Datasets:

Languages:
English
ArXiv:
License:
File size: 12,207 Bytes
19391a5
a3e1ce4
 
 
60b610a
c28734d
 
60b610a
c28734d
 
60b610a
c28734d
 
ae07650
c28734d
 
ae07650
c28734d
 
19391a5
 
5b95481
 
 
 
9a436ff
5b95481
 
 
 
 
 
 
 
 
 
 
 
bde7d96
 
 
5b95481
 
 
 
1b10b57
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5b95481
 
 
 
 
 
 
 
 
 
9a436ff
 
 
 
5b95481
 
 
 
 
 
 
 
 
9a436ff
5b95481
 
 
 
9a436ff
5b95481
 
 
 
 
 
 
 
 
 
 
 
19391a5
 
 
5b95481
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a3e1ce4
5b95481
 
 
 
 
1b10b57
 
60b610a
5b95481
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19391a5
 
5b95481
 
e9bf213
 
39fc020
e9bf213
 
 
5b95481
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
911f874
5b95481
 
 
 
5f0221d
8ea930a
1b10b57
 
 
 
5b95481
 
 
 
 
 
 
19391a5
 
8b720cd
5b95481
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bde7d96
 
51db099
 
bde7d96
 
 
1693d6c
fe2e381
5b95481
 
 
 
 
 
 
 
 
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
249
250
251
'''
wandb offline
export WANDB_DISABLED='true'
export RAY_RESULTS='ray_results'

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"
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"

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"
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"

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"
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"

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"
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"

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"
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"
'''

import argparse
import json
import logging
import os
import math
import shutil
import urllib.request
import multiprocessing
from os.path import join as pj

import torch
import numpy as np
from huggingface_hub import create_repo
from datasets import load_dataset, load_metric
from transformers import AutoTokenizer, AutoModelForSequenceClassification, TrainingArguments, Trainer
from ray import tune

from readme import get_readme


logging.basicConfig(format='%(asctime)s %(levelname)-8s %(message)s', level=logging.INFO, datefmt='%Y-%m-%d %H:%M:%S')

PARALLEL = bool(int(os.getenv("PARALLEL", 1)))
RAY_RESULTS = os.getenv("RAY_RESULTS", "ray_results")
LABEL2ID = {
    "arts_&_culture": 0,
    "business_&_entrepreneurs": 1,
    "celebrity_&_pop_culture": 2,
    "diaries_&_daily_life": 3,
    "family": 4,
    "fashion_&_style": 5,
    "film_tv_&_video": 6,
    "fitness_&_health": 7,
    "food_&_dining": 8,
    "gaming": 9,
    "learning_&_educational": 10,
    "music": 11,
    "news_&_social_concern": 12,
    "other_hobbies": 13,
    "relationships": 14,
    "science_&_technology": 15,
    "sports": 16,
    "travel_&_adventure": 17,
    "youth_&_student_life": 18
}
ID2LABEL = {v: k for k, v in LABEL2ID.items()}


def internet_connection(host='http://google.com'):
    try:
        urllib.request.urlopen(host)
        return True
    except:
        return False


def sigmoid(x):
  return 1 / (1 + math.exp(-x))


def get_metrics():
    metric_accuracy = load_metric("accuracy", "multilabel")
    metric_f1 = load_metric("f1", "multilabel")

    # metric_f1.compute(predictions=[[0, 1, 1], [1, 1, 0]], references=[[0, 1, 1], [0, 1, 0]], average='micro')
    # metric_accuracy.compute(predictions=[[0, 1, 1], [1, 1, 0]], references=[[0, 1, 1], [0, 1, 0]])

    def compute_metric_search(eval_pred):
        logits, labels = eval_pred
        predictions = np.array([[int(sigmoid(j) > 0.5) for j in i] for i in logits])
        return metric_f1.compute(predictions=predictions, references=labels, average='micro')

    def compute_metric_all(eval_pred):
        logits, labels = eval_pred
        predictions = np.array([[int(sigmoid(j) > 0.5) for j in i] for i in logits])
        return {
            'f1': metric_f1.compute(predictions=predictions, references=labels, average='micro')['f1'],
            'f1_macro': metric_f1.compute(predictions=predictions, references=labels, average='macro')['f1'],
            'accuracy': metric_accuracy.compute(predictions=predictions, references=labels)['accuracy']
        }
    return compute_metric_search, compute_metric_all


def main():
    parser = argparse.ArgumentParser(description='Fine-tuning language model.')
    parser.add_argument('-m', '--model', help='transformer LM', default='roberta-base', type=str)
    parser.add_argument('-d', '--dataset', help='', default='cardiffnlp/tweet_topic_multi', type=str)
    parser.add_argument('--split-train', help='', required=True, type=str)
    parser.add_argument('--split-validation', help='', required=True, type=str)
    parser.add_argument('--split-test', help='', required=True, type=str)
    parser.add_argument('-l', '--seq-length', help='', default=128, type=int)
    parser.add_argument('--random-seed', help='', default=42, type=int)
    parser.add_argument('--eval-step', help='', default=50, type=int)
    parser.add_argument('-o', '--output-dir', help='Directory to output', default='ckpt_tmp', type=str)
    parser.add_argument('-t', '--n-trials', default=10, type=int)
    parser.add_argument('--push-to-hub', action='store_true')
    parser.add_argument('--use-auth-token', action='store_true')
    parser.add_argument('--hf-organization', default=None, type=str)
    parser.add_argument('-a', '--model-alias', help='', default=None, type=str)
    parser.add_argument('--summary-file', default='metric_summary.json', type=str)
    parser.add_argument('--skip-train', action='store_true')
    parser.add_argument('--skip-eval', action='store_true')
    opt = parser.parse_args()
    assert opt.summary_file.endswith('.json'), f'`--summary-file` should be a json file {opt.summary_file}'
    # setup data
    dataset = load_dataset(opt.dataset)
    network = internet_connection()
    # setup model
    tokenizer = AutoTokenizer.from_pretrained(opt.model, local_files_only=not network)
    model = AutoModelForSequenceClassification.from_pretrained(
        opt.model,
        id2label=ID2LABEL,
        label2id=LABEL2ID,
        num_labels=len(dataset[opt.split_train]['label'][0]),
        local_files_only=not network,
        problem_type="multi_label_classification"
    )
    tokenized_datasets = dataset.map(
        lambda x: tokenizer(x["text"], padding="max_length", truncation=True, max_length=opt.seq_length),
        batched=True)
    # setup metrics
    compute_metric_search, compute_metric_all = get_metrics()

    if not opt.skip_train:
        # setup trainer
        trainer = Trainer(
            model=model,
            args=TrainingArguments(
                output_dir=opt.output_dir,
                evaluation_strategy="steps",
                eval_steps=opt.eval_step,
                seed=opt.random_seed
            ),
            train_dataset=tokenized_datasets[opt.split_train],
            eval_dataset=tokenized_datasets[opt.split_validation],
            compute_metrics=compute_metric_search,
            model_init=lambda x: AutoModelForSequenceClassification.from_pretrained(
                opt.model,
                return_dict=True,
                num_labels=len(dataset[opt.split_train]['label'][0]),
                id2label=ID2LABEL,
                label2id=LABEL2ID
            )
        )
        # parameter search
        if PARALLEL:
            best_run = trainer.hyperparameter_search(
                hp_space=lambda x: {
                    "learning_rate": tune.loguniform(1e-6, 1e-4),
                    "num_train_epochs": tune.choice(list(range(1, 6))),
                    "per_device_train_batch_size": tune.choice([4, 8, 16, 32, 64]),
                },
                local_dir=RAY_RESULTS, direction="maximize", backend="ray", n_trials=opt.n_trials,
                resources_per_trial={'cpu': multiprocessing.cpu_count(), "gpu": torch.cuda.device_count()},

            )
        else:
            best_run = trainer.hyperparameter_search(
                hp_space=lambda x: {
                    "learning_rate": tune.loguniform(1e-6, 1e-4),
                    "num_train_epochs": tune.choice(list(range(1, 6))),
                    "per_device_train_batch_size": tune.choice([4, 8, 16, 32, 64]),
                },
                local_dir=RAY_RESULTS, direction="maximize", backend="ray", n_trials=opt.n_trials
            )
        # finetuning
        for n, v in best_run.hyperparameters.items():
            setattr(trainer.args, n, v)
        trainer.train()
        trainer.save_model(pj(opt.output_dir, 'best_model'))
        best_model_path = pj(opt.output_dir, 'best_model')
    else:
        best_model_path = pj(opt.output_dir, 'best_model')

    # evaluation
    model = AutoModelForSequenceClassification.from_pretrained(
        best_model_path,
        num_labels=len(dataset[opt.split_train]['label'][0]),
        local_files_only=not network,
        problem_type="multi_label_classification",
        id2label=ID2LABEL,
        label2id=LABEL2ID
    )
    trainer = Trainer(
        model=model,
        args=TrainingArguments(
            output_dir=opt.output_dir,
            evaluation_strategy="no",
            seed=opt.random_seed
        ),
        train_dataset=tokenized_datasets[opt.split_train],
        eval_dataset=tokenized_datasets[opt.split_test],
        compute_metrics=compute_metric_all
    )
    summary_file = pj(opt.output_dir, opt.summary_file)
    if not opt.skip_eval:
        result = {f'test/{k}': v for k, v in trainer.evaluate().items()}
        logging.info(json.dumps(result, indent=4))
        with open(summary_file, 'w') as f:
            json.dump(result, f)

    if opt.push_to_hub:
        assert opt.hf_organization is not None, f'specify hf organization `--hf-organization`'
        assert opt.model_alias is not None, f'specify hf organization `--model-alias`'
        url = create_repo(opt.model_alias, organization=opt.hf_organization, exist_ok=True)
        # if not opt.skip_train:
        args = {"use_auth_token": opt.use_auth_token, "repo_url": url, "organization": opt.hf_organization}
        trainer.model.push_to_hub(opt.model_alias, **args)
        tokenizer.push_to_hub(opt.model_alias, **args)
        if os.path.exists(summary_file):
            shutil.copy2(summary_file, opt.model_alias)
        extra_desc = f"This model is fine-tuned on `{opt.split_train}` split and validated on `{opt.split_test}` split of tweet_topic."
        readme = get_readme(
            model_name=f"{opt.hf_organization}/{opt.model_alias}",
            metric=summary_file,
            language_model=opt.model,
            extra_desc= extra_desc
            )
        with open(f"{opt.model_alias}/README.md", "w") as f:
            f.write(readme)
        os.system(
            f"cd {opt.model_alias} && git lfs install && git add . && git commit -m 'model update' && git push && cd ../")
        shutil.rmtree(f"{opt.model_alias}")  # clean up the cloned repo


if __name__ == '__main__':
    main()