-
Notifications
You must be signed in to change notification settings - Fork 0
/
finetune_on_ful.py
188 lines (133 loc) · 4.56 KB
/
finetune_on_ful.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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import torch
import os
import argparse
import logging
import datetime
import gc
date_time = str(datetime.date.today()) + "_" + ":".join(str(datetime.datetime.now()).split()[1].split(":")[:2])
parser = argparse.ArgumentParser()
parser.add_argument(
"--dataset",
type = str,
help = "select dataset / task",
default = "spanish_xnli",
# choices = ["french_xnli" "french_paws" french_csl # spanish_csl
#choices = ["ant", "csl","ChnSentiCorp", "sst", "evinf", "agnews", "multirc", "evinf_FA"]
)
parser.add_argument(
"--data_dir",
type = str,
help = "directory of saved processed data",
default = "datasets/"
)
parser.add_argument(
"--model_dir",
type = str,
help = "directory to save models, mannually modify it for multi and mono",
default = "test_trained_models/" # macbert bert zhbert french_bert
)
parser.add_argument(
"--seed",
type = int,
help = "random seed for experiment",
default = 15
)
parser.add_argument(
'--evaluate_models',
help='test predictive performance in and out of domain',
action='store_true',
default=False,
)
user_args = vars(parser.parse_args())
user_args["importance_metric"] = None
### used only for data stats
data_dir_plain = user_args["data_dir"]
log_dir = "experiment_logs/train_" + user_args["dataset"] + "_seed-" + str(user_args["seed"]) + "_" + date_time + "/"
config_dir = "experiment_config/train_" + user_args["dataset"] + "_seed-" + str(user_args["seed"]) + "_" + date_time + "/"
os.makedirs(log_dir, exist_ok = True)
os.makedirs(config_dir, exist_ok = True)
import config.cfg
config.cfg.config_directory = config_dir
logging.basicConfig(
filename= log_dir + "/out.log",
format='%(asctime)s %(levelname)-8s %(message)s',
level=logging.INFO,
datefmt='%Y-%m-%d %H:%M:%S'
)
# device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
if torch.cuda.is_available():
device = torch.device("cuda:0")
print("running on the GPU")
else:
device = torch.device("cpu")
print("running on the CPU")
logging.info("Running on cuda : {}".format(torch.cuda.is_available()))
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
from src.common_code.initialiser import initial_preparations
import datetime
# creating unique config from stage_config.json file and model_config.json file
args = initial_preparations(user_args, stage = "train")
logging.info("config : \n ----------------------")
[logging.info(k + " : " + str(v)) for k,v in args.items()]
logging.info("\n ----------------------")
if args['model_abbreviation'] == 't5m':
from src.data_functions.dataholder import mT5_HOLDER as dataholder
print(' ')
print(' ')
print('using T5')
else: from src.data_functions.dataholder import BERT_HOLDER as dataholder
from src.tRpipeline import train_and_save, train_and_save_t5, test_predictive_performance, keep_best_model_
from src.data_functions.useful_functions import describe_data_stats
data_desc = describe_data_stats(
path_to_data = args["data_dir"],
path_to_stats = os.path.join(
data_dir_plain,
args["dataset"],
""
)
)
import pprint
logging.info(
pprint.pformat(data_desc, indent = 4)
)
del data_desc
gc.collect()
# training the models and evaluating their predictive performance
# on the full text length
data = dataholder(
path = args["data_dir"],
b_size = args["batch_size"],
stage = "train"
)
print(' done loading data')
## evaluating finetuned models
if args["evaluate_models"]:
## in domain evaluation
test_stats = test_predictive_performance(
test_data_loader = data.test_loader,
for_rationale = False,
output_dims = data.nu_of_labels,
save_output_probs = True,
)
del data
gc.collect()
## shows which model performed best on dev F1 (in-domain)
## if keep_models = False then will remove the rest of the models to save space
keep_best_model_(keep_models = False)
else:
if args['model_abbreviation'] == 't5m':
train_and_save_t5(
train_data_loader = data.train_loader,
dev_data_loader = data.dev_loader,
for_rationale = False,
output_dims = data.nu_of_labels,
)
else: train_and_save(
train_data_loader = data.train_loader,
dev_data_loader = data.dev_loader,
for_rationale = False,
output_dims = data.nu_of_labels,
)