run_multiple_choice.py 28.6 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
# Copyright (c) 2018, NVIDIA CORPORATION.  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.
erenup's avatar
erenup committed
16
""" Finetuning the library models for multiple choice (Bert, Roberta, XLNet)."""
17
18
19
20
21
22
23
24
25
26
27
28

from __future__ import absolute_import, division, print_function

import argparse
import glob
import logging
import os
import random


import numpy as np
import torch
29
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
30
from torch.utils.data.distributed import DistributedSampler
31
32
33
34
35
36

try:
    from torch.utils.tensorboard import SummaryWriter
except:
    from tensorboardX import SummaryWriter

37
38
from tqdm import tqdm, trange

39
40
41
42
43
44
45
46
47
48
49
50
from transformers import (
    WEIGHTS_NAME,
    BertConfig,
    BertForMultipleChoice,
    BertTokenizer,
    XLNetConfig,
    XLNetForMultipleChoice,
    XLNetTokenizer,
    RobertaConfig,
    RobertaForMultipleChoice,
    RobertaTokenizer,
)
51

52
from transformers import AdamW, get_linear_schedule_with_warmup
53

54
from utils_multiple_choice import convert_examples_to_features, processors
55
56
57

logger = logging.getLogger(__name__)

58
59
60
ALL_MODELS = sum(
    (tuple(conf.pretrained_config_archive_map.keys()) for conf in (BertConfig, XLNetConfig, RobertaConfig)), ()
)
61
62

MODEL_CLASSES = {
63
64
65
    "bert": (BertConfig, BertForMultipleChoice, BertTokenizer),
    "xlnet": (XLNetConfig, XLNetForMultipleChoice, XLNetTokenizer),
    "roberta": (RobertaConfig, RobertaForMultipleChoice, RobertaTokenizer),
66
67
}

68

69
def select_field(features, field):
70
    return [[choice[field] for choice in feature.choices_features] for feature in features]
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


def simple_accuracy(preds, labels):
    return (preds == labels).mean()


def set_seed(args):
    random.seed(args.seed)
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    if args.n_gpu > 0:
        torch.cuda.manual_seed_all(args.seed)


def train(args, train_dataset, model, tokenizer):
    """ Train the model """
    if args.local_rank in [-1, 0]:
        tb_writer = SummaryWriter()

    args.train_batch_size = args.per_gpu_train_batch_size * max(1, args.n_gpu)
    train_sampler = RandomSampler(train_dataset) if args.local_rank == -1 else DistributedSampler(train_dataset)
    train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size)

    if args.max_steps > 0:
        t_total = args.max_steps
        args.num_train_epochs = args.max_steps // (len(train_dataloader) // args.gradient_accumulation_steps) + 1
    else:
        t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs

    # Prepare optimizer and schedule (linear warmup and decay)
101
    no_decay = ["bias", "LayerNorm.weight"]
102
    optimizer_grouped_parameters = [
103
104
105
106
107
108
        {
            "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},
    ]
109
    optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
110
111
112
    scheduler = get_linear_schedule_with_warmup(
        optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
    )
113
114
115
116
117
118
119
120
121
122
123
124
125
    if args.fp16:
        try:
            from apex import amp
        except ImportError:
            raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
        model, optimizer = amp.initialize(model, optimizer, opt_level=args.fp16_opt_level)

    # multi-gpu training (should be after apex fp16 initialization)
    if args.n_gpu > 1:
        model = torch.nn.DataParallel(model)

    # Distributed training (should be after apex fp16 initialization)
    if args.local_rank != -1:
126
127
128
        model = torch.nn.parallel.DistributedDataParallel(
            model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=True
        )
129
130
131
132
133
134

    # Train!
    logger.info("***** Running training *****")
    logger.info("  Num examples = %d", len(train_dataset))
    logger.info("  Num Epochs = %d", args.num_train_epochs)
    logger.info("  Instantaneous batch size per GPU = %d", args.per_gpu_train_batch_size)
135
136
137
138
139
140
    logger.info(
        "  Total train batch size (w. parallel, distributed & accumulation) = %d",
        args.train_batch_size
        * args.gradient_accumulation_steps
        * (torch.distributed.get_world_size() if args.local_rank != -1 else 1),
    )
141
142
143
144
145
    logger.info("  Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
    logger.info("  Total optimization steps = %d", t_total)

    global_step = 0
    tr_loss, logging_loss = 0.0, 0.0
erenup's avatar
erenup committed
146
    best_dev_acc, best_dev_loss = 0.0, 99999999999.0
erenup's avatar
erenup committed
147
    best_steps = 0
148
149
150
151
152
153
154
155
    model.zero_grad()
    train_iterator = trange(int(args.num_train_epochs), desc="Epoch", disable=args.local_rank not in [-1, 0])
    set_seed(args)  # Added here for reproductibility (even between python 2 and 3)
    for _ in train_iterator:
        epoch_iterator = tqdm(train_dataloader, desc="Iteration", disable=args.local_rank not in [-1, 0])
        for step, batch in enumerate(epoch_iterator):
            model.train()
            batch = tuple(t.to(args.device) for t in batch)
156
157
158
159
160
161
162
163
            inputs = {
                "input_ids": batch[0],
                "attention_mask": batch[1],
                "token_type_ids": batch[2]
                if args.model_type in ["bert", "xlnet"]
                else None,  # XLM don't use segment_ids
                "labels": batch[3],
            }
164
            outputs = model(**inputs)
165
            loss = outputs[0]  # model outputs are always tuple in transformers (see doc)
166
167

            if args.n_gpu > 1:
168
                loss = loss.mean()  # mean() to average on multi-gpu parallel training
169
170
171
172
173
174
175
176
177
178
179
180
181
            if args.gradient_accumulation_steps > 1:
                loss = loss / args.gradient_accumulation_steps

            if args.fp16:
                with amp.scale_loss(loss, optimizer) as scaled_loss:
                    scaled_loss.backward()
                torch.nn.utils.clip_grad_norm_(amp.master_params(optimizer), args.max_grad_norm)
            else:
                loss.backward()
                torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)

            tr_loss += loss.item()
            if (step + 1) % args.gradient_accumulation_steps == 0:
182

183
                optimizer.step()
184
                scheduler.step()  # Update learning rate schedule
185
186
187
188
189
                model.zero_grad()
                global_step += 1

                if args.local_rank in [-1, 0] and args.logging_steps > 0 and global_step % args.logging_steps == 0:
                    # Log metrics
190
191
192
                    if (
                        args.local_rank == -1 and args.evaluate_during_training
                    ):  # Only evaluate when single GPU otherwise metrics may not average well
193
194
                        results = evaluate(args, model, tokenizer)
                        for key, value in results.items():
195
                            tb_writer.add_scalar("eval_{}".format(key), value, global_step)
erenup's avatar
erenup committed
196
                        if results["eval_acc"] > best_dev_acc:
erenup's avatar
erenup committed
197
198
                            best_dev_acc = results["eval_acc"]
                            best_dev_loss = results["eval_loss"]
erenup's avatar
erenup committed
199
200
201
202
                            best_steps = global_step
                            if args.do_test:
                                results_test = evaluate(args, model, tokenizer, test=True)
                                for key, value in results_test.items():
203
204
205
206
207
208
209
210
211
212
213
214
215
216
                                    tb_writer.add_scalar("test_{}".format(key), value, global_step)
                                logger.info(
                                    "test acc: %s, loss: %s, global steps: %s",
                                    str(results_test["eval_acc"]),
                                    str(results_test["eval_loss"]),
                                    str(global_step),
                                )
                    tb_writer.add_scalar("lr", scheduler.get_lr()[0], global_step)
                    tb_writer.add_scalar("loss", (tr_loss - logging_loss) / args.logging_steps, global_step)
                    logger.info(
                        "Average loss: %s at global step: %s",
                        str((tr_loss - logging_loss) / args.logging_steps),
                        str(global_step),
                    )
217
218
219
220
                    logging_loss = tr_loss

                if args.local_rank in [-1, 0] and args.save_steps > 0 and global_step % args.save_steps == 0:
                    # Save model checkpoint
221
                    output_dir = os.path.join(args.output_dir, "checkpoint-{}".format(global_step))
222
223
                    if not os.path.exists(output_dir):
                        os.makedirs(output_dir)
224
225
226
                    model_to_save = (
                        model.module if hasattr(model, "module") else model
                    )  # Take care of distributed/parallel training
227
228
                    model_to_save.save_pretrained(output_dir)
                    tokenizer.save_vocabulary(output_dir)
229
                    torch.save(args, os.path.join(output_dir, "training_args.bin"))
230
231
232
233
234
235
236
237
238
239
240
241
                    logger.info("Saving model checkpoint to %s", output_dir)

            if args.max_steps > 0 and global_step > args.max_steps:
                epoch_iterator.close()
                break
        if args.max_steps > 0 and global_step > args.max_steps:
            train_iterator.close()
            break

    if args.local_rank in [-1, 0]:
        tb_writer.close()

erenup's avatar
erenup committed
242
    return global_step, tr_loss / global_step, best_steps
243
244


erenup's avatar
erenup committed
245
def evaluate(args, model, tokenizer, prefix="", test=False):
246
247
248
249
250
    eval_task_names = (args.task_name,)
    eval_outputs_dirs = (args.output_dir,)

    results = {}
    for eval_task, eval_output_dir in zip(eval_task_names, eval_outputs_dirs):
erenup's avatar
erenup committed
251
        eval_dataset = load_and_cache_examples(args, eval_task, tokenizer, evaluate=not test, test=test)
252
253
254
255
256
257

        if not os.path.exists(eval_output_dir) and args.local_rank in [-1, 0]:
            os.makedirs(eval_output_dir)

        args.eval_batch_size = args.per_gpu_eval_batch_size * max(1, args.n_gpu)
        # Note that DistributedSampler samples randomly
258
        eval_sampler = SequentialSampler(eval_dataset)
259
260
        eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)

ronakice's avatar
ronakice committed
261
262
263
264
        # multi-gpu evaluate
        if args.n_gpu > 1:
            model = torch.nn.DataParallel(model)

265
266
267
268
269
270
271
272
273
274
275
276
277
        # Eval!
        logger.info("***** Running evaluation {} *****".format(prefix))
        logger.info("  Num examples = %d", len(eval_dataset))
        logger.info("  Batch size = %d", args.eval_batch_size)
        eval_loss = 0.0
        nb_eval_steps = 0
        preds = None
        out_label_ids = None
        for batch in tqdm(eval_dataloader, desc="Evaluating"):
            model.eval()
            batch = tuple(t.to(args.device) for t in batch)

            with torch.no_grad():
278
279
280
281
282
283
284
285
                inputs = {
                    "input_ids": batch[0],
                    "attention_mask": batch[1],
                    "token_type_ids": batch[2]
                    if args.model_type in ["bert", "xlnet"]
                    else None,  # XLM don't use segment_ids
                    "labels": batch[3],
                }
286
287
288
289
290
291
292
                outputs = model(**inputs)
                tmp_eval_loss, logits = outputs[:2]

                eval_loss += tmp_eval_loss.mean().item()
            nb_eval_steps += 1
            if preds is None:
                preds = logits.detach().cpu().numpy()
293
                out_label_ids = inputs["labels"].detach().cpu().numpy()
294
295
            else:
                preds = np.append(preds, logits.detach().cpu().numpy(), axis=0)
296
                out_label_ids = np.append(out_label_ids, inputs["labels"].detach().cpu().numpy(), axis=0)
297
298
299
300
301
302
303

        eval_loss = eval_loss / nb_eval_steps
        preds = np.argmax(preds, axis=1)
        acc = simple_accuracy(preds, out_label_ids)
        result = {"eval_acc": acc, "eval_loss": eval_loss}
        results.update(result)

erenup's avatar
erenup committed
304
        output_eval_file = os.path.join(eval_output_dir, "is_test_" + str(test).lower() + "_eval_results.txt")
erenup's avatar
erenup committed
305

306
        with open(output_eval_file, "w") as writer:
erenup's avatar
erenup committed
307
            logger.info("***** Eval results {} *****".format(str(prefix) + " is test:" + str(test)))
erenup's avatar
erenup committed
308
            writer.write("model           =%s\n" % str(args.model_name_or_path))
309
310
311
312
313
314
315
316
            writer.write(
                "total batch size=%d\n"
                % (
                    args.per_gpu_train_batch_size
                    * args.gradient_accumulation_steps
                    * (torch.distributed.get_world_size() if args.local_rank != -1 else 1)
                )
            )
erenup's avatar
erenup committed
317
318
319
            writer.write("train num epochs=%d\n" % args.num_train_epochs)
            writer.write("fp16            =%s\n" % args.fp16)
            writer.write("max seq length  =%d\n" % args.max_seq_length)
320
321
322
323
324
325
            for key in sorted(result.keys()):
                logger.info("  %s = %s", key, str(result[key]))
                writer.write("%s = %s\n" % (key, str(result[key])))
    return results


erenup's avatar
erenup committed
326
def load_and_cache_examples(args, task, tokenizer, evaluate=False, test=False):
327
328
329
330
331
    if args.local_rank not in [-1, 0]:
        torch.distributed.barrier()  # Make sure only the first process in distributed training process the dataset, and the others will use the cache

    processor = processors[task]()
    # Load data features from cache or dataset file
erenup's avatar
erenup committed
332
    if evaluate:
333
        cached_mode = "dev"
erenup's avatar
erenup committed
334
    elif test:
335
        cached_mode = "test"
erenup's avatar
erenup committed
336
    else:
337
        cached_mode = "train"
erenup's avatar
erenup committed
338
    assert (evaluate == True and test == True) == False
339
340
341
342
343
344
345
346
347
    cached_features_file = os.path.join(
        args.data_dir,
        "cached_{}_{}_{}_{}".format(
            cached_mode,
            list(filter(None, args.model_name_or_path.split("/"))).pop(),
            str(args.max_seq_length),
            str(task),
        ),
    )
348
    if os.path.exists(cached_features_file) and not args.overwrite_cache:
349
350
351
352
353
        logger.info("Loading features from cached file %s", cached_features_file)
        features = torch.load(cached_features_file)
    else:
        logger.info("Creating features from dataset file at %s", args.data_dir)
        label_list = processor.get_labels()
erenup's avatar
erenup committed
354
355
356
357
358
359
        if evaluate:
            examples = processor.get_dev_examples(args.data_dir)
        elif test:
            examples = processor.get_test_examples(args.data_dir)
        else:
            examples = processor.get_train_examples(args.data_dir)
360
        logger.info("Training number: %s", str(len(examples)))
361
362
363
364
365
        features = convert_examples_to_features(
            examples,
            label_list,
            args.max_seq_length,
            tokenizer,
366
367
            pad_on_left=bool(args.model_type in ["xlnet"]),  # pad on the left for xlnet
            pad_token_segment_id=4 if args.model_type in ["xlnet"] else 0,
368
        )
369
370
371
372
373
374
375
376
        if args.local_rank in [-1, 0]:
            logger.info("Saving features into cached file %s", cached_features_file)
            torch.save(features, cached_features_file)

    if args.local_rank == 0:
        torch.distributed.barrier()  # Make sure only the first process in distributed training process the dataset, and the others will use the cache

    # Convert to Tensors and build dataset
377
378
379
    all_input_ids = torch.tensor(select_field(features, "input_ids"), dtype=torch.long)
    all_input_mask = torch.tensor(select_field(features, "input_mask"), dtype=torch.long)
    all_segment_ids = torch.tensor(select_field(features, "segment_ids"), dtype=torch.long)
380
381
382
383
384
385
386
387
388
389
    all_label_ids = torch.tensor([f.label for f in features], dtype=torch.long)

    dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_label_ids)
    return dataset


def main():
    parser = argparse.ArgumentParser()

    ## Required parameters
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
    parser.add_argument(
        "--data_dir",
        default=None,
        type=str,
        required=True,
        help="The input data dir. Should contain the .tsv files (or other data files) for the task.",
    )
    parser.add_argument(
        "--model_type",
        default=None,
        type=str,
        required=True,
        help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
    )
    parser.add_argument(
        "--model_name_or_path",
        default=None,
        type=str,
        required=True,
        help="Path to pre-trained model or shortcut name selected in the list: " + ", ".join(ALL_MODELS),
    )
    parser.add_argument(
        "--task_name",
        default=None,
        type=str,
        required=True,
        help="The name of the task to train selected in the list: " + ", ".join(processors.keys()),
    )
    parser.add_argument(
        "--output_dir",
        default=None,
        type=str,
        required=True,
        help="The output directory where the model predictions and checkpoints will be written.",
    )
425
426

    ## Other parameters
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
    parser.add_argument(
        "--config_name", default="", type=str, help="Pretrained config name or path if not the same as model_name"
    )
    parser.add_argument(
        "--tokenizer_name",
        default="",
        type=str,
        help="Pretrained tokenizer name or path if not the same as model_name",
    )
    parser.add_argument(
        "--cache_dir",
        default="",
        type=str,
        help="Where do you want to store the pre-trained models downloaded from s3",
    )
    parser.add_argument(
        "--max_seq_length",
        default=128,
        type=int,
        help="The maximum total input sequence length after tokenization. Sequences longer "
        "than this will be truncated, sequences shorter will be padded.",
    )
    parser.add_argument("--do_train", action="store_true", help="Whether to run training.")
    parser.add_argument("--do_eval", action="store_true", help="Whether to run eval on the dev set.")
    parser.add_argument("--do_test", action="store_true", help="Whether to run test on the test set")
    parser.add_argument(
        "--evaluate_during_training", action="store_true", help="Run evaluation during training at each logging step."
    )
    parser.add_argument(
        "--do_lower_case", action="store_true", help="Set this flag if you are using an uncased model."
    )

    parser.add_argument("--per_gpu_train_batch_size", default=8, type=int, help="Batch size per GPU/CPU for training.")
    parser.add_argument(
        "--per_gpu_eval_batch_size", default=8, type=int, help="Batch size per GPU/CPU for evaluation."
    )
    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("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
    parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight deay if we apply some.")
    parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
    parser.add_argument(
        "--num_train_epochs", default=3.0, type=float, help="Total number of training epochs to perform."
    )
    parser.add_argument(
        "--max_steps",
        default=-1,
        type=int,
        help="If > 0: set total number of training steps to perform. Override num_train_epochs.",
    )
    parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")

    parser.add_argument("--logging_steps", type=int, default=50, help="Log every X updates steps.")
    parser.add_argument("--save_steps", type=int, default=50, help="Save checkpoint every X updates steps.")
    parser.add_argument(
        "--eval_all_checkpoints",
        action="store_true",
        help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number",
    )
    parser.add_argument("--no_cuda", action="store_true", help="Avoid using CUDA when available")
    parser.add_argument(
        "--overwrite_output_dir", action="store_true", help="Overwrite the content of the output directory"
    )
    parser.add_argument(
        "--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
    )
    parser.add_argument("--seed", type=int, default=42, help="random seed for initialization")

    parser.add_argument(
        "--fp16",
        action="store_true",
        help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit",
    )
    parser.add_argument(
        "--fp16_opt_level",
        type=str,
        default="O1",
        help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
        "See details at https://nvidia.github.io/apex/amp.html",
    )
    parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
    parser.add_argument("--server_ip", type=str, default="", help="For distant debugging.")
    parser.add_argument("--server_port", type=str, default="", help="For distant debugging.")
515
516
    args = parser.parse_args()

517
518
519
520
521
522
523
524
525
526
527
    if (
        os.path.exists(args.output_dir)
        and os.listdir(args.output_dir)
        and args.do_train
        and not args.overwrite_output_dir
    ):
        raise ValueError(
            "Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(
                args.output_dir
            )
        )
528
529
530
531
532

    # Setup distant debugging if needed
    if args.server_ip and args.server_port:
        # Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
        import ptvsd
533

534
535
536
537
538
539
540
541
542
543
544
        print("Waiting for debugger attach")
        ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
        ptvsd.wait_for_attach()

    # Setup CUDA, GPU & distributed training
    if args.local_rank == -1 or args.no_cuda:
        device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
        args.n_gpu = torch.cuda.device_count()
    else:  # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
        torch.cuda.set_device(args.local_rank)
        device = torch.device("cuda", args.local_rank)
545
        torch.distributed.init_process_group(backend="nccl")
546
547
548
549
        args.n_gpu = 1
    args.device = device

    # Setup logging
550
551
552
553
554
555
556
557
558
559
560
561
562
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN,
    )
    logger.warning(
        "Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
        args.local_rank,
        device,
        args.n_gpu,
        bool(args.local_rank != -1),
        args.fp16,
    )
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580

    # Set seed
    set_seed(args)

    # Prepare GLUE task
    args.task_name = args.task_name.lower()
    if args.task_name not in processors:
        raise ValueError("Task not found: %s" % (args.task_name))
    processor = processors[args.task_name]()
    label_list = processor.get_labels()
    num_labels = len(label_list)

    # Load pretrained model and tokenizer
    if args.local_rank not in [-1, 0]:
        torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab

    args.model_type = args.model_type.lower()
    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
    config = config_class.from_pretrained(
        args.config_name if args.config_name else args.model_name_or_path,
        num_labels=num_labels,
        finetuning_task=args.task_name,
        cache_dir=args.cache_dir if args.cache_dir else None,
    )
    tokenizer = tokenizer_class.from_pretrained(
        args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
        do_lower_case=args.do_lower_case,
        cache_dir=args.cache_dir if args.cache_dir else None,
    )
    model = model_class.from_pretrained(
        args.model_name_or_path,
        from_tf=bool(".ckpt" in args.model_name_or_path),
        config=config,
        cache_dir=args.cache_dir if args.cache_dir else None,
    )
598
599
600
601
602
603
604

    if args.local_rank == 0:
        torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab

    model.to(args.device)

    logger.info("Training/evaluation parameters %s", args)
605
    best_steps = 0
606
607
608
609

    # Training
    if args.do_train:
        train_dataset = load_and_cache_examples(args, args.task_name, tokenizer, evaluate=False)
610
        global_step, tr_loss, best_steps = train(args, train_dataset, model, tokenizer)
611
612
613
614
615
616
617
618
619
620
621
        logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)

    # Saving best-practices: if you use defaults names for the model, you can reload it using from_pretrained()
    if args.do_train and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
        # Create output directory if needed
        if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
            os.makedirs(args.output_dir)

        logger.info("Saving model checkpoint to %s", args.output_dir)
        # Save a trained model, configuration and tokenizer using `save_pretrained()`.
        # They can then be reloaded using `from_pretrained()`
622
623
624
        model_to_save = (
            model.module if hasattr(model, "module") else model
        )  # Take care of distributed/parallel training
625
626
627
628
        model_to_save.save_pretrained(args.output_dir)
        tokenizer.save_pretrained(args.output_dir)

        # Good practice: save your training arguments together with the trained model
629
        torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
630
631
632
633
634
635
636
637
638
639
640
641
642

        # Load a trained model and vocabulary that you have fine-tuned
        model = model_class.from_pretrained(args.output_dir)
        tokenizer = tokenizer_class.from_pretrained(args.output_dir)
        model.to(args.device)

    # Evaluation
    results = {}
    if args.do_eval and args.local_rank in [-1, 0]:
        if not args.do_train:
            args.output_dir = args.model_name_or_path
        checkpoints = [args.output_dir]
        if args.eval_all_checkpoints:
643
644
645
            checkpoints = list(
                os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + "/**/" + WEIGHTS_NAME, recursive=True))
            )
646
            logging.getLogger("transformers.modeling_utils").setLevel(logging.WARN)  # Reduce logging
647
648
        logger.info("Evaluate the following checkpoints: %s", checkpoints)
        for checkpoint in checkpoints:
649
650
651
            global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
            prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""

652
653
            model = model_class.from_pretrained(checkpoint)
            model.to(args.device)
654
            result = evaluate(args, model, tokenizer, prefix=prefix)
655
            result = dict((k + "_{}".format(global_step), v) for k, v in result.items())
656
657
            results.update(result)

erenup's avatar
erenup committed
658
659
660
661
    if args.do_test and args.local_rank in [-1, 0]:
        if not args.do_train:
            args.output_dir = args.model_name_or_path
        checkpoints = [args.output_dir]
erenup's avatar
erenup committed
662
663
        # if args.eval_all_checkpoints: # can not use this to do test!!
        #     checkpoints = list(os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME, recursive=True)))
664
        #     logging.getLogger("transformers.modeling_utils").setLevel(logging.WARN)  # Reduce logging
erenup's avatar
erenup committed
665
666
        logger.info("Evaluate the following checkpoints: %s", checkpoints)
        for checkpoint in checkpoints:
667
668
669
            global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
            prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""

erenup's avatar
erenup committed
670
671
            model = model_class.from_pretrained(checkpoint)
            model.to(args.device)
672
            result = evaluate(args, model, tokenizer, prefix=prefix, test=True)
673
            result = dict((k + "_{}".format(global_step), v) for k, v in result.items())
erenup's avatar
erenup committed
674
            results.update(result)
675
676
    if best_steps:
        logger.info("best steps of eval acc is the following checkpoints: %s", best_steps)
677
678
679
680
681
    return results


if __name__ == "__main__":
    main()