-
-
Notifications
You must be signed in to change notification settings - Fork 32
/
run_finetune.py
248 lines (209 loc) · 8.9 KB
/
run_finetune.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
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
# coding=utf-8
import argparse
import json
import os
import numpy as np
import tensorflow.compat.v1 as tf
import time
import tqdm
from copy import copy
from encode_bpe import BPEEncoder_ja
import model
if int(tf.__version__[0]) > 1:
from model import HParams as HParams
else:
from tensorflow.contrib.training import HParams
CHECKPOINT_DIR = 'checkpoint'
SAMPLE_DIR = 'samples'
parser = argparse.ArgumentParser(
description='Pretraining GPT2-JA on your custom dataset.',
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--dataset', metavar='PATH', type=str, required=True, help='Input npz file')
parser.add_argument('--base_model', type=str, default='gpt2ja-small', help='a path to a model file')
parser.add_argument('--batch_size', metavar='SIZE', type=int, default=1, help='Batch size')
parser.add_argument('--optim', type=str, default='adam', help='"adam", "adagrad", or "sgd" to use optimizer')
parser.add_argument('--learning_rate', metavar='LR', type=float, default=5e-5, help='Learning rate for optimizer')
parser.add_argument('--warmup_steps', metavar='WR', type=int, default=0, help='Learning rate warming up steps')
parser.add_argument('--run_name', type=str, default='gpt2ja_finetune', help='Run id. Name of subdirectory in checkpoint/')
parser.add_argument('--save_every', metavar='N', type=int, default=1000, help='Write a checkpoint every N steps')
parser.add_argument('--gpu', default='0', help='visible gpu number.')
def maketree(path):
try:
os.makedirs(path)
except:
pass
with open('ja-bpe.txt', encoding='utf-8') as f:
bpe = f.read().split('\n')
with open('emoji.json', encoding='utf-8') as f:
emoji = json.loads(f.read())
enc = BPEEncoder_ja(bpe, emoji)
n_vocab = len(enc)
def main():
args = parser.parse_args()
if os.path.isfile(args.base_model+'/hparams.json'):
with open(args.base_model+'/hparams.json', encoding='utf-8') as f:
params = json.loads(f.read())
hparams = HParams(**params)
elif 'small' in args.base_model:
hparams = HParams(**{
"n_vocab": n_vocab,
"n_ctx": 1024,
"n_embd": 768,
"n_head": 12,
"n_layer": 12
})
elif 'medium' in args.base_model:
hparams = HParams(**{
"n_vocab": n_vocab,
"n_ctx": 1024,
"n_embd": 1024,
"n_head": 16,
"n_layer": 24
})
elif 'large' in args.base_model:
hparams = HParams(**{
"n_vocab": n_vocab,
"n_ctx": 1024,
"n_embd": 1280,
"n_head": 20,
"n_layer": 36
})
else:
raise ValueError('invalid model name.')
config = tf.ConfigProto()
if int(args.gpu) >= 0:
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = args.gpu
with tf.Session(config=config,graph=tf.Graph()) as sess:
context = tf.placeholder(tf.int32, [None, None])
output = model.model(hparams=hparams, X=context, past=None, reuse=tf.AUTO_REUSE)
loss = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(
labels=context[:, 1:], logits=output['logits'][:, :-1]))
saver = tf.train.Saver()
ckpt = tf.train.latest_checkpoint(args.base_model)
saver.restore(sess, ckpt)
train_vars = tf.trainable_variables()
global_step = tf.Variable(0, trainable=False)
if args.warmup_steps > 0:
learning_rate = tf.compat.v1.train.polynomial_decay(
learning_rate=1e-10,
end_learning_rate=args.learning_rate,
global_step=global_step,
decay_steps=args.warmup_steps
)
else:
learning_rate = args.learning_rate
if args.optim=='adam':
opt = tf.train.AdamOptimizer(learning_rate=learning_rate,
beta1=0.9,
beta2=0.98,
epsilon=1e-7)
elif args.optim=='adagrad':
opt = tf.train.AdagradOptimizer(learning_rate=learning_rate)
elif args.optim=='sgd':
opt = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
else:
raise ValueError('invalid optimizer name.')
train_vars = tf.trainable_variables()
opt_grads = tf.gradients(loss, train_vars)
opt_grads = list(zip(opt_grads, train_vars))
opt_apply = opt.apply_gradients(opt_grads)
summaries = tf.summary.scalar('loss', loss)
summary_log = tf.summary.FileWriter(
os.path.join(CHECKPOINT_DIR, args.run_name))
saver = tf.train.Saver(
var_list=train_vars,
max_to_keep=5,
keep_checkpoint_every_n_hours=2)
sess.run(tf.global_variables_initializer())
ckpt = tf.train.latest_checkpoint(args.base_model)
saver.restore(sess, ckpt)
print('Loading checkpoint', ckpt)
print('Loading dataset...')
global_chunks = []
with np.load(args.dataset) as npz:
current_token = []
for inditem, item in enumerate(npz.files):
token_chunk = npz[item]
for ind in range(0,len(token_chunk)):
current_token.append(np.uint16(token_chunk[ind]))
if len(current_token) == hparams.n_ctx:
global_chunks.append(current_token)
current_token = []
if len(current_token) > 1:
global_chunks.append(current_token + [n_vocab-1]*(hparams.n_ctx-len(current_token)))
current_token = []
global_chunk_index = np.random.permutation(len(global_chunks))
global_chunk_step = 0
print('Training...')
def sample_feature():
nonlocal global_chunks,global_chunk_index,global_chunk_step
p_input_ids = []
for b in range(args.batch_size): # FULL-SENTENCES
idx = global_chunk_index[global_chunk_step]
global_chunk_step += 1
if global_chunk_step >= len(global_chunk_index):
global_chunk_step = 0
global_chunk_index = np.random.permutation(len(global_chunks))
sampled_token = global_chunks[idx]
# Make Sequence
ids = copy(global_chunks[idx])
p_input_ids.append(ids)
return {context:p_input_ids}
counter = 1
counter_path = os.path.join(CHECKPOINT_DIR, args.run_name, 'counter')
hparams_path = os.path.join(CHECKPOINT_DIR, args.run_name, 'hparams.json')
if os.path.exists(counter_path):
# Load the step number if we're resuming a run
# Add 1 so we don't immediately try to save again
with open(counter_path, 'r', encoding='utf-8') as fp:
counter = int(fp.read()) + 1
maketree(os.path.join(CHECKPOINT_DIR, args.run_name))
def save():
maketree(os.path.join(CHECKPOINT_DIR, args.run_name))
print(
'Saving',
os.path.join(CHECKPOINT_DIR, args.run_name,
'model-{}').format(counter))
saver.save(
sess,
os.path.join(CHECKPOINT_DIR, args.run_name, 'model'),
global_step=counter)
with open(counter_path, 'w', encoding='utf-8') as fp:
fp.write(str(counter) + '\n')
with open(hparams_path, 'w', encoding='utf-8') as fp:
fp.write(json.dumps({
"n_vocab": int(hparams.n_vocab),
"n_ctx": int(hparams.n_ctx),
"n_embd": int(hparams.n_embd),
"n_head": int(hparams.n_head),
"n_layer": int(hparams.n_layer),
}))
avg_loss = (0.0, 0.0)
start_time = time.time()
try:
while True:
if counter % args.save_every == 0:
save()
(_, v_loss, v_summary) = sess.run(
(opt_apply, loss, summaries),
feed_dict=sample_feature())
summary_log.add_summary(v_summary, counter)
avg_loss = (avg_loss[0] * 0.99 + v_loss,
avg_loss[1] * 0.99 + 1.0)
print(
'[{counter} | {time:2.2f}] loss={loss:2.2f} avg={avg:2.2f}'
.format(
counter=counter,
time=time.time() - start_time,
loss=v_loss,
avg=avg_loss[0] / avg_loss[1]))
counter = counter+1
if args.warmup_steps > 0:
global_step = global_step+1
except KeyboardInterrupt:
print('interrupted')
save()
if __name__ == '__main__':
main()