-
Notifications
You must be signed in to change notification settings - Fork 2
/
run.py
948 lines (837 loc) · 33.2 KB
/
run.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
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
#!/usr/bin/env python
# coding=utf-8
# Copyright The HuggingFace Team and The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Fine-tuning a 🤗 Transformers model on summarization.
"""
# You can also adapt this script on your own summarization task. Pointers for this are left as comments.
import os
import re
import argparse
import logging
import math
from random import choice
import datasets
import numpy as np
import torch
from datasets import load_dataset, load_metric
from torch.utils.data import DataLoader
from tqdm.auto import tqdm
import transformers
from accelerate import Accelerator
from transformers import (
CONFIG_MAPPING,
MODEL_MAPPING,
AdamW,
AutoConfig,
AutoModelForSeq2SeqLM,
AutoTokenizer,
DataCollatorWithPadding,
DataCollatorForSeq2Seq,
SchedulerType,
get_scheduler,
set_seed,
)
from models.unified.prefixtuning import Model as T5PrefixTuning
from parent import parent_score
logger = logging.getLogger(__name__)
# You should update this to your particular problem to have better documentation of `model_type`
MODEL_CONFIG_CLASSES = list(MODEL_MAPPING.keys())
MODEL_TYPES = tuple(conf.model_type for conf in MODEL_CONFIG_CLASSES)
def convert_text(text):
#return text
text = text.lower()
text = ' '.join(re.split('(\W)', text))
text = ' '.join(text.split())
return text
class Bert:
def __init__(self, location):
self.location = location
class PrefixTuning:
def __init__(self, prefix_sequence_length):
self.prefix_sequence_length = prefix_sequence_length
self.mid_dim = 512
self.prefix_dropout = 0.0
class Model:
def __init__(self):
self.use_description = False
self.knowledge_usage = "concatenate"
self.freeze_plm = True
self.freeze_prefix = False
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument(
"--dataset_name",
type=str,
default="webnlg",
choices=["webnlg", "dart"],
help="The name of the dataset to use.",
)
parser.add_argument(
"--dataset_config_name",
type=str,
default=None,
help="The configuration name of the dataset to use (via the datasets library).",
)
parser.add_argument(
"--train_file",
type=str,
default=None,
help="A csv or a json file containing the training data."
)
parser.add_argument(
"--valid_file",
type=str,
default=None,
help="A csv or a json file containing the validation data (also the data for generation if --do_gen is passed)."
)
parser.add_argument(
"--ignore_pad_token_for_loss",
type=bool,
default=True,
help="Whether to ignore the tokens corresponding to "
"padded labels in the loss computation or not.",
)
parser.add_argument(
"--max_source_length",
type=int,
default=512,
help="The maximum total input sequence length after "
"tokenization.Sequences longer than this will be truncated, sequences shorter will be padded.",
)
parser.add_argument(
"--source_prefix",
type=str,
default='paraphrase: ',
help="A prefix to add before every source text " "(useful for T5 models).",
)
parser.add_argument(
"--preprocessing_num_workers",
type=int,
default=None,
help="The number of processes to use for the preprocessing.",
)
parser.add_argument(
"--max_target_length",
type=int,
default=512,
help="The maximum total sequence length for target text after "
"tokenization. Sequences longer than this will be truncated, sequences shorter will be padded."
"during ``evaluate`` and ``predict``.",
)
parser.add_argument(
"--val_max_target_length",
type=int,
default=None,
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``.",
)
parser.add_argument(
"--num_beams",
type=int,
default=None,
help="Number of beams to use for evaluation. This argument will be "
"passed to ``model.generate``, which is used during ``evaluate`` and ``predict``.",
)
parser.add_argument(
"--pad_to_max_length",
action="store_true",
help="If passed, pad all samples to `max_length`. Otherwise, dynamic padding is used.",
)
parser.add_argument(
"--model_name_or_path",
type=str,
default=None,
help="Path to pretrained model or model identifier from huggingface.co/models.",
required=True,
)
parser.add_argument(
"--config_name",
type=str,
default=None,
help="Pretrained config name or path if not the same as model_name",
)
parser.add_argument(
"--tokenizer_name",
type=str,
default=None,
help="Pretrained tokenizer name or path if not the same as model_name",
)
parser.add_argument(
"--input_column",
type=str,
default="input",
help="The name of the column in the datasets containing the inputs.",
)
parser.add_argument(
"--target_column",
type=str,
default="ref",
help="The name of the column in the datasets containing the targets.",
)
parser.add_argument(
"--use_slow_tokenizer",
action="store_true",
help="If passed, will use a slow tokenizer (not backed by the 🤗 Tokenizers library).",
)
parser.add_argument(
"--per_device_train_batch_size",
type=int,
default=8,
help="Batch size (per device) for the training dataloader.",
)
parser.add_argument(
"--per_device_eval_batch_size",
type=int,
default=8,
help="Batch size (per device) for the evaluation dataloader.",
)
parser.add_argument(
"--learning_rate",
type=float,
default=5e-5,
help="Initial learning rate (after the potential warmup period) to use.",
)
parser.add_argument(
"--weight_decay",
type=float,
default=0.0,
help="Weight decay to use."
)
parser.add_argument(
"--num_train_epochs",
type=int,
default=3,
help="Total number of training epochs to perform."
)
parser.add_argument(
"--max_train_steps",
type=int,
default=None,
help="Total number of training steps to perform. If provided, overrides num_train_epochs.",
)
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument(
"--lr_scheduler_type",
type=SchedulerType,
default="linear",
help="The scheduler type to use.",
choices=["linear", "cosine", "cosine_with_restarts", "polynomial", "constant", "constant_with_warmup"],
)
parser.add_argument(
"--num_warmup_steps",
type=int,
default=0,
help="Number of steps for the warmup in the lr scheduler."
)
parser.add_argument(
"--output_dir",
type=str,
default=None,
help="Where to store the final model."
)
parser.add_argument(
"--seed",
type=int,
default=100,
help="A random seed for reproducible training."
)
parser.add_argument(
"--model_type",
type=str,
default=None,
help="Model type to use if training from scratch.",
choices=MODEL_TYPES,
)
parser.add_argument(
"--use_sacrebleu",
action="store_true",
help="Use sacrebleu for evaluation."
)
parser.add_argument(
"--logging_steps",
type=int,
default=50,
help='Log training information every N steps'
)
parser.add_argument(
"--no_progress_bar",
action="store_true",
help="If passed, will log information of the training progress regularly instead of showing a progress bar"
)
parser.add_argument(
"--do_eval",
action="store_true",
help="If passed, only do evaluation"
)
parser.add_argument(
"--do_gen",
action="store_true",
help="If passed, only do generation"
)
parser.add_argument(
"--eval_after_each_epoch",
type=int,
default=1,
help="Do evaluation when one epoch training ends."
)
parser.add_argument(
"--eval_steps",
type=int,
default=0,
help="Evaluate the model every N steps"
)
parser.add_argument(
"--add_special_tokens",
action="store_true",
help="Add <H>, <R>, <T> tokens for graph translation"
)
parser.add_argument(
"--early_stopping_patience",
type=int,
default=15,
help="The patience for early stopping"
)
parser.add_argument(
"--prefix_tuning",
action="store_true",
help="If set, do prefix tuning instead of fine-tuning"
)
parser.add_argument(
"--prefix_sequence_length",
type=int,
default=10,
help="The length of prefix sequence for prefix tuning"
)
parser.add_argument(
"--not_use_existing_ref",
action="store_true",
help="If set, write reference file instead of using existing ones for evaluation"
)
args = parser.parse_args()
# Sanity checks
if args.dataset_name is None and args.train_file is None and args.valid_file is None:
raise ValueError("Need either a dataset name or a training/validation file.")
else:
if args.train_file is not None:
extension = args.train_file.split(".")[-1]
assert extension in ["csv", "json"], "`train_file` should be a csv or a json file."
if args.valid_file is not None:
extension = args.valid_file.split(".")[-1]
assert extension in ["csv", "json"], "valid_file` should be a csv or a json file."
if args.prefix_tuning:
args.bert = Bert(args.model_name_or_path)
args.prefix_tuning = PrefixTuning(args.prefix_sequence_length)
args.model = Model()
return args
# class for early stopping
class EarlyStopping:
def __init__(self, patience):
self.patience = patience
self.counter = 0
self.best_score = -float("inf")
def __call__(self, score=None):
if score != None:
if score < self.best_score:
self.counter += 1
else:
self.best_score = score
self.counter = 0
return self.counter == self.patience
def train(args, logger, datasets, tokenizer, model, config, accelerator):
label_pad_token_id = -100 if args.ignore_pad_token_for_loss else tokenizer.pad_token_id
data_collator = DataCollatorForSeq2Seq(
tokenizer,
model=model,
label_pad_token_id=label_pad_token_id,
pad_to_multiple_of=8 if accelerator.use_fp16 else None,
)
dataset = datasets["train"]
dataloader = DataLoader(
dataset,
shuffle=True,
collate_fn=data_collator,
batch_size=args.per_device_train_batch_size,
)
# Optimizer
# Split weights in two groups, one with weight decay and the other not.
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{
"params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{
"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)],
"weight_decay": 0.0,
},
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate)
optimizer, model, dataloader = accelerator.prepare(optimizer, model, dataloader)
# Note -> the training dataloader needs to be prepared before we grab his length below (cause its length will be
# shorter in multiprocess)
# Scheduler and math around the number of training steps.
num_update_steps_per_epoch = math.ceil(len(dataloader) / args.gradient_accumulation_steps)
if args.max_train_steps is None:
args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
else:
args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
lr_scheduler = get_scheduler(
name=args.lr_scheduler_type,
optimizer=optimizer,
num_warmup_steps=args.num_warmup_steps,
num_training_steps=args.max_train_steps,
)
def save_checkpoint(checkpoint_dir, model):
if args.output_dir is not None:
save_dir = "{}/{}".format(args.output_dir, checkpoint_dir)
accelerator.wait_for_everyone()
unwrapped_model = accelerator.unwrap_model(model)
# disable logging in save_pretrained function
transformers.utils.logging.set_verbosity_error()
unwrapped_model.save_pretrained(
save_dir,
save_function=accelerator.save,
)
if accelerator.is_main_process:
tokenizer.save_pretrained(save_dir)
# enable logging again
if accelerator.is_local_main_process:
transformers.utils.logging.set_verbosity_info()
# Train!
total_batch_size = args.per_device_train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
logger.info("***** Running training *****")
logger.info(f" Num examples = {len(dataset)}")
logger.info(f" Num Epochs = {args.num_train_epochs}")
logger.info(f" Instantaneous batch size per device = {args.per_device_train_batch_size}")
logger.info(f" Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
logger.info(f" Gradient Accumulation steps = {args.gradient_accumulation_steps}")
logger.info(f" Total optimization steps = {args.max_train_steps}")
if not args.no_progress_bar:
# Only show the progress bar once on each machine.
progress_bar = tqdm(range(args.max_train_steps), disable=not accelerator.is_local_main_process)
completed_steps = 0
tr_loss, logging_loss = 0., 0.
best_score = 0.
early_stopping = EarlyStopping(args.early_stopping_patience)
for epoch in range(args.num_train_epochs):
model.train()
for step, batch in enumerate(dataloader):
outputs = model(**batch)
if args.prefix_tuning:
loss = outputs["loss"]
else:
loss = outputs.loss
loss = loss / args.gradient_accumulation_steps
accelerator.backward(loss)
tr_loss += loss.item()
if (step + 1) % args.gradient_accumulation_steps == 0 or step == len(dataloader) - 1:
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
completed_steps += 1
if not args.no_progress_bar:
progress_bar.update(1)
if completed_steps % args.logging_steps == 0:
cur_loss = (tr_loss - logging_loss) / args.logging_steps
if args.no_progress_bar:
logger.info(f" Epoch={epoch+1}, step={completed_steps}, loss={cur_loss:.4f}")
else:
progress_bar.set_postfix(
{"loss": cur_loss}
)
logging_loss = tr_loss
if args.eval_steps > 0 and completed_steps % args.eval_steps == 0:
score = eval(
args,
logger,
datasets,
tokenizer,
accelerator.unwrap_model(model),
config,
accelerator
)
if score >= best_score:
best_score = score
save_checkpoint("checkpoint_best", model)
save_checkpoint("checkpoint_last", model)
logger.info(f" Evaluation after step {completed_steps}: score={score:.2f}, best score={best_score:.2f}")
if early_stopping(score):
break
if completed_steps >= args.max_train_steps:
break
save_checkpoint("checkpoint_last", model)
if args.eval_after_each_epoch:
score = eval(
args,
logger,
datasets,
tokenizer,
accelerator.unwrap_model(model),
config,
accelerator
)
if score >= best_score:
best_score = score
save_checkpoint("checkpoint_best", model)
logger.info(f" Evaluation after epoch {epoch+1}: score={score:.2f}, best score={best_score:.2f}")
if early_stopping(score):
break
def eval(args, logger, datasets, tokenizer, model, config, accelerator):
data_collator = DataCollatorWithPadding(
tokenizer,
pad_to_multiple_of=8 if accelerator.use_fp16 else None,
)
dataset = datasets["valid"]
dataset, refs = dataset["dataset"], dataset["refs"]
if not args.use_sacrebleu:
refs, triples = refs["refs"], refs["triples"]
dataloader = DataLoader(
dataset,
collate_fn=data_collator,
batch_size=args.per_device_eval_batch_size,
)
model, dataloader = accelerator.prepare(model, dataloader)
# Metric
if args.use_sacrebleu:
metric = load_metric("sacrebleu")
else:
metric = None
# Evaluate!
model.eval()
if args.val_max_target_length is not None:
max_length = args.val_max_target_length
elif args.max_target_length is not None:
max_length = args.max_target_length
else:
max_length = config.max_length
gen_kwargs = {
"max_length": max_length,
"num_beams": args.num_beams,
}
if args.do_eval:
logger.info("***** Running evaluation *****")
logger.info(f" Num examples = {len(dataset)}")
logger.info(f" Total evaluation batch size = {args.per_device_eval_batch_size}")
output_preds = []
for step, batch in enumerate(dataloader):
with torch.no_grad():
generated_tokens = accelerator.unwrap_model(model).generate(
batch["input_ids"],
attention_mask=batch["attention_mask"],
**gen_kwargs,
)
generated_tokens = accelerator.pad_across_processes(
generated_tokens, dim=1, pad_index=tokenizer.pad_token_id
)
generated_tokens = accelerator.gather(generated_tokens).cpu().numpy()
if isinstance(generated_tokens, tuple):
generated_tokens = generated_tokens[0]
decoded_preds = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
output_preds += decoded_preds
# The accelerator dataloader will return the first few items in the dataset
# at the end of the epoch to keep the batches on the same size on all GPUs
# Thus we need to truncate the output
output_preds = output_preds[:len(refs)]
if metric is not None:
metric.add_batch(predictions=output_preds, references=refs)
score = metric.compute()["score"]
if args.do_eval:
logger.info(f" Evaluation results: score={score:.2f}")
# TODO: not sure its behavior in multi-processing
else:
dir_path = os.path.dirname(os.path.realpath(__file__)) + "/data"
pred_path = f"{args.output_dir}/predictions.txt"
with open(pred_path, "w", encoding="utf-8") as f:
f.writelines(convert_text(s) + "\n" for s in output_preds)
split = re.findall(r".*/(.*).json", args.valid_file)[0]
if args.not_use_existing_ref:
file_prefix = f"{args.output_dir}/{split}.target"
else:
file_prefix = os.path.dirname(os.path.realpath(__file__)) + f"/ref_data/{args.dataset_name}/{split}.target"
if args.not_use_existing_ref:
# multi-bleu.perl for evaluation
with open(f"{file_prefix}_eval", "w") as f1, open(f"{file_prefix}2_eval", "w") as f2, open(f"{file_prefix}3_eval", "w") as f3:
for flag, ref in enumerate(refs):
if flag > 0:
text_prefix = "\n"
else:
text_prefix = ""
for i, f in enumerate([f1, f2, f3]):
if i < len(ref):
text = text_prefix + convert_text(ref[i])
f.write(text)
cmd_string = f"perl {dir_path}/multi-bleu.perl -lc " \
+ f"{file_prefix}_eval {file_prefix}2_eval {file_prefix}3_eval " \
+ f"< {pred_path}"
res = os.popen(cmd_string).read()
score = float(re.findall(r"BLEU = (.*), .*\(.*\)", res)[0])
if args.not_use_existing_ref:
os.remove(f"{file_prefix}_eval")
os.remove(f"{file_prefix}2_eval")
os.remove(f"{file_prefix}3_eval")
if args.do_eval:
# meteor
if args.not_use_existing_ref:
text_prefix = ""
with open(f"{file_prefix}_eval_meteor", "w") as f:
for flag, ref in enumerate(refs):
for i in range(3):
if i < len(ref):
text = text_prefix + convert_text(ref[i])
f.write(text)
text_prefix = "\n"
cmd_string = f"java -jar {dir_path}/utils/meteor-1.5.jar {pred_path} " \
+ f"{file_prefix}_eval_meteor -l en -norm -r 3"
res = os.popen(cmd_string).read()
meteor_score = float(res.split()[-1]) * 100
if args.not_use_existing_ref:
os.remove(f"{file_prefix}_eval_meteor")
# parent
_, _, f1 = parent_score(output_preds, refs, triples)
logger.info(f" Evaluation results: BLEU={score:.2f}, METEOR={meteor_score:.2f}, PARENT-F1={100*f1:.2f}")
os.remove(pred_path)
return score
def gen(args, logger, datasets, tokenizer, model, config, accelerator):
data_collator = DataCollatorWithPadding(
tokenizer,
pad_to_multiple_of=8 if accelerator.use_fp16 else None,
)
dataset, refs = datasets["valid"]["dataset"], datasets["valid"]["refs"]
if not args.use_sacrebleu:
refs, triples = refs["refs"], refs["triples"]
dataloader = DataLoader(
dataset,
collate_fn=data_collator,
batch_size=args.per_device_eval_batch_size,
)
model, dataloader = accelerator.prepare(model, dataloader)
# Generate!
model.eval()
if args.val_max_target_length is not None:
max_length = args.val_max_target_length
elif args.max_target_length is not None:
max_length = args.max_target_length
else:
max_length = config.max_length
gen_kwargs = {
"max_length": max_length,
"num_beams": args.num_beams,
}
logger.info("***** Running generation *****")
logger.info(f" Num examples = {len(dataset)}")
logger.info(f" Total evaluation batch size = {args.per_device_eval_batch_size * accelerator.num_processes}")
output_preds = []
for step, batch in enumerate(dataloader):
generated_tokens = accelerator.unwrap_model(model).generate(
batch["input_ids"],
attention_mask=batch["attention_mask"],
**gen_kwargs,
)
generated_tokens = accelerator.pad_across_processes(
generated_tokens, dim=1, pad_index=tokenizer.pad_token_id
)
generated_tokens = accelerator.gather(generated_tokens).cpu().numpy()
if isinstance(generated_tokens, tuple):
generated_tokens = generated_tokens[0]
decoded_preds = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
output_preds += decoded_preds
# The accelerator dataloader will return the first few items in the dataset
# at the end of the epoch to keep the batches on the same size on all GPUs
# Thus we need to truncate the output
output_preds = output_preds[:len(refs)]
with open(f"{args.output_dir}/gen_output.txt", "w", encoding="utf-8") as f:
for pred in output_preds:
f.write(pred + "\n")
logger.info(f" Generation finished!")
def main(args):
args.do_train = not (args.do_eval or args.do_gen)
if args.do_train:
args.use_sacrebleu = True
# Initialize the accelerator. We will let the accelerator handle device placement for us in this example.
accelerator = Accelerator()
logging_handlers = []
if accelerator.is_main_process and args.output_dir is not None:
if args.do_train:
os.makedirs(args.output_dir + "/checkpoint_best", exist_ok=True)
os.makedirs(args.output_dir + "/checkpoint_last", exist_ok=True)
logging_handlers.append(logging.FileHandler(os.path.join(args.output_dir, 'train.log')))
else:
os.makedirs(args.output_dir, exist_ok=True)
if args.do_eval:
logging_handlers.append(logging.FileHandler(os.path.join(args.output_dir, 'eval.log')))
# Make one log on every process with the configuration for debugging.
logging_handlers.append(logging.StreamHandler())
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
handlers=logging_handlers,
)
logger.info(accelerator.state)
# Setup logging, we only want one process per machine to log things on the screen.
# accelerator.is_local_main_process is only True for one process per machine.
logger.setLevel(logging.INFO if accelerator.is_local_main_process else logging.ERROR)
if accelerator.is_local_main_process:
datasets.utils.logging.set_verbosity_warning()
transformers.utils.logging.set_verbosity_info()
else:
datasets.utils.logging.set_verbosity_error()
transformers.utils.logging.set_verbosity_error()
# If passed along, set the training seed now.
if args.seed is not None:
set_seed(args.seed)
accelerator.wait_for_everyone()
if accelerator.is_main_process:
logger.info(args)
# Load pretrained model and tokenizer
#
# In distributed training, the .from_pretrained methods guarantee that only one local process can concurrently
# download model & vocab.
if args.config_name:
config = AutoConfig.from_pretrained(args.config_name)
elif args.model_name_or_path:
config = AutoConfig.from_pretrained(args.model_name_or_path)
else:
config = CONFIG_MAPPING[args.model_type]()
logger.warning("You are instantiating a new config instance from scratch.")
if args.tokenizer_name:
tokenizer = AutoTokenizer.from_pretrained(
args.tokenizer_name,
use_fast=not args.use_slow_tokenizer,
)
elif args.model_name_or_path:
tokenizer = AutoTokenizer.from_pretrained(
args.model_name_or_path,
use_fast=not args.use_slow_tokenizer,
)
else:
raise ValueError(
"You are instantiating a new tokenizer from scratch. "
"This is not supported by this script. "
"You can do it from another script, save it, and load it from here, "
"using --tokenizer_name."
)
if args.add_special_tokens:
new_tokens = ['<H>', '<R>', '<T>']
new_tokens_vocab = {}
new_tokens_vocab['additional_special_tokens'] = []
for idx, t in enumerate(new_tokens):
new_tokens_vocab['additional_special_tokens'].append(t)
num_added_toks = tokenizer.add_special_tokens(new_tokens_vocab)
if args.model_name_or_path:
# prefixtuning
if args.prefix_tuning:
model = T5PrefixTuning(args)
if not args.do_train:
state_dict = torch.load(os.path.join(args.model_name_or_path, transformers.WEIGHTS_NAME), map_location="cpu")
model.load_state_dict(state_dict, strict=True)
# release memory
del state_dict
# finetuning
else:
model = AutoModelForSeq2SeqLM.from_pretrained(
args.model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
)
else:
logger.info("Training new model from scratch")
model = AutoModelForSeq2SeqLM.from_config(config)
'''
model.resize_token_embeddings(len(tokenizer))
if model.config.decoder_start_token_id is None:
raise ValueError(
"Make sure that `config.decoder_start_token_id` is correctly defined"
)
'''
prefix = args.source_prefix if args.source_prefix is not None and not args.prefix_tuning else ""
extension = args.valid_file.split(".")[-1]
raw_datasets = {}
if args.do_train and args.train_file is not None:
raw_datasets["train"] = load_dataset(extension, data_files={"data": args.train_file})
# validation set is multi-references so we don't load it together with training set
if args.valid_file is not None:
raw_datasets["valid"] = load_dataset(extension, data_files={"data": args.valid_file})
# Preprocessing the datasets.
column_names_to_remove = raw_datasets["valid"]["data"].column_names
# Temporarily set max_target_length for training.
max_target_length = args.max_target_length
padding = "max_length" if args.pad_to_max_length else False
def preprocess_function(examples):
inputs = examples[args.input_column]
targets = examples[args.target_column]
inputs = [prefix + inp for inp in inputs]
model_inputs = tokenizer(
inputs,
max_length=args.max_source_length,
padding=padding,
truncation=True,
)
# multiple references
if type(targets[0]) is not list:
# Setup the tokenizer for targets
with tokenizer.as_target_tokenizer():
labels = tokenizer(
targets,
max_length=max_target_length,
padding=padding,
truncation=True,
)
# If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore
# padding in the loss.
if padding == "max_length" and args.ignore_pad_token_for_loss:
labels["input_ids"] = [
[(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
]
model_inputs["labels"] = labels["input_ids"]
return model_inputs
refs = [sample["ref"] for sample in raw_datasets["valid"]["data"]]
# process for sacrebleu
if args.use_sacrebleu:
max_num_ref = max([len(ref) for ref in refs])
refs = [(ref if type(ref) is list else [ref]) + [""] * (max_num_ref - len(ref)) for ref in refs]
# save both refs and triples for PARENT metric computation
else:
triples = [sample["triple"] for sample in raw_datasets["valid"]["data"]]
refs = {"refs":refs, "triples": triples}
with accelerator.main_process_first():
processed_datasets = {}
for split, dataset in raw_datasets.items():
processed_datasets[split] = dataset.map(
preprocess_function,
batched=True,
remove_columns=column_names_to_remove,
load_from_cache_file=False,
desc="Running tokenizer on {} dataset".format(split),
)
if "train" in processed_datasets:
processed_datasets["train"] = processed_datasets["train"]["data"]
processed_datasets["valid"] = {"dataset": processed_datasets["valid"]["data"], "refs": refs}
if args.do_train:
train(args, logger, processed_datasets, tokenizer, model, config, accelerator)
elif args.do_eval:
eval(args, logger, processed_datasets, tokenizer, model, config, accelerator)
else:
gen(args, logger, processed_datasets, tokenizer, model, config, accelerator)
if __name__ == "__main__":
main(parse_args())