-
Notifications
You must be signed in to change notification settings - Fork 1
/
run_gegl_constrained.py
158 lines (138 loc) · 6.76 KB
/
run_gegl_constrained.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
import random
import json
import argparse
import os
import numpy as np
import torch
from torch.utils.data import DataLoader
from torch.optim import Adam
from runner.gegl_trainer import GeneticExpertGuidedLearningTrainer
from runner.guacamol_generator import GeneticExpertGuidedLearningGenerator
from model.neural_apprentice import SmilesGenerator, SmilesGeneratorHandler
from model.genetic_expert import GeneticOperatorHandler
from util.storage.priority_queue import MaxRewardPriorityQueue
from util.storage.recorder import Recorder
from util.chemistry.benchmarks import (
similarity_constrained_penalized_logp_atomrings,
similarity_constrained_penalized_logp_cyclebasis,
penalized_logp_atomrings,
penalized_logp_cyclebasis,
TanimotoScoringFunction,
)
from util.smiles.char_dict import SmilesCharDictionary
from util.smiles.dataset import load_dataset
from util.selfiesutil.load_function import load_genetic_experts
import neptune
import sys
sys.stdout = None
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="", formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
parser.add_argument("--smi_id_min", type=int, default=441)
parser.add_argument("--smi_id_max", type=int, default=1350)
parser.add_argument("--dataset", type=str, default="zinc")
#parser.add_argument("--dataset_path", type=str, default="./resource/data/zinc/logp_800.txt")
parser.add_argument("--dataset_path", type=str, default="./resource/data/zinc/2rgp_1350.txt")
parser.add_argument("--max_smiles_length", type=int, default=120)
parser.add_argument("--similarity_threshold", type=float, default=0.4)#0.4/0.6
parser.add_argument("--apprentice_load_dir", type=str, default="")
parser.add_argument("--learning_rate", type=float, default=1e-3)
parser.add_argument("--sample_batch_size", type=int, default=512)
parser.add_argument("--optimize_batch_size", type=int, default=256)
parser.add_argument("--mutation_rate", type=float, default=0.01)
parser.add_argument("--num_steps", type=int, default=50)
parser.add_argument("--num_keep", type=int, default=1024)
parser.add_argument("--max_sampling_batch_size", type=int, default=1024)
parser.add_argument("--apprentice_sampling_batch_size", type=int, default=1024)
parser.add_argument("--expert_sampling_batch_size", type=int, default=1024)
parser.add_argument("--apprentice_training_batch_size", type=int, default=256)
parser.add_argument("--num_apprentice_training_steps", type=int, default=4)
parser.add_argument("--num_jobs", type=int, default=8)
parser.add_argument("--record_filtered", action="store_true")
parser.add_argument("--use_atomrings", action="store_true")
parser.add_argument("--genetic_experts", type=str, nargs="+", default=["SELFIES"])
args = parser.parse_args()
args.algorithm = "gegl_constrained"
random.seed(0)
device = torch.device(0)
neptune.init(project_qualified_name="",
api_token='',
)
experiment = neptune.create_experiment(name=args.algorithm, params=vars(args))
neptune.append_tag(
f"{args.smi_id_min:03d}_{args.smi_id_max:03d}_{args.similarity_threshold}".replace(".", "")
)
char_dict = SmilesCharDictionary(dataset=args.dataset, max_smi_len=args.max_smiles_length)
dataset = load_dataset(char_dict=char_dict, smi_path=args.dataset_path)
if args.use_atomrings:
similarity_constrained_penalized_logp = similarity_constrained_penalized_logp_atomrings
penalized_logp_score_func = penalized_logp_atomrings().wrapped_objective.score
else:
similarity_constrained_penalized_logp = similarity_constrained_penalized_logp_cyclebasis
penalized_logp_score_func = penalized_logp_cyclebasis().wrapped_objective.score
for smi_id in range(args.smi_id_min, args.smi_id_max):
print(f"ID: {smi_id}")
reference_smi = dataset[smi_id]
benchmark = similarity_constrained_penalized_logp(
smiles=reference_smi, name=str(smi_id), threshold=args.similarity_threshold
)
scoring_num_list = [1]
apprentice_storage = MaxRewardPriorityQueue()
expert_storage = MaxRewardPriorityQueue()
apprentice = SmilesGenerator.load(load_dir=args.apprentice_load_dir)
apprentice = apprentice.to(device)
apprentice.train()
apprentice_optimizer = Adam(apprentice.parameters(), lr=args.learning_rate)
apprentice_handler = SmilesGeneratorHandler(
model=apprentice,
optimizer=apprentice_optimizer,
char_dict=char_dict,
max_sampling_batch_size=args.max_sampling_batch_size,
)
expert_handler = load_genetic_experts(
args.genetic_experts,
args=args,
)
#expert_handler = GeneticOperatorHandler(mutation_rate=args.mutation_rate)
trainer = GeneticExpertGuidedLearningTrainer(
apprentice_storage=apprentice_storage,
expert_storage=expert_storage,
apprentice_handler=apprentice_handler,
expert_handler=expert_handler,
char_dict=char_dict,
num_keep=args.num_keep,
apprentice_sampling_batch_size=args.apprentice_sampling_batch_size,
expert_sampling_batch_size=args.expert_sampling_batch_size,
apprentice_training_batch_size=args.apprentice_training_batch_size,
num_apprentice_training_steps=args.num_apprentice_training_steps,
init_smis=[reference_smi],
)
recorder = Recorder(scoring_num_list=scoring_num_list, record_filtered=args.record_filtered)
exp_generator = GeneticExpertGuidedLearningGenerator(
trainer=trainer,
recorder=recorder,
num_steps=args.num_steps,
device=device,
scoring_num_list=scoring_num_list,
num_jobs=args.num_jobs,
)
result = benchmark.assess_model(exp_generator)
optimized_smi, score = result.optimized_molecules[0]
print('********')
print(optimized_smi)
print('********')
output_path = 'out.txt'
with open(output_path, 'w', encoding='utf-8') as file1:
print(optimized_smi, file=file1)
print('\n')
reference_score = penalized_logp_score_func(reference_smi)
optimized_score = penalized_logp_score_func(optimized_smi)
similarity = TanimotoScoringFunction(target=reference_smi, fp_type="ECFP4").score(
optimized_smi
)
neptune.log_metric("id", smi_id)
neptune.log_text("reference_smi", reference_smi)
neptune.log_metric("reference_penalized_logp_score", reference_score)
neptune.log_metric("optimized_penalized_logp_score", optimized_score)
neptune.log_metric("similarity", similarity)