Datasets:
Tasks:
Text Classification
Modalities:
Text
Sub-tasks:
sentiment-classification
Languages:
English
Size:
10K - 100K
ArXiv:
License:
Update lm_finetuning.py
Browse files- lm_finetuning.py +34 -12
lm_finetuning.py
CHANGED
@@ -1,3 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import argparse
|
2 |
import json
|
3 |
import logging
|
@@ -52,7 +73,9 @@ def main():
|
|
52 |
parser = argparse.ArgumentParser(description='Fine-tuning language model.')
|
53 |
parser.add_argument('-m', '--model', help='transformer LM', default='roberta-base', type=str)
|
54 |
parser.add_argument('-d', '--dataset', help='', default='cardiffnlp/tweet_topic_single', type=str)
|
55 |
-
parser.add_argument('--
|
|
|
|
|
56 |
parser.add_argument('-l', '--seq-length', help='', default=128, type=int)
|
57 |
parser.add_argument('--random-seed', help='', default=42, type=int)
|
58 |
parser.add_argument('--eval-step', help='', default=50, type=int)
|
@@ -68,7 +91,7 @@ def main():
|
|
68 |
opt = parser.parse_args()
|
69 |
assert opt.summary_file.endswith('.json'), f'`--summary-file` should be a json file {opt.summary_file}'
|
70 |
# setup data
|
71 |
-
dataset = load_dataset(opt.dataset
|
72 |
network = internet_connection()
|
73 |
# setup model
|
74 |
tokenizer = AutoTokenizer.from_pretrained(opt.model, local_files_only=not network)
|
@@ -90,11 +113,11 @@ def main():
|
|
90 |
eval_steps=opt.eval_step,
|
91 |
seed=opt.random_seed
|
92 |
),
|
93 |
-
train_dataset=tokenized_datasets[
|
94 |
-
eval_dataset=tokenized_datasets[
|
95 |
compute_metrics=compute_metric_search,
|
96 |
-
model_init=lambda x: AutoModelForSequenceClassification.from_pretrained(
|
97 |
-
|
98 |
)
|
99 |
# parameter search
|
100 |
if PARALLEL:
|
@@ -128,9 +151,7 @@ def main():
|
|
128 |
|
129 |
# evaluation
|
130 |
model = AutoModelForSequenceClassification.from_pretrained(
|
131 |
-
|
132 |
-
num_labels=dataset['train'].features['label'].num_classes,
|
133 |
-
local_files_only=not network)
|
134 |
trainer = Trainer(
|
135 |
model=model,
|
136 |
args=TrainingArguments(
|
@@ -138,11 +159,11 @@ def main():
|
|
138 |
evaluation_strategy="no",
|
139 |
seed=opt.random_seed
|
140 |
),
|
141 |
-
train_dataset=tokenized_datasets[
|
142 |
-
eval_dataset=tokenized_datasets[
|
143 |
compute_metrics=compute_metric_all,
|
144 |
model_init=lambda x: AutoModelForSequenceClassification.from_pretrained(
|
145 |
-
|
146 |
)
|
147 |
summary_file = pj(opt.output_dir, opt.summary_file)
|
148 |
if not opt.skip_eval:
|
@@ -168,3 +189,4 @@ def main():
|
|
168 |
|
169 |
if __name__ == '__main__':
|
170 |
main()
|
|
|
|
1 |
+
'''
|
2 |
+
wandb offline
|
3 |
+
export WANDB_DISABLED='true'
|
4 |
+
export RAY_RESULTS='ray_results'
|
5 |
+
|
6 |
+
python lm_finetuning.py -m "roberta-large" -o "ckpt/2021/roberta-large" --push-to-hub --hf-organization "cardiffnlp" -a "roberta-large-tweet-topic-multi-all" --split-train "train_all" --split-valid "validation_2021" --split-test "test_2021"
|
7 |
+
python lm_finetuning.py -m "roberta-large" -o "ckpt/2020/roberta-large" --push-to-hub --hf-organization "cardiffnlp" -a "roberta-large-tweet-topic-multi-2020" --split-train "train_2020" --split-valid "validation_2020" --split-test "test_2021"
|
8 |
+
|
9 |
+
python lm_finetuning.py -m "roberta-base" -o "ckpt/2021/roberta_base" --push-to-hub --hf-organization "cardiffnlp" -a "roberta-base-tweet-topic-multi-all" --split-train "train_all" --split-valid "validation_2021" --split-test "test_2021"
|
10 |
+
python lm_finetuning.py -m "roberta-base" -o "ckpt/2020/roberta_base" --push-to-hub --hf-organization "cardiffnlp" -a "roberta-base-tweet-topic-multi-2020" --split-train "train_2020" --split-valid "validation_2020" --split-test "test_2021"
|
11 |
+
|
12 |
+
python lm_finetuning.py -m "cardiffnlp/twitter-roberta-base-2019-90m" -o "ckpt/2021/twitter-roberta-base-2019-90m" --push-to-hub --hf-organization "cardiffnlp" -a "twitter-roberta-base-2019-90m-tweet-topic-multi-all" --split-train "train_all" --split-valid "validation_2021" --split-test "test_2021"
|
13 |
+
python lm_finetuning.py -m "cardiffnlp/twitter-roberta-base-2019-90m" -o "ckpt/2020/twitter-roberta-base-2019-90m" --push-to-hub --hf-organization "cardiffnlp" -a "twitter-roberta-base-2019-90m-tweet-topic-multi-2020" --split-train "train_2020" --split-valid "validation_2020" --split-test "test_2021"
|
14 |
+
|
15 |
+
python lm_finetuning.py -m "cardiffnlp/twitter-roberta-base-dec2020" -o "ckpt/2021/twitter-roberta-base-dec2020" --push-to-hub --hf-organization "cardiffnlp" -a "twitter-roberta-base-dec2020-tweet-topic-multi-all" --split-train "train_all" --split-valid "validation_2021" --split-test "test_2021"
|
16 |
+
python lm_finetuning.py -m "cardiffnlp/twitter-roberta-base-dec2020" -o "ckpt/2020/twitter-roberta-base-dec2020" --push-to-hub --hf-organization "cardiffnlp" -a "twitter-roberta-base-dec2020-tweet-topic-multi-2020" --split-train "train_2020" --split-valid "validation_2020" --split-test "test_2021"
|
17 |
+
|
18 |
+
python lm_finetuning.py -m "cardiffnlp/twitter-roberta-base-dec2021" -o "ckpt/2021/twitter-roberta-base-dec2021" --push-to-hub --hf-organization "cardiffnlp" -a "twitter-roberta-base-dec2021-tweet-topic-multi-all" --split-train "train_all" --split-valid "validation_2021" --split-test "test_2021"
|
19 |
+
python lm_finetuning.py -m "cardiffnlp/twitter-roberta-base-dec2021" -o "ckpt/2020/twitter-roberta-base-dec2021" --push-to-hub --hf-organization "cardiffnlp" -a "twitter-roberta-base-dec2021-tweet-topic-multi-2020" --split-train "train_2020" --split-valid "validation_2020" --split-test "test_2021"
|
20 |
+
'''
|
21 |
+
|
22 |
import argparse
|
23 |
import json
|
24 |
import logging
|
|
|
73 |
parser = argparse.ArgumentParser(description='Fine-tuning language model.')
|
74 |
parser.add_argument('-m', '--model', help='transformer LM', default='roberta-base', type=str)
|
75 |
parser.add_argument('-d', '--dataset', help='', default='cardiffnlp/tweet_topic_single', type=str)
|
76 |
+
parser.add_argument('--split-train', help='', required=True, type=str)
|
77 |
+
parser.add_argument('--split-validation', help='', required=True, type=str)
|
78 |
+
parser.add_argument('--split-test', help='', required=True, type=str)
|
79 |
parser.add_argument('-l', '--seq-length', help='', default=128, type=int)
|
80 |
parser.add_argument('--random-seed', help='', default=42, type=int)
|
81 |
parser.add_argument('--eval-step', help='', default=50, type=int)
|
|
|
91 |
opt = parser.parse_args()
|
92 |
assert opt.summary_file.endswith('.json'), f'`--summary-file` should be a json file {opt.summary_file}'
|
93 |
# setup data
|
94 |
+
dataset = load_dataset(opt.dataset)
|
95 |
network = internet_connection()
|
96 |
# setup model
|
97 |
tokenizer = AutoTokenizer.from_pretrained(opt.model, local_files_only=not network)
|
|
|
113 |
eval_steps=opt.eval_step,
|
114 |
seed=opt.random_seed
|
115 |
),
|
116 |
+
train_dataset=tokenized_datasets[opt.split_train],
|
117 |
+
eval_dataset=tokenized_datasets[opt.split_validation],
|
118 |
compute_metrics=compute_metric_search,
|
119 |
+
model_init=lambda x: AutoModelForSequenceClassification.from_pretrained(opt.model, num_labels=dataset['train'].features['label'].num_classes, local_files_only=not network, return_dict=True)
|
120 |
+
)
|
121 |
)
|
122 |
# parameter search
|
123 |
if PARALLEL:
|
|
|
151 |
|
152 |
# evaluation
|
153 |
model = AutoModelForSequenceClassification.from_pretrained(
|
154 |
+
opt.model, num_labels=dataset['train'].features['label'].num_classes, local_files_only=not network)
|
|
|
|
|
155 |
trainer = Trainer(
|
156 |
model=model,
|
157 |
args=TrainingArguments(
|
|
|
159 |
evaluation_strategy="no",
|
160 |
seed=opt.random_seed
|
161 |
),
|
162 |
+
train_dataset=tokenized_datasets[opt.split_train],
|
163 |
+
eval_dataset=tokenized_datasets[opt.split_test],
|
164 |
compute_metrics=compute_metric_all,
|
165 |
model_init=lambda x: AutoModelForSequenceClassification.from_pretrained(
|
166 |
+
opt.model, num_labels=dataset['train'].features['label'].num_classes, local_files_only=not network, return_dict=True)
|
167 |
)
|
168 |
summary_file = pj(opt.output_dir, opt.summary_file)
|
169 |
if not opt.skip_eval:
|
|
|
189 |
|
190 |
if __name__ == '__main__':
|
191 |
main()
|
192 |
+
|