, image_size.height, image_size.width)
target_shape = Size(width=50, height=50)
# adjustable network parameters
import torch
from torchvision.models import resnet18
dataset_size = 1281216
def plot_lr( optim, args, hyper ):
lr_hist = []
batch_size = hyper.batch_size
n_per_epoch = int( dataset_size / batch_size )
print( "number of iterations per epoch:{}".format( n_per_epoch ) )
start_epoch = args.start_epoch - 1
end_epoch = start_epoch + args.epochs
for epoch in range( start_epoch, end_epoch ):
for i in range( n_per_epoch ):
niter = epoch * n_per_epoch + i
lr = adjust_learning_rate( optim, niter, hyper )
lr_hist.append( lr )
index = list( range( n_per_epoch * args.epochs ) )
plt.plot( index, lr_hist )
plt.show()
# create a dummy optimizer
args = parse_args()
hyper = HyperParams( args.__dict__ )
optim = torch.optim.SGD( resnet18().parameters(), lr=args.base_lr )
plot_lr( optim, args, hyper )
if args.model_file != None:
if os.path.exists(args.model_file):
print("Loading model from file: {}".format(args.model_file))
model.load_state_dict(torch.load(args.model_file))
else:
print("Model file does not exist, training from scratch...")
# Check for cuda usage
device = torch.device("cuda:0") if args.cuda else "cpu"
if not args.e:
train_net(
model,
batch_iter,
args.epochs,
optimizer,
args.val_freq,
device=device,
print_interval=args.print_interval,
)
if args.model_file != None:
torch.save(model.cpu().state_dict(), args.model_file)
else:
model = model.to(device)
eval_net(model, batch_iter, args.val_freq, target_vocab, device=device)
if __name__ == "__main__":
main(parse_args())
def main():
# --- argument parsing ---
(model_name, epochs, min_count, cores, checkpoint_every, cache_in_memory,
lowercase, _, args) = parse_args(default_model_name='d2v',
default_epochs=20)
# --- init logging ---
logger = init_logging(name=model_name,
basic=True,
to_file=True,
to_stdout=False)
log_args(logger, args)
input_dir = join(SMPL_PATH, 'dewiki')
model_dir = join(EMB_PATH, model_name)
if not exists(model_dir):
makedirs(model_dir)
logger.info('model dir: ' + model_dir)
t0 = time()
documents = Documents(input_dir=input_dir,
logger=logger,
lowercase=lowercase)
if cache_in_memory:
documents = list(documents)
gc.collect()
# Model initialization
logger.info('Initializing new model')
model = Doc2Vec(
vector_size=300,
window=15,
min_count=20,
sample=1e-5,
negative=5,
hs=0,
dm=0,
dbow_words=1,
dm_concat=0,
seed=42,
epochs=epochs,
workers=cores,
)
logger.info('Building vocab')
model.build_vocab(documents)
# Model Training
epoch_saver = EpochSaver(model_name, model_dir, checkpoint_every)
epoch_logger = EpochLogger(logger)
logger.info('Training {:d} epochs'.format(epochs))
model.train(
documents,
total_examples=model.corpus_count,
epochs=model.epochs,
report_delay=60,
callbacks=[epoch_logger, epoch_saver],
)
# saving model
file_path = join(model_dir, model_name)
logger.info('Writing model to ' + file_path)
model.callbacks = ()
model.save(file_path)
t1 = int(time() - t0)
logger.info("all done in {:02d}:{:02d}:{:02d}".format(
t1 // 3600, (t1 // 60) % 60, t1 % 60))
def main():
# --- argument parsing ---
(
model_name, epochs, min_count, cores, checkpoint_every,
cache_in_memory, lowercase, fasttext, args
) = parse_args(default_model_name='w2v_default', default_epochs=100)
# --- init logging ---
logger = init_logging(name=model_name, basic=True, to_file=True, to_stdout=False)
log_args(logger, args)
input_dir = join(SMPL_PATH, 'dewiki')
model_dir = join(EMB_PATH, model_name)
if not exists(model_dir):
makedirs(model_dir)
logger.info('model dir: ' + model_dir)
t0 = time()
if cache_in_memory:
# needs approx. 25GB of RAM
logger.info('cache data in memory')
sentences = [s for s in Sentences(input_dir, logger, lowercase=lowercase)]
else:
sentences = Sentences(input_dir, logger, use_file_cache=True, lowercase=lowercase)
gc.collect()
# Model initialization
logger.info('Initializing new model')
if fasttext:
model = FastText(
size=300,
window=5,
min_count=min_count,
sample=1e-5,
negative=5,
sg=1,
seed=42,
iter=epochs,
workers=cores,
min_n=3,
max_n=6,
)
else:
model = Word2Vec(
size=300,
window=5,
min_count=min_count,
sample=1e-5,
negative=5,
sg=1,
seed=42,
iter=epochs,
workers=cores,
)
logger.info('Building vocab')
model.build_vocab(sentences, progress_per=100_000)
# Model Training
epoch_saver = EpochSaver(model_name, model_dir, checkpoint_every)
epoch_logger = EpochLogger(logger)
logger.info('Training {:d} epochs'.format(epochs))
model.train(
sentences,
total_examples=model.corpus_count,
epochs=model.epochs,
report_delay=60,
callbacks=[epoch_logger, epoch_saver],
)
# saving model
file_path = join(model_dir, model_name)
logger.info('Writing model to ' + file_path)
model.callbacks = ()
model.save(file_path)
t1 = int(time() - t0)
logger.info("all done in {:02d}:{:02d}:{:02d}".format(t1//3600, (t1//60) % 60, t1 % 60))
from data_io.file_iter import FileBasedIter
from data_io.lstm_iter import LSTMIter, InitStateLSTMIter
from initializers.spn_initializer import SPNInitializer, ShapeAgnosticLoad
from metrics.base import STNCrossEntropy, STNAccuracy
from networks.svhn import SVHNMultiLineResNetNetwork
from utils.create_gif import make_gif
from utils.create_video import make_video
from utils.datatypes import Size
from operations.debug import *
from operations.ones import *
from operations.disable_shearing import *
from utils.plot_log import LogPlotter
if __name__ == '__main__':
parser = utils.parse_args()
args = parser.parse_args()
if args.send_bboxes and args.ip is None:
parser.print_usage()
raise ValueError(
"You must specify an upstream ip if you want to send the bboxes of each iteration"
)
time = datetime.datetime.now().isoformat()
args.log_dir = os.path.join(args.log_dir,
"{}_{}".format(time, args.log_name))
args.log_file = os.path.join(args.log_dir, 'log')
image_size = Size(width=200, height=200)
source_shape = (args.batch_size, 1, image_size.height, image_size.width)