-
Notifications
You must be signed in to change notification settings - Fork 38
/
Copy pathrun_model.py
180 lines (163 loc) · 7.85 KB
/
run_model.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
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT license.
import sys
from argparse import ArgumentParser
from fairseq_cli.train import cli_main as fairseq_train
from fairseq_cli.generate import cli_main as fairseq_generate
import logging
import shlex
import re
from tapex.model_interface import TAPEXModelInterface
from tapex.model_eval import evaluate_generate_file
import os
logger = logging.getLogger(__name__)
def set_train_parser(parser_group):
train_parser = parser_group.add_parser("train")
train_parser.add_argument("--dataset-dir", type=str, required=True, default="",
help="dataset directory where train.src is located in")
train_parser.add_argument("--exp-dir", type=str, default="checkpoints",
help="experiment directory which stores the checkpoint weights")
train_parser.add_argument("--model-path", type=str, default="bart.base/model.pt",
help="the directory of pre-trained model path")
train_parser.add_argument("--model-arch", type=str, default="bart_base", choices=["bart_large", "bart_base"],
help="tapex large should correspond to bart_large, and tapex base should be bart_base")
train_parser.add_argument("--max-tokens", type=int, default=1536,
help="if you train a large model on 16GB memory, max-tokens should be empirically "
"set as 1536, and can be near-linearly increased according to your GPU memory.")
train_parser.add_argument("--gradient-accumulation", type=int, default=8 * 8,
help="the accumulation steps to arrive a equal batch size on one card, and"
" you can also reduce it to a proper value for you.")
train_parser.add_argument("--total-num-update", type=int, default=50000,
help="the total optimization training steps")
train_parser.add_argument("--learning-rate", type=float, default=3e-5,
help="the peak learning rate for model training")
def set_eval_parser(parser_group):
eval_parser = parser_group.add_parser("eval")
eval_parser.add_argument("--dataset-dir", type=str, required=True, default="",
help="dataset directory where train.src is located in")
eval_parser.add_argument("--model-path", type=str, default="tapex.base/model.pt",
help="the directory of fine-tuned model path such as tapex.base.wikisql/model.pt")
eval_parser.add_argument("--sub-dir", type=str, default="valid", choices=["train", "valid"],
help="the directory of pre-trained model path, and the default should be in"
"{bart.base, bart.large, tapex.base, tapex.large}.")
eval_parser.add_argument("--max-tokens", type=int, default=1536 * 4,
help="the max tokens can be larger than training when in inference.")
eval_parser.add_argument("--predict-dir", type=str, default="predict",
help="the predict folder of generated result.")
def set_predict_parser(parser_group):
predict_parser = parser_group.add_parser("predict")
predict_parser.add_argument("--resource-dir", type=str, required=True, default="./tapex.base",
help="the resource dir which contains the model weights, vocab.bpe, "
"dict.src.txt, dict.tgt.txt and encoder.json.")
predict_parser.add_argument("--checkpoint-name", type=str, default="model.pt",
help="the model weight's name in the resource directory")
def train_fairseq_model(args):
cmd = f"""
fairseq-train {args.dataset_dir}/bin \
--save-dir {args.exp_dir} \
--restore-file {args.model_path} \
--arch {args.model_arch} \
--memory-efficient-fp16 \
--task translation \
--criterion label_smoothed_cross_entropy \
--source-lang src \
--target-lang tgt \
--truncate-source \
--label-smoothing 0.1 \
--max-source-positions 1024 \
--max-tokens {args.max_tokens} \
--update-freq {args.gradient_accumulation} \
--max-update {args.total_num_update} \
--required-batch-size-multiple 1 \
--dropout 0.1 \
--attention-dropout 0.1 \
--relu-dropout 0.0 \
--weight-decay 0.01 \
--optimizer adam \
--adam-eps 1e-08 \
--clip-norm 0.1 \
--lr-scheduler polynomial_decay \
--lr {args.learning_rate} \
--total-num-update {args.total_num_update} \
--warmup-updates 5000 \
--ddp-backend no_c10d \
--num-workers 20 \
--reset-meters \
--reset-optimizer \
--reset-dataloader \
--share-all-embeddings \
--layernorm-embedding \
--share-decoder-input-output-embed \
--skip-invalid-size-inputs-valid-test \
--log-format json \
--log-interval 10 \
--save-interval-updates 100 \
--validate-interval 50 \
--save-interval 50 \
--patience 200 \
--report-accuracy
"""
sys.argv = shlex.split(cmd)
logger.info("Begin to train model for dataset {}".format(args.dataset_dir))
logger.info("Running command {}".format(re.sub("\s+", " ", cmd.replace("\n", " "))))
fairseq_train()
def evaluate_fairseq_model(args):
cmd = f"""
fairseq-generate
--path {args.model_path} \
{args.dataset_dir}/bin \
--truncate-source \
--gen-subset {args.sub_dir} \
--max-tokens {args.max_tokens} \
--nbest 1 \
--source-lang src \
--target-lang tgt \
--results-path {args.predict_dir} \
--beam 5 \
--bpe gpt2 \
--remove-bpe \
--num-workers 20 \
--skip-invalid-size-inputs-valid-test
"""
sys.argv = shlex.split(cmd)
logger.info("Begin to evaluate model on the {} subset of dataset {}".format(args.sub_dir, args.dataset_dir))
logger.info("Running command {}".format(re.sub("\s+", " ", cmd.replace("\n", " "))))
fairseq_generate()
# after generation, we should call TAPEX evaluate function to evaluate the result
generate_file = os.path.join(args.predict_dir, "generate-{}.txt".format(args.sub_dir))
# the delimiter is the answer delimiter used in training, which by default is a comma
evaluate_generate_file(generate_file, target_delimiter=", ")
def predict_demo(args):
demo_interface = TAPEXModelInterface(resource_dir=args.resource_dir,
checkpoint_name=args.checkpoint_name)
table_context = {
"header": ["Year", "City", "Country", "Nations"],
"rows": [
[1896, "Athens", "Greece", 14],
[1900, "Paris", "France", 24],
[1904, "St. Louis", "USA", 12],
[2004, "Athens", "Greece", 201],
[2012, "Beijing", "China", 204],
[2008, "London", "UK", 204]
]
}
while True:
# something like `SELECT Year WHERE Country = Greece`
question = input()
answer = demo_interface.predict(question=question,
table_context=table_context)
logger.info("Receive SQL query as : {}".format(question))
logger.info("The neural SQL execution result is : {}".format(answer))
if __name__ == '__main__':
parser = ArgumentParser()
subparsers = parser.add_subparsers(dest="subcommand")
set_train_parser(subparsers)
set_eval_parser(subparsers)
set_predict_parser(subparsers)
args = parser.parse_args()
if args.subcommand == "train":
train_fairseq_model(args)
elif args.subcommand == "eval":
evaluate_fairseq_model(args)
elif args.subcommand == "predict":
predict_demo(args)