def data_provider(k, flags):
tokens = get_default_tokens()
demo_data = GeneratorData(training_data_path=flags.demo_file,
delimiter='\t',
cols_to_read=[0],
keep_header=True,
pad_symbol=' ',
max_len=120,
tokens=tokens,
use_cuda=use_cuda)
unbiased_data = GeneratorData(training_data_path=flags.unbiased_file,
delimiter='\t',
cols_to_read=[0],
keep_header=True,
pad_symbol=' ',
max_len=120,
tokens=tokens,
use_cuda=use_cuda)
prior_data = GeneratorData(training_data_path=flags.prior_data,
delimiter='\t',
cols_to_read=[0],
keep_header=True,
pad_symbol=' ',
max_len=120,
tokens=tokens,
use_cuda=use_cuda)
return {
'demo_data': demo_data,
'unbiased_data': unbiased_data,
'prior_data': prior_data
}
def data_provider(k, flags):
tokens = get_default_tokens()
gen_data = GeneratorData(training_data_path=flags.data_file,
delimiter='\t',
cols_to_read=[0],
keep_header=True,
pad_symbol=' ',
max_len=120,
tokens=tokens,
use_cuda=use_cuda)
return {"train": gen_data, "val": gen_data, "test": gen_data}
help='The directory to save the created dataset')
parser.add_argument('--filename',
type=str,
default='drd2_active.smi',
help='The filename for the created dataset')
args = parser.parse_args()
assert (os.path.exists(args.svc))
assert (os.path.exists(args.data))
assert (0 < args.threshold < 1)
# Load file containing SMILES
gen_data = GeneratorData(training_data_path=args.data,
delimiter='\t',
cols_to_read=[0],
keep_header=True,
pad_symbol=' ',
max_len=120,
tokens=get_default_tokens(),
use_cuda=False)
# Load classifier
clf = DRD2Model(args.svc)
# Screen SMILES in data file and write active compounds to file.
os.makedirs(args.save_dir, exist_ok=True)
num_active = 0
with open(os.path.join(args.save_dir, args.filename), 'w') as f:
for i in trange(gen_data.file_len, desc='Screening compounds...'):
smiles = gen_data.file[i][1:-1]
p = clf(smiles)
if p >= args.threshold:
from irelease.data import GeneratorData
from irelease.env import MoleculeEnv
from irelease.model import Encoder, PositionalEncoding, StackDecoderLayer, LinearOut, StackRNN, RNNLinearOut, RewardNetRNN
from irelease.reward import RewardFunction
from irelease.rl import PolicyAgent, MolEnvProbabilityActionSelector, REINFORCE, GuidedRewardLearningIRL, \
StateActionProbRegistry
from irelease.stackrnn import StackRNNCell
from irelease.utils import init_hidden, init_stack, get_default_tokens, init_hidden_2d, init_stack_2d, init_cell, seq2tensor
gen_data_path = '../data/chembl_xsmall.smi'
tokens = get_default_tokens()
# print(f'Number of tokens = {len(tokens)}')
gen_data = GeneratorData(training_data_path=gen_data_path,
delimiter='\t',
cols_to_read=[0],
keep_header=True,
tokens=tokens,
tokens_reload=True)
bz = 32
class MyTestCase(unittest.TestCase):
def test_batch(self):
batch = gen_data.random_training_set(batch_size=bz)
assert (len(batch[0]) == bz and len(batch[1]) == bz)
def test_embeddings(self):
x, y = gen_data.random_training_set(batch_size=bz)
encoder = Encoder(gen_data.n_characters, 128,
gen_data.char2idx[gen_data.pad_symbol])