forked from le-schlo/EnTdecker
-
Notifications
You must be signed in to change notification settings - Fork 0
/
eval.py
144 lines (105 loc) · 4.52 KB
/
eval.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
import sys
import os
import json
import time
import torch
from pathlib import Path
import numpy as np
import pandas as pd
from EasyChemML.DataImport.DataImporter import DataImporter
from EasyChemML.DataImport.Module.CSV import CSV
from EasyChemML.Encoder import BertTokenizer, MolRdkitConverter
from EasyChemML.Environment import Environment
from EasyChemML.Encoder.impl_Tokenizer.SmilesTokenizer_SchwallerEtAll import SmilesTokenzier
from EasyChemML.Metrik.Module.DensityEvaluation import DensityMetrics
from EasyChemML.Model.impl_Pytorch.Models.BERT.FP2MOL_BERT_Trans import FP2MOL_Bert
# ----------------------------------- Data Preprocessing -----------------------
settings_path ="/Models/spin_population/settings.json"
with open(settings_path, "r") as settings_file:
setting_dict = json.load(settings_file)
dir_name = setting_dict.get("dir_name")
file_path = setting_dict.get("file_path")
device = "cuda:0"
d_model = 512
heads = 4
N = 8
src_vocab_size = setting_dict.get("src_vocab_size")
trg_vocab_size = 136
dropout = 0.1
max_seq_len = setting_dict.get("src_len")
model_object = FP2MOL_Bert(src_vocab_size, trg_vocab_size, N, heads, d_model, dropout, max_seq_len, device)
p_fname = "/Models/spin_population/spin_population.pt"
model_object.load_model_parameters(p_fname)
env = Environment(WORKING_path_addRelativ='Output')
dataloader = {'EnTdecker_data': CSV(file_path, columns=['SMILES', 'SMILES_SD'])}
di = DataImporter(env)
bp = di.load_data_InNewBatchPartition(dataloader)
val_data_size = len(bp['EnTdecker_data'])
print('Start BertTokenizer')
tokenizer = BertTokenizer()
tokenizer.convert(datatable=bp['EnTdecker_data'], columns=['SMILES', 'SMILES_SD'], n_jobs=4)
# ----------------------------------- Evaluation -----------------------
batch_size = setting_dict.get("batch_size")
print_every = setting_dict.get("print_every")
s = SmilesTokenzier()
total_loss = 0
positive_prediction = []
smi2smi = []
smi2smi_l = []
R2_perMol = []
R2_perMol_l = []
Ranked = []
Ranked_l = []
start_mol = 0
iter_steps = 0
start = time.time()
temp = start
iteration_per_chunk = val_data_size / batch_size
for i in range(int(iteration_per_chunk)):
end_mol = start_mol + batch_size
true_smiSD = bp['EnTdecker_data'][start_mol:end_mol]['SMILES_SD_ids']
input_smi = bp['EnTdecker_data'][start_mol:end_mol]['SMILES_ids']
input_smi = input_smi[:, 0:setting_dict.get("src_len")]
true_smi = bp['EnTdecker_data'][start_mol:end_mol]['SMILES']
true_smiSD = torch.LongTensor(true_smiSD).to(torch.device(device))
input_smi = torch.LongTensor(input_smi).to(torch.device(device))
model_eval = model_object.fit_eval(input_smi, true_smiSD, method='greedy')
loss, outputs = model_eval.loss, model_eval.outputs
total_loss += torch.Tensor.item(loss.data)
start_mol = end_mol
iter_steps += 1
for example in range(batch_size):
pred_smi, pred_SD_string, DensityArray = s.getSmilesfromoutputwithSD(outputs[example])
_, _, true_DensityArray = s.getSmilesfromoutputwithSD(true_smiSD[example])
if torch.equal(outputs[example], true_smiSD[example]):
positive_prediction.append(1)
else:
positive_prediction.append((0))
if pred_smi == true_smi[example].decode("utf-8"):
smi2smi.append(1)
else:
smi2smi.append(0)
num_highest = 10
Arrays4metric = DensityMetrics(true_DensityArray, DensityArray)
R2_perMol.append(Arrays4metric.PearsonR2_np())
Ranked.append(Arrays4metric.RankDensities(num_highest=num_highest))
if (i + 1) % print_every == 0:
accuracy = np.sum(positive_prediction) / (iter_steps * batch_size)
smi_acc = np.sum(smi2smi) / (iter_steps * batch_size)
R2_acc = np.sum(R2_perMol) / (iter_steps * batch_size)
Ran = np.sum(Ranked) / (iter_steps * batch_size)
print(f'exact accuracy = {accuracy: .3f}, ' f'smi2smi accuracy = {smi_acc: .3f}, '
f'Top {num_highest/2} ' f'Ranked Score = {Ran: .3f}, '
f'Average R2 = {R2_acc: .3f}, ' f'{(time.time() - temp): .3f}s per {print_every*batch_size} molecules')
temp = time.time()
positive_prediction.append(accuracy)
smi2smi_l.append(smi_acc)
R2_perMol_l.append(R2_acc)
Ranked_l.append(Ran)
loss_store_df = pd.DataFrame({'positive_prediction': positive_prediction,
'smi2smi': smi2smi_l,
'R2 per mol': R2_perMol_l,
'Ranked': Ranked_l,
})
loss_store_df.to_csv(dir_name + 'eval.csv')
print('well_done')