-
Notifications
You must be signed in to change notification settings - Fork 3
/
train.py
168 lines (138 loc) · 7.71 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
162
163
164
165
166
167
168
import os
import torch
import torch.nn as nn
from torch import LongTensor
import rtdl
import typing as ty
import yaml
import numpy as np
from dataset import *
from utils import *
from metrics import *
from model import common
from sklearn.model_selection import KFold, StratifiedKFold
import wandb
from collections import OrderedDict
def model_train(args : ty.Any, config: ty.Dict[str, ty.List[str]]) -> None:
"""
args have device info (CPU, GPU, etc..), data, path [check main.py File].
(Default Model using torcn.nn.parallel.)
config have model parameters and model info. check model.yaml.
- If you question or Error, leave an Issue.
"""
train_dict, val_dict, test_dict, dataset_info_dict, d_out, y_std = load_dataset(args.data_path)
print("loaded Dataset..")
model = common.load_model(config, dataset_info_dict)
print("loaded Model..")
wandb.init( name = config["model"] + "_" + str(config["count"]),
project = config["model"] + '_' + args.data)
wandb.config = config
optimizer = get_optimizer(model, config)
loss_fn = get_loss(dataset_info_dict)
print("loaded optimizer and loss..")
if int(config["fold"]) > 2: # fold102235
print("Fold Training..")
kf = KFold(get_splits = 15)
for idxx, (train_idx, temp_idx) in enumerate(kf.split(train_dict["N_train"], train_dict["y_train"])): # X_train, X_temp, y_train, y_temp =
fold_train_dict = {}
fold_train_dict["N_train"] = train_dict["N_train"][train_idx]
fold_train_dict["y_train"] = train_dict["y_train"][train_idx]
train_dataloader, valid_dataloader, test_dataloader = get_DataLoader(fold_train_dict, val_dict, test_dict, config)
else: # Default
print("Single[default] Training..")
train_dataloader, valid_dataloader, test_dataloader = get_DataLoader(train_dict, val_dict, test_dict, config)
model_run(model, optimizer, loss_fn, train_dataloader, valid_dataloader, test_dataloader, dataset_info_dict, args, config, y_std)
#
def model_run(model, optimizer, loss_fn, train_dataloader, valid_dataloader, test_dataloader, dataset_info_dict, args, config, y_std):
seed_everything(0)
model.to(args.device)
model = nn.DataParallel(model)
method = "ensemble" if config["fold"] > 0 else "default"
json_info_output_path = os.path.join(str(args.savepath), config["model"], str(args.data), method)
print("Ready to run model...")
# Best Score
if dataset_info_dict["task_type"] == "regression": # RMSE
best_valid = 1e10
else: # Accuracy
best_valid = 0
for epoch in range(int(config["epochs"])):
train_loss_score, valid_loss_score = 0, 0
train_pred, valid_pred, test_pred = np.array([]), np.array([]), np.array([])
train_label, valid_label, test_label = np.array([]), np.array([]), np.array([])
model.train() # Train DataLoader
for X_data, y_label in train_dataloader: # Train
optimizer.zero_grad()
X_data, y_label = X_data.to(args.device), y_label.to(args.device)
y_pred = model(x_num = X_data, x_cat = None)
if dataset_info_dict["task_type"] == "regression":
loss = loss_fn(y_pred.to(torch.float64).squeeze(1), y_label.to(torch.float64))
elif dataset_info_dict["task_type"] == "binclass": # ERROR
loss = loss_fn(y_pred.squeeze(1), y_label.to(torch.float32))
else:
loss = loss_fn(y_pred, y_label)
loss.backward()
optimizer.step()
train_loss_score += loss.item()
if dataset_info_dict["task_type"] == "regression":
train_pred = np.append(train_pred, y_pred.cpu().detach().numpy())
train_label = np.append(train_label, y_label.cpu().detach().numpy())
elif dataset_info_dict["task_type"] == "binclass": # ERROR
train_pred = np.append(train_pred, y_pred.cpu().detach().numpy())
train_label = np.append(train_label, y_label.cpu().detach().numpy())
else:
_, indics = torch.max(y_pred, 1)
train_pred = np.append(train_pred, indics.cpu().detach().numpy())
train_label = np.append(train_label, y_label.cpu().detach().numpy())
model.eval() # Valid DataLoader
for X_data, y_label in valid_dataloader:
X_data, y_label = X_data.to(args.device), y_label.to(args.device)
y_pred = model(x_num = X_data, x_cat = None)
if dataset_info_dict["task_type"] == "regression":
valid_pred = np.append(valid_pred, y_pred.cpu().detach().numpy())
valid_label = np.append(valid_label, y_label.cpu().detach().numpy())
elif dataset_info_dict["task_type"] == "binclass": # ERROR
valid_pred = np.append(valid_pred, y_pred.cpu().detach().numpy())
valid_label = np.append(valid_label, y_label.cpu().detach().numpy())
else:
_, indics = torch.max(y_pred, 1)
valid_pred = np.append(valid_pred, indics.cpu().detach().numpy())
valid_label = np.append(valid_label, y_label.cpu().detach().numpy())
model.eval() # Test DataLoader
for X_data, y_label in test_dataloader:
X_data, y_label = X_data.to(args.device), y_label.to(args.device)
y_pred = model(x_num = X_data, x_cat = None)
if dataset_info_dict["task_type"] == "regression":
test_pred = np.append(test_pred, y_pred.cpu().detach().numpy())
test_label = np.append(test_label, y_label.cpu().detach().numpy())
elif dataset_info_dict["task_type"] == "binclass": # ERROR
test_pred = np.append(test_pred, y_pred.cpu().detach().numpy())
test_label = np.append(test_label, y_label.cpu().detach().numpy())
else:
_, indics = torch.max(y_pred, 1)
test_pred = np.append(test_pred, indics.cpu().detach().numpy())
test_label = np.append(test_label, y_label.cpu().detach().numpy())
if dataset_info_dict["task_type"] == "regression":
train_score, valid_score = get_rmse_score(train_pred, train_label, y_std), get_rmse_score(valid_pred, valid_label, y_std)
if best_valid > valid_score:
best_valid = valid_score
test_score = get_rmse_score(test_pred, test_label, y_std)
config["valid_rmse"] = valid_score
config["test_rmse"] = test_score
save_mode_with_json(model, config, json_info_output_path)
else:
if dataset_info_dict["task_type"] == "binclass": # ERROR
train_score, valid_score = get_accuracy_score(train_pred, train_label, dataset_info_dict), get_accuracy_score(valid_pred, valid_label, dataset_info_dict)
else:
train_score, valid_score = get_accuracy_score(train_pred, train_label, dataset_info_dict), get_accuracy_score(valid_pred, valid_label, dataset_info_dict)
if best_valid < valid_score:
best_valid = valid_score
test_score = get_accuracy_score(test_pred, test_label, dataset_info_dict)
config["valid_accuracy"] = valid_score
config["test_accuracy"] = test_score
save_mode_with_json(model, config, json_info_output_path)
wandb.log({
"train_score" : train_score,
"train_loss" : train_loss_score,
"valid_score" : valid_score,
"test_score" : test_score
})