-
Notifications
You must be signed in to change notification settings - Fork 6
/
train.py
161 lines (132 loc) · 6.15 KB
/
train.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
159
160
161
import argparse
import pandas as pd
import numpy as np
from functools import partial
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from fastai.callbacks import SaveModelCallback
from fastai.basic_data import DataBunch, DeviceDataLoader, DatasetType
from fastai.basic_train import Learner
from fastai.train import *
from fastai.distributed import *
from moldataset import MoleculeDataset, collate_parallel_fn
from model import Transformer
from utils import scale_features, set_seed, store_submit, store_oof
from callbacks import GradientClipping, GroupMeanLogMAE
from losses_and_metrics import rmse, mae, contribs_rmse_loss
import constants as C
# parse arguments
parser = argparse.ArgumentParser()
parser.add_argument('--batch_size', type=int, default=10)
parser.add_argument('--epochs', type=int, default=100)
parser.add_argument('--lr', type=float, default=4e-5, help='learning rate')
parser.add_argument('--d_model', type=int, default=650,
help='dimenstion of node state vector')
parser.add_argument('--N', type=int, default=10,
help='number of encoding layers')
parser.add_argument('--h', type=int, default=10,
help='number of attention heads')
parser.add_argument('--wd', type=float, default=1e-2, help='weight decay')
parser.add_argument('--dropout', type=float, default=0.0)
parser.add_argument('--start_epoch', type=int, default=0)
parser.add_argument('--fold_id', type=int, default=1)
parser.add_argument('--version', type=int, default=1)
parser.add_argument('--local_rank', type=int)
args = parser.parse_args()
# check if distributed training is possible and set model description
distributed_train = torch.cuda.device_count() > 1
model_str = f'mol_transformer_v{args.version}_fold{args.fold_id}'
# import data
train_df = pd.read_csv(C.PROC_DATA_PATH+'train_proc_df.csv', index_col=0)
test_df = pd.read_csv(C.PROC_DATA_PATH+'test_proc_df.csv', index_col=0)
atom_df = pd.read_csv(C.PROC_DATA_PATH+'atom_df.csv', index_col=0)
bond_df = pd.read_csv(C.PROC_DATA_PATH+'bond_df.csv', index_col=0)
angle_in_df = pd.read_csv(C.PROC_DATA_PATH+'angle_in_df.csv', index_col=0)
angle_out_df = pd.read_csv(C.PROC_DATA_PATH+'angle_out_df.csv', index_col=0)
graph_dist_df = pd.read_csv(
C.PROC_DATA_PATH+'graph_dist_df.csv', index_col=0, dtype=np.int32)
structures_df = pd.read_csv(
C.PROC_DATA_PATH+'structures_proc_df.csv', index_col=0)
train_mol_ids = pd.read_csv(C.PROC_DATA_PATH+'train_idxs_8_fold_cv.csv',
usecols=[0, args.fold_id], index_col=0
).dropna().astype(int).iloc[:,0]
val_mol_ids = pd.read_csv(C.PROC_DATA_PATH+'val_idxs_8_fold_cv.csv',
usecols=[0, args.fold_id], index_col=0
).dropna().astype(int).iloc[:,0]
test_mol_ids = pd.Series(test_df['molecule_id'].unique())
# scale features
train_df, sc_feat_means, sc_feat_stds = scale_features(
train_df, C.SC_FEATS_TO_SCALE, train_mol_ids, return_mean_and_std=True)
test_df = scale_features(
test_df, C.SC_FEATS_TO_SCALE, means=sc_feat_means, stds=sc_feat_stds)
atom_df = scale_features(atom_df, C.ATOM_FEATS_TO_SCALE, train_mol_ids)
bond_df = scale_features(bond_df, C.BOND_FEATS_TO_SCALE, train_mol_ids)
# group data by molecule id
gb_mol_sc = train_df.groupby('molecule_id')
test_gb_mol_sc = test_df.groupby('molecule_id')
gb_mol_atom = atom_df.groupby('molecule_id')
gb_mol_bond = bond_df.groupby('molecule_id')
gb_mol_struct = structures_df.groupby('molecule_id')
gb_mol_angle_in = angle_in_df.groupby('molecule_id')
gb_mol_angle_out = angle_out_df.groupby('molecule_id')
gb_mol_graph_dist = graph_dist_df.groupby('molecule_id')
# create dataloaders and fastai DataBunch
set_seed(100)
train_ds = MoleculeDataset(
train_mol_ids, gb_mol_sc, gb_mol_atom, gb_mol_bond, gb_mol_struct,
gb_mol_angle_in, gb_mol_angle_out, gb_mol_graph_dist
)
val_ds = MoleculeDataset(
val_mol_ids, gb_mol_sc, gb_mol_atom, gb_mol_bond, gb_mol_struct,
gb_mol_angle_in, gb_mol_angle_out, gb_mol_graph_dist
)
test_ds = MoleculeDataset(
test_mol_ids, test_gb_mol_sc, gb_mol_atom, gb_mol_bond, gb_mol_struct,
gb_mol_angle_in, gb_mol_angle_out, gb_mol_graph_dist
)
train_dl = DataLoader(train_ds, args.batch_size, shuffle=True, num_workers=8)
val_dl = DataLoader(val_ds, args.batch_size, num_workers=8)
test_dl = DeviceDataLoader.create(
test_ds, args.batch_size, num_workers=8,
collate_fn=partial(collate_parallel_fn, test=True)
)
db = DataBunch(train_dl, val_dl, collate_fn=collate_parallel_fn)
db.test_dl = test_dl
# set up model
set_seed(100)
d_model = args.d_model
enn_args = dict(layers=3*[d_model], dropout=3*[0.0], layer_norm=True)
ann_args = dict(layers=1*[d_model], dropout=1*[0.0], layer_norm=True,
out_act=nn.Tanh())
model = Transformer(
C.N_ATOM_FEATURES, C.N_BOND_FEATURES, C.N_SC_EDGE_FEATURES,
C.N_SC_MOL_FEATURES, N=args.N, d_model=d_model, d_ff=d_model*4,
d_ff_contrib=d_model//4, h=args.h, dropout=args.dropout,
kernel_sz=min(128, d_model), enn_args=enn_args, ann_args=ann_args
)
# initialize distributed
if distributed_train:
torch.cuda.set_device(args.local_rank)
torch.distributed.init_process_group(backend='nccl', init_method='env://')
# train model
callback_fns = [
partial(GradientClipping, clip=10), GroupMeanLogMAE,
partial(SaveModelCallback, every='improvement', mode='min',
monitor='group_mean_log_mae', name=model_str)
]
learn = Learner(db, model, metrics=[rmse, mae], callback_fns=callback_fns,
wd=args.wd, loss_func=contribs_rmse_loss)
if args.start_epoch > 0:
learn.load(model_str)
torch.cuda.empty_cache()
if distributed_train: learn = learn.to_distributed(args.local_rank)
learn.fit_one_cycle(args.epochs, max_lr=args.lr, start_epoch=args.start_epoch)
# make predictions
val_contrib_preds = learn.get_preds(DatasetType.Valid)
test_contrib_preds = learn.get_preds(DatasetType.Test)
val_preds = val_contrib_preds[0][:,-1].detach().numpy() * C.SC_STD + C.SC_MEAN
test_preds = test_contrib_preds[0][:,-1].detach().numpy() * C.SC_STD + C.SC_MEAN
# store results
store_submit(test_preds, model_str, print_head=True)
store_oof(val_preds, model_str, print_head=True)