def initialize(args):
sys.setrecursionlimit(10000)
lg = rdkit.RDLogger.logger()
lg.setLevel(rdkit.RDLogger.CRITICAL)
torch.manual_seed(args.seed)
torch.cuda.set_device(args.device)
arg_info = '_%s_LR_%f_HS_%d_RS_%d_AR_%.2f_AI_%d_%s' % (
args.task_tag, args.lr, args.hidden_size, args.rand_size,
args.anneal_rate, args.anneal_interval,
'share_embedding' if args.share_embedding else 'not_share_embedding')
LOG.init(file_name=current_time + '_' + arg_info)
logger = logging.getLogger('logger')
logger.info(args)
# create the tensorboard log saved folder
if not os.path.isdir(args.tensorboard_save_dir):
os.makedirs(args.tensorboard_save_dir)
# set the tensorboard writer
train_tb_log_dir = os.path.join(args.tensorboard_save_dir,
current_time + '_' + arg_info + '_train')
tb_suffix = '_' + arg_info
train_writer = SummaryWriter(log_dir=train_tb_log_dir,
filename_suffix=tb_suffix)
# create the model saved folder
if args.model_save_dir is not None:
# args.model_save_dir = os.path.join(args.model_save_dir, current_time + '_' + arg_info)
args.model_save_dir = os.path.join(args.model_save_dir,
f'{args.task_tag}')
if not os.path.isdir(args.model_save_dir):
os.makedirs(args.model_save_dir)
# save the model config
with open(os.path.join(args.model_save_dir, 'model_config.json'),
'w') as f:
json.dump(
{
'vocab': args.vocab,
'hidden_size': args.hidden_size,
'rand_size': args.rand_size,
'share_embedding': args.share_embedding,
'use_molatt': args.use_molatt,
'depthT': args.depthT,
'depthG': args.depthG
}, f)
return logger, train_writer
def initialize(args):
current_time = '{:%Y-%m-%d-%H-%M-%S}'.format(datetime.now())
sys.setrecursionlimit(10000)
lg = rdkit.RDLogger.logger()
lg.setLevel(rdkit.RDLogger.CRITICAL)
arg_info = '%s_%s_HS_%d_RS_%d' % (
args.task_tag,
args.metric_type,
args.hidden_size,
args.rand_size
)
LOG.init(file_name=current_time + '_Evaluation' + '_' + arg_info)
logger = logging.getLogger('logger')
logger.info(args)
torch.cuda.set_device(args.device)
torch.manual_seed(args.seed)
return logger