-
Notifications
You must be signed in to change notification settings - Fork 0
/
pl2text_codebert.py
593 lines (537 loc) · 25.4 KB
/
pl2text_codebert.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
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
import pickle as pkl
from numpy import pad
from tqdm import tqdm
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler
from torch.optim import Adam, AdamW, lr_scheduler
import torch.nn as nn
import torch.nn.functional as F
import torch
from torch.nn.utils.rnn import pad_sequence
from transformers import get_linear_schedule_with_warmup
import os
import json
import itertools
import wandb
from sklearn.metrics import precision_score, recall_score, f1_score
import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Optional
import datasets
import nltk # Here to have a nice missing dependency error message early on
import numpy as np
from datasets import load_dataset, concatenate_datasets
import evaluate
import statistics
import transformers
from filelock import FileLock
from transformers import (
AutoConfig,
AutoModelForSeq2SeqLM,
AutoTokenizer,
DataCollatorForSeq2Seq,
HfArgumentParser,
Seq2SeqTrainer,
Seq2SeqTrainingArguments,
set_seed,
EncoderDecoderModel,
BertTokenizer,
EncoderDecoderConfig,
GPT2Tokenizer,
GPT2Config,
GPT2LMHeadModel,
RobertaModel,
RobertaConfig
)
from transformers.trainer_utils import get_last_checkpoint
device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
# os.environ['WANDB_NOTEBOOK_NAME'] = 'pl2text'
# wandb.init(project="pl2text")
logger = logging.getLogger(__name__)
nltk.download('punkt')
# try:
# nltk.data.find("tokenizers/punkt")
# except (LookupError, OSError):
# if is_offline_mode():
# raise LookupError(
# "Offline mode: run this script without TRANSFORMERS_OFFLINE first to download nltk data files"
# )
# with FileLock(".lock") as lock:
# nltk.download("punkt", quiet=True)
@dataclass
class ModelArguments:
"""
Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
"""
model_name_or_path: str = field(
metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
)
config_name: Optional[str] = field(
default=None, metadata={"help": "Pretrained config name or path if not the same as model_name"}
)
tokenizer_name: Optional[str] = field(
default=None, metadata={"help": "Pretrained tokenizer name or path if not the same as model_name"}
)
cache_dir: Optional[str] = field(
default=None,
metadata={"help": "Where to store the pretrained models downloaded from huggingface.co"},
)
use_fast_tokenizer: bool = field(
default=True,
metadata={"help": "Whether to use one of the fast tokenizer (backed by the tokenizers library) or not."},
)
model_revision: str = field(
default="main",
metadata={"help": "The specific model version to use (can be a branch name, tag name or commit id)."},
)
use_auth_token: bool = field(
default=False,
metadata={
"help": (
"Will use the token generated when running `huggingface-cli login` (necessary to use this script "
"with private models)."
)
},
)
resize_position_embeddings: Optional[bool] = field(
default=None,
metadata={
"help": (
"Whether to automatically resize the position embeddings if `max_source_length` exceeds "
"the model's position embeddings."
)
},
)
@dataclass
class DataTrainingArguments:
"""
Arguments pertaining to what data we are going to input our model for training and eval.
"""
source_lang: str = field(
default=None, metadata={"help": "Source language id for translation."})
target_lang: str = field(
default=None, metadata={"help": "Target language id for translation."})
dataset_name: Optional[str] = field(
default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
)
dataset_config_name: Optional[str] = field(
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
)
text_column: Optional[str] = field(
default=None,
metadata={"help": "The name of the column in the datasets containing the full texts (for summarization)."},
)
summary_column: Optional[str] = field(
default=None,
metadata={"help": "The name of the column in the datasets containing the summaries (for summarization)."},
)
train_file: Optional[str] = field(
default=None, metadata={"help": "The input training data file (a jsonlines or csv file)."}
)
validation_file: Optional[str] = field(
default=None,
metadata={
"help": (
"An optional input evaluation data file to evaluate the metrics (rouge) on (a jsonlines or csv file)."
)
},
)
test_file: Optional[str] = field(
default=None,
metadata={
"help": "An optional input test data file to evaluate the metrics (rouge) on (a jsonlines or csv file)."
},
)
overwrite_cache: bool = field(
default=True, metadata={"help": "Overwrite the cached training and evaluation sets"}
)
preprocessing_num_workers: Optional[int] = field(
default=None,
metadata={"help": "The number of processes to use for the preprocessing."},
)
max_source_length: Optional[int] = field(
default=1024,
metadata={
"help": (
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
max_target_length: Optional[int] = field(
default=128,
metadata={
"help": (
"The maximum total sequence length for target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded."
)
},
)
val_max_target_length: Optional[int] = field(
default=None,
metadata={
"help": (
"The maximum total sequence length for validation target text after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded. Will default to `max_target_length`."
"This argument is also used to override the ``max_length`` param of ``model.generate``, which is used "
"during ``evaluate`` and ``predict``."
)
},
)
pad_to_max_length: bool = field(
default=False,
metadata={
"help": (
"Whether to pad all samples to model maximum sentence length. "
"If False, will pad the samples dynamically when batching to the maximum length in the batch. More "
"efficient on GPU but very bad for TPU."
)
},
)
max_train_samples: Optional[int] = field(
default=None,
metadata={
"help": (
"For debugging purposes or quicker training, truncate the number of training examples to this "
"value if set."
)
},
)
max_eval_samples: Optional[int] = field(
default=None,
metadata={
"help": (
"For debugging purposes or quicker training, truncate the number of evaluation examples to this "
"value if set."
)
},
)
max_predict_samples: Optional[int] = field(
default=None,
metadata={
"help": (
"For debugging purposes or quicker training, truncate the number of prediction examples to this "
"value if set."
)
},
)
num_beams: Optional[int] = field(
default=None,
metadata={
"help": (
"Number of beams to use for evaluation. This argument will be passed to ``model.generate``, "
"which is used during ``evaluate`` and ``predict``."
)
},
)
ignore_pad_token_for_loss: bool = field(
default=True,
metadata={
"help": "Whether to ignore the tokens corresponding to padded labels in the loss computation or not."
},
)
source_prefix: Optional[str] = field(
default="", metadata={"help": "A prefix to add before every source text (useful for T5 models)."}
)
forced_bos_token: Optional[str] = field(
default=None,
metadata={
"help": (
"The token to force as the first generated token after the decoder_start_token_id."
"Useful for multilingual models like mBART where the first generated token"
"needs to be the target language token (Usually it is the target language token)"
)
},
)
def __post_init__(self):
if self.dataset_name is None and self.train_file is None and self.validation_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if self.train_file is not None:
extension = self.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if self.validation_file is not None:
extension = self.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
if self.val_max_target_length is None:
self.val_max_target_length = self.max_target_length
def convert_data(code_file, desc_file):
code_data, desc_data = [], []
with open(code_file, 'r') as f1, open(desc_file, 'r') as f2:
temp_code_data = f1.read()
temp_desc_data = f2.read()
temp_code_data = temp_code_data.split('\n')
temp_desc_data = temp_desc_data.split('\n')
print('Converting data...')
for i in tqdm(range(len(temp_code_data))):
code, desc = temp_code_data[i].split(), temp_desc_data[i].split()
if 200 >= len(code) >= 4 and 60 >= len(desc) >= 4:
code_data.append(temp_code_data[i])
desc_data.append(temp_desc_data[i].lower())
return code_data, desc_data
parser = HfArgumentParser((ModelArguments, DataTrainingArguments, Seq2SeqTrainingArguments))
model_args, data_args, training_args = parser.parse_args_into_dataclasses([
"--train_file", "train_data_text_label.json",
"--validation_file", "val_data_text_label.json",
"--test_file", "test_data_text_label.json",
"--model_name_or_path", "huggingface/CodeBERTa-small-v1",
"--dataset_name", "codesc",
"--output_dir", "/localscratch/vjain312/pl2text",
"--source_lang", "java",
"--target_lang", "en",
"--ignore_pad_token_for_loss", "True",
"--do_train", "True",
"--do_eval", "False",
"--do_predict", "True",
"--learning_rate", "5e-05",
"--per_device_train_batch_size", "16",
"--per_device_eval_batch_size", "16",
"--num_train_epochs", "2",
"--overwrite_output_dir",
"--run_name", "codebert",
"--logging_steps", "20",
"--save_strategy", "epoch",
"--evaluation_strategy", "steps",
"--eval_steps", "4000",
"--save_total_limit", "1",
"--max_source_length", "250",
"--max_target_length", "80",
"--predict_with_generate", "True"
])
set_seed(training_args.seed)
SRC_LANG, TGT_LANG = f"<{data_args.source_lang}>", f"<{data_args.target_lang}>"
if 'checkpoint' in model_args.model_name_or_path:
training_args.output_dir = os.path.join(training_args.output_dir, f'{data_args.dataset_name}_codebert')
training_args.run_name = f'{training_args.run_name}_{data_args.dataset_name}_{data_args.source_lang}_{data_args.target_lang}'
else:
training_args.output_dir = os.path.join(training_args.output_dir, f'{data_args.dataset_name}_codebert')
training_args.run_name = f'{training_args.run_name}_{data_args.dataset_name}_{data_args.source_lang}_{data_args.target_lang}'
if not os.path.exists(training_args.output_dir):
os.makedirs(training_args.output_dir)
if data_args.dataset_name == "codesearchnet":
code_data_train, desc_data_train = convert_data(f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.code.train.txt",
f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.desc.train.txt")
elif data_args.dataset_name == "codesc":
code_data_train, desc_data_train = convert_data(f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.code.train.txt",
f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.desc.train.txt")
with open(os.path.join(training_args.output_dir, 'train_data_text_label.json'), 'w') as openfile:
for code, desc in zip(code_data_train, desc_data_train):
temp = dict()
temp["text"] = code
temp["label"] = desc
json.dump(temp, openfile)
openfile.write('\n')
if data_args.dataset_name == "codesearchnet":
code_data_val, desc_data_val = convert_data(f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.code.val.txt",
f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.desc.val.txt")
elif data_args.dataset_name == "codesc":
code_data_val, desc_data_val = convert_data(f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.code.val.txt",
f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.desc.val.txt")
with open(os.path.join(training_args.output_dir, 'val_data_text_label.json'), 'w') as openfile:
for code, desc in zip(code_data_val, desc_data_val):
temp = dict()
temp["text"] = code
temp["label"] = desc
json.dump(temp, openfile)
openfile.write('\n')
if data_args.dataset_name == "codesearchnet":
code_data_test, desc_data_test = convert_data(f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.code.test.txt",
f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.desc.test.txt")
elif data_args.dataset_name == "codesc":
code_data_test, desc_data_test = convert_data(f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.code.test.txt",
f"/localscratch/vjain312/DL-project-data/{data_args.dataset_name}/{data_args.source_lang}.desc.test.txt")
with open(os.path.join(training_args.output_dir, 'test_data_text_label.json'), 'w') as openfile:
for code, desc in zip(code_data_test, desc_data_test):
temp = dict()
temp["text"] = code
temp["label"] = desc
json.dump(temp, openfile)
openfile.write('\n')
extension = "json"
raw_train_dataset = load_dataset(extension, data_files=os.path.join(training_args.output_dir, data_args.train_file), split="train")
raw_validation_dataset = load_dataset(extension, data_files=os.path.join(training_args.output_dir, data_args.validation_file), split="train")
raw_test_dataset = load_dataset(extension, data_files=os.path.join(training_args.output_dir, data_args.test_file), split="train")
# model = EncoderDecoderModel.from_encoder_decoder_pretrained(model_args.model_name_or_path, "gpt2")
encoder = RobertaModel.from_pretrained(model_args.model_name_or_path)
roberta_tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path,
use_fast=model_args.use_fast_tokenizer,
)
roberta_tokenizer.bos_token = roberta_tokenizer.cls_token
roberta_tokenizer.eos_token = roberta_tokenizer.sep_token
decoder_config = GPT2Config()
decoder_config.vocab_size = len(roberta_tokenizer)
decoder_config.n_layer = 6
decoder = GPT2LMHeadModel(decoder_config)
model_config = EncoderDecoderConfig.from_encoder_decoder_configs(encoder.config, decoder.config)
# model.decoder.config.use_cache = False
# make sure GPT2 appends EOS in begin and end
def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
outputs = [self.bos_token_id] + token_ids_0 + [self.eos_token_id]
return outputs
# GPT2Tokenizer.build_inputs_with_special_tokens = build_inputs_with_special_tokens
# gpt2_tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
# set pad_token_id to unk_token_id -> be careful here as unk_token_id == eos_token_id == bos_token_id
# gpt2_tokenizer.pad_token = gpt2_tokenizer.unk_token
# model.config.decoder_start_token_id = gpt2_tokenizer.bos_token_id
model_config.decoder.decoder_start_token_id = roberta_tokenizer.eos_token_id
model_config.decoder_start_token_id = roberta_tokenizer.eos_token_id
# model.config.eos_token_id = gpt2_tokenizer.eos_token_id
model_config.eos_token_id = roberta_tokenizer.eos_token_id
model_config.decoder.eos_token_id = roberta_tokenizer.eos_token_id
model_config.decoder.bos_token_id = roberta_tokenizer.bos_token_id
model_config.decoder.sep_token_id = roberta_tokenizer.eos_token_id
model_config.decoder.pad_token_id = roberta_tokenizer.pad_token_id
model_config.decoder.early_stopping = True
model_config.decoder.num_beams = 10
model_config.decoder.no_repeat_ngram_size = 3
model_config.decoder.max_length = 80
model_config.decoder.min_length = 4
model_config.decoder.length_penalty = 2.0
model_config.max_length = 80
model_config.min_length = 20
model_config.no_repeat_ngram_size = 3
model_config.decoder.add_cross_attention = True
model = EncoderDecoderModel(model_config)
model.early_stopping = True
model.length_penalty = 2.0
model.num_beams = 10
# tokenizer = AutoTokenizer.from_pretrained(
# model_args.model_name_or_path,
# use_fast=model_args.use_fast_tokenizer,
# )
# model = EncoderDecoderModel.from_encoder_decoder_pretrained(
# model_args.model_name_or_path,
# model_args.model_name_or_path,
# )
# tokenizer.add_tokens([SRC_LANG, TGT_LANG], special_tokens=True)
# tokenizer.bos_token = tokenizer.cls_token
# tokenizer.eos_token = tokenizer.sep_token
# model.encoder.resize_token_embeddings(len(tokenizer))
# model.decoder.resize_token_embeddings(len(tokenizer))
# model.config.decoder_start_token_id = tokenizer.eos_token_id #tokenizer.convert_tokens_to_ids(TGT_LANG)
# model.config.eos_token_id = tokenizer.eos_token_id
# model.config.pad_token_id = tokenizer.pad_token_id
# model.config.vocab_size = model.config.decoder.vocab_size
# model.config.max_length = 80
# model.config.min_length = 4
# model.config.no_repeat_ngram_size = 3
# model.config.early_stopping = True
# model.config.length_penalty = 2.0
# model.config.num_beams = training_args.generation_num_beams
print()
src_text = raw_train_dataset[0]["text"]
print(f'Raw code: {src_text}')
inputs = roberta_tokenizer(src_text, max_length=data_args.max_source_length, padding="max_length", return_tensors="pt").input_ids
print(f'Tokenized code: {roberta_tokenizer.convert_ids_to_tokens(list(inputs[0]))}')
print()
tgt_text = raw_train_dataset[0]["label"]
print(f'Raw description: {tgt_text}')
labels = roberta_tokenizer(tgt_text, max_length=data_args.max_target_length, padding="max_length", return_tensors="pt").input_ids
print(f'Tokenized description: {roberta_tokenizer.convert_ids_to_tokens(list(labels[0]))}')
print()
max_target_length = data_args.max_target_length
padding = "max_length" if data_args.pad_to_max_length else False
if training_args.label_smoothing_factor > 0 and not hasattr(model, "prepare_decoder_input_ids_from_labels"):
logger.warning(
"label_smoothing is enabled but the `prepare_decoder_input_ids_from_labels` method is not defined for"
f"`{model.__class__.__name__}`. This will lead to loss being calculated twice and will take up more memory"
)
prefix = data_args.source_prefix if data_args.source_prefix is not None else ""
padding = "max_length" if data_args.pad_to_max_length else False
def preprocess_function(batch):
# for i in range(len(batch["text"])):
# batch['text'][i] = f"{tokenizer.convert_ids_to_tokens(tokenizer.bos_token_id)}{batch['text'][i]}{tokenizer.convert_ids_to_tokens(tokenizer.eos_token_id)}{SRC_LANG}"
# batch['label'][i] = f"{tokenizer.convert_ids_to_tokens(tokenizer.bos_token_id)}{batch['label'][i]}{tokenizer.convert_ids_to_tokens(tokenizer.eos_token_id)}{TGT_LANG}"
inputs = roberta_tokenizer(batch["text"], max_length=data_args.max_source_length, padding="max_length", truncation=True)
labels = roberta_tokenizer(batch["label"], max_length=data_args.max_target_length, padding="max_length", truncation=True)
batch["input_ids"] = inputs.input_ids
batch["attention_mask"] = inputs.attention_mask
batch["decoder_input_ids"] = labels.input_ids
batch["decoder_attention_mask"] = labels.attention_mask
batch["labels"] = labels.input_ids.copy()
# if padding == "max_length" and data_args.ignore_pad_token_for_loss:
# batch["labels"] = [[-100 if token == gpt2_tokenizer.pad_token_id else token for token in labels] for labels in batch["labels"]]
batch["labels"] = [
[-100 if mask == 0 else token for mask, token in mask_and_tokens] for mask_and_tokens in [zip(masks, labels) for masks, labels in zip(batch["decoder_attention_mask"], batch["labels"])]
]
return batch
column_names = raw_train_dataset.column_names
train_dataset = raw_train_dataset.map(
preprocess_function,
batched=True,
batch_size=training_args.per_device_train_batch_size,
remove_columns=column_names,
num_proc=data_args.preprocessing_num_workers,
desc="Running tokenizer on train dataset",
)
val_dataset = raw_validation_dataset.map(
preprocess_function,
batched=True,
batch_size=training_args.per_device_train_batch_size,
remove_columns=column_names,
num_proc=data_args.preprocessing_num_workers,
desc="Running tokenizer on validation dataset"
)
test_dataset = raw_test_dataset.map(
preprocess_function,
batched=True,
batch_size=training_args.per_device_train_batch_size,
remove_columns=column_names,
num_proc=data_args.preprocessing_num_workers,
desc="Running tokenizer on test dataset",
)
print('Vocab size: ', model.config.encoder.vocab_size)
print('Decoder start token: ', roberta_tokenizer.convert_ids_to_tokens(model.config.decoder_start_token_id))
ids = list(train_dataset[0]["input_ids"])
print(f'Example input ids: {ids}')
print(f'Example input tokens: {roberta_tokenizer.convert_ids_to_tokens(ids)}')
print()
label_ids = list(train_dataset[0]["decoder_input_ids"])
print(f'Example label ids: {label_ids}')
print(f'Example label tokens: {roberta_tokenizer.convert_ids_to_tokens(label_ids)}')
print()
sacrebleu = evaluate.load("sacrebleu")
bleu = evaluate.load("bleu")
rouge = evaluate.load("rouge")
def compute_metrics(eval_preds):
label_ids = eval_preds.label_ids
pred_ids = eval_preds.predictions
pred_str = roberta_tokenizer.batch_decode(pred_ids, skip_special_tokens=True)
label_ids[label_ids == -100] = roberta_tokenizer.eos_token_id
label_str = roberta_tokenizer.batch_decode(label_ids, skip_special_tokens=True)
result = {}
result["sacrebleu"] = sacrebleu.compute(predictions=pred_str, references=label_str)
result["bleu_1"] = bleu.compute(predictions=pred_str, references=label_str, max_order=1)
result["bleu_2"] = bleu.compute(predictions=pred_str, references=label_str, max_order=2)
result["bleu_3"] = bleu.compute(predictions=pred_str, references=label_str, max_order=3)
result["bleu_4"] = bleu.compute(predictions=pred_str, references=label_str, max_order=4)
result["rouge"] = rouge.compute(predictions=pred_str, references=label_str)
# result = {k: round(v, 4) for k, v in result.items()}
return result
# Initialize our Trainer
trainer = Seq2SeqTrainer(
model=model,
args=training_args,
train_dataset=train_dataset if training_args.do_train else None,
eval_dataset=val_dataset if training_args.do_eval else None,
tokenizer=roberta_tokenizer,
compute_metrics=compute_metrics
)
if training_args.do_train:
train_result = trainer.train()
trainer.save_model() # Saves the tokenizer too for easy upload
trainer.save_metrics("train", train_result.metrics)
print(train_result)
if training_args.do_predict:
predict_results = trainer.predict(test_dataset, metric_key_prefix="predict", max_length=80, num_beams=10)
trainer.save_metrics("predict", predict_results.metrics)
print(predict_results)
predictions = roberta_tokenizer.batch_decode(
predict_results.predictions, skip_special_tokens=True, clean_up_tokenization_spaces=True
)
predictions = [pred.strip() for pred in predictions]
output_prediction_file = os.path.join(training_args.output_dir, "generated_predictions.txt")
with open(output_prediction_file, "w", encoding="utf-8") as writer:
writer.write("\n".join(predictions))