def main():
global args
use_cuda = torch.cuda.is_available()
initialize_environment(random_seed=cfg.RNG_SEED, use_cuda=use_cuda)
args = parser.parse_args()
datadir = args.db_dir
outputdir = get_output_dir(args.db_dir)
nepoch = args.nepoch
step = args.step_epoch
dropout = args.dropout
n_layers = cfg.N_LAYERS
input_dim = cfg.INPUT_DIM
hidden_dims = cfg.HIDDEN_DIMS
# logging information
loggin_dir = os.path.join(outputdir, 'runs', 'pretraining')
if not os.path.exists(loggin_dir):
os.makedirs(loggin_dir)
tensorboard_logger.configure(os.path.join(loggin_dir, '%s' % (args.id)))
trainset = EncodedTextDataset(root=datadir, train=True)
testset = EncodedTextDataset(root=datadir, train=False)
kwargs = {'num_workers': 0, 'pin_memory': True} if use_cuda else {}
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=args.batchsize,
shuffle=True,
**kwargs)
testloader = torch.utils.data.DataLoader(testset,
batch_size=100,
shuffle=True,
**kwargs)
pretrain(
outputdir, {
'nlayers': n_layers,
'dropout': dropout,
'reluslope': 0.0,
'nepoch': nepoch,
'lrate': [args.lr],
'wdecay': [0.0],
'step': step,
'input_dim': input_dim,
'hidden_dims': hidden_dims
}, use_cuda, trainloader, testloader)
def check_prelim():
"""Initial setup code. Eventually this will set options."""
directory_structure = utils.initialize_environment(config)
return directory_structure
return args
args = get_args()
# n_clusters = 4
# data_dir = 'data/ag_news/'
data_dir = args.data_dir
n_clusters = args.n_clusters
use_cuda = torch.cuda.is_available()
random_seed = args.seed
recons_lam = args.recons_lam
cluster_lam = args.cluster_lam
batch_size = args.batch_size
tol = args.tol
lr = args.lr
initialize_environment(random_seed=random_seed, use_cuda=use_cuda)
feat_path = os.path.join(data_dir, cfg.TRAIN_TEXT_FEAT_FILE_NAME)
feat, labels, ids = load_feat(feat_path)
outputdir = get_output_dir(data_dir)
net_filename = os.path.join(outputdir, cfg.PRETRAINED_FAE_FILENAME)
checkpoint = torch.load(net_filename)
net = extract_sdae_model(input_dim=cfg.INPUT_DIM,
hidden_dims=cfg.HIDDEN_DIMS)
net.load_state_dict(checkpoint['state_dict'])
if use_cuda:
net.cuda()
dcn = DCN(n_clusters,
net,
cfg.HIDDEN_DIMS[-1],
default=100,
help='the number of seed for each class')
parser.add_argument('--verbose',
help='whether to print log',
action='store_true')
args = parser.parse_args()
return args
args = get_args()
data_dir = args.data_dir
random_seed = args.seed
seed_num = args.seed_num
verbose = args.verbose
initialize_environment(random_seed=random_seed)
_, labels, ids = load_csv_corpus(
os.path.join(data_dir, cfg.TRAIN_DATA_NAME + '.csv'))
dic = defaultdict(list)
for tmp_id, tmp_label in zip(ids, labels):
dic[tmp_label].append(tmp_id)
results = []
for l, tmp_ids in dic.items():
random.shuffle(tmp_ids)
tmp_ids = tmp_ids[:seed_num]
results.extend([(tmp_id, l) for tmp_id in tmp_ids])
results.sort()
else:
with bz2.open(memory_path, 'rb') as zipped_pickle_file:
return pickle.load(zipped_pickle_file)
def save_memory(memory, memory_path, disable_bzip):
if disable_bzip:
with open(memory_path, 'wb') as pickle_file:
pickle.dump(memory, pickle_file)
else:
with bz2.open(memory_path, 'wb') as zipped_pickle_file:
pickle.dump(memory, zipped_pickle_file)
# Environment
env, test_env = initialize_environment(args)
n_actions = env.action_space.n
# Agent
dqn = Agent(args, env)
# If a model is provided, and evaluate is fale, presumably we want to resume, so try to load memory
if args.model is not None and not args.evaluate:
if not args.memory:
raise ValueError('Cannot resume training without memory save path. Aborting...')
elif not os.path.exists(args.memory):
raise ValueError(
'Could not find memory file at {path}. Aborting...'.format(path=args.memory))
mem = load_memory(args.memory, args.disable_bzip_memory)
def check_prelim():
"""Initial setup code. Eventually this will set options."""
directory_structure = utils.initialize_environment(config)
return directory_structure