main.py

# 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.


import argparse
import glob
import logging
import os
import random

import numpy as np
import torch
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm, trange


from transformers import (
    WEIGHTS_NAME,
    AdamW,
    BertConfig,
    BertTokenizer,
    get_linear_schedule_with_warmup,
)
from utils import convert_examples_to_features, processors
from models import BertForMultipleChoice


logger = logging.getLogger(__name__)


MODEL_CLASSES = {
    "bert": (BertConfig, BertForMultipleChoice, BertTokenizer),
}


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


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 """
    args.train_batch_size = args.per_gpu_train_batch_size * max(1, args.n_gpu)
    train_sampler = RandomSampler(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)
    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, eps=args.adam_epsilon)
    scheduler = get_linear_schedule_with_warmup(
        optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
    )

    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)

    # 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)
    logger.info(
        "  Total train batch size (w. parallel, distributed & accumulation) = %d",
        args.train_batch_size
        * args.gradient_accumulation_steps
    )
    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
    best_dev_acc = 0.0
    best_steps = 0
    model.zero_grad()
    train_iterator = trange(int(args.num_train_epochs), desc="Epoch", disable=False)
    set_seed(args)
    for _ in train_iterator:
        epoch_iterator = tqdm(train_dataloader, desc="Iteration", disable=False)
        for step, batch in enumerate(epoch_iterator):
            model.train()
            batch = tuple(t.to(args.device) for t in batch)
            inputs = {
                "input_ids": batch[0],
                "attention_mask": batch[1],
                "token_type_ids": batch[2]
                if args.model_type in ["bert", "xlnet"]
                else None,
                "predicate_positions": batch[3],
                "n_choices": batch[4],
                "labels": batch[5],
            }
            outputs = model(**inputs)
            loss = outputs[0]
            logits = outputs[1]
            preds = logits.detach().cpu().numpy()
            out_label_ids = inputs["labels"].detach().cpu().numpy()
            preds = np.argmax(preds, axis=1)
            acc = simple_accuracy(preds, out_label_ids)

            if args.n_gpu > 1:
                loss = loss.mean()  # mean() to average on multi-gpu parallel training
            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:
                optimizer.step()
                scheduler.step()  # Update learning rate schedule
                model.zero_grad()
                global_step += 1

                if args.logging_steps > 0 and global_step % args.logging_steps == 0:
                    # Log metrics
                    if args.evaluate_during_training:
                        # Only evaluate when single GPU otherwise metrics may not average well
                        results = evaluate(args, model, tokenizer)
                        if results["eval_acc"] > best_dev_acc:
                            best_dev_acc = results["eval_acc"]
                            best_steps = global_step
                    logger.info(
                        "Average loss: %s at global step: %s",
                        str((tr_loss - logging_loss) / args.logging_steps),
                        str(global_step),
                    )
                    logging_loss = tr_loss

                if args.save_steps > 0 and global_step % args.save_steps == 0:
                    # Save model checkpoint
                    output_dir = os.path.join(args.output_dir, "checkpoint-{}".format(global_step))
                    if not os.path.exists(output_dir):
                        os.makedirs(output_dir)
                    model_to_save = (
                        model.module if hasattr(model, "module") else model
                    )  # Take care of distributed/parallel training
                    model_to_save.save_pretrained(output_dir)
                    tokenizer.save_vocabulary(output_dir)
                    torch.save(args, os.path.join(output_dir, "training_args.bin"))
                    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

    return global_step, tr_loss / global_step, best_steps


def evaluate(args, model, tokenizer, prefix="", test=False):
    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):
        eval_dataset = load_and_cache_examples(args, eval_task, tokenizer, evaluate=not test, test=test)

        if not os.path.exists(eval_output_dir):
            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
        eval_sampler = SequentialSampler(eval_dataset)
        eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)

        if args.n_gpu > 1:
            model = torch.nn.DataParallel(model)

        # 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():
                inputs = {
                    "input_ids": batch[0],
                    "attention_mask": batch[1],
                    "token_type_ids": batch[2] if args.model_type in ["bert", "xlnet"] else None,
                    "predicate_positions": batch[3],
                    "n_choices": batch[4],
                    "labels": batch[5] if not test else None,
                }
                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()
                if not test:
                    out_label_ids = inputs["labels"].detach().cpu().numpy()
            else:
                preds = np.append(preds, logits.detach().cpu().numpy(), axis=0)
                if not test:
                    out_label_ids = np.append(out_label_ids, inputs["labels"].detach().cpu().numpy(), axis=0)

        eval_loss = eval_loss / nb_eval_steps
        preds_scores = preds
        preds = np.argmax(preds, axis=1)
        
        output_prefix = 'eval' if not test else 'test'
        if not test:
            acc = simple_accuracy(preds, out_label_ids)
            result = {"eval_acc": acc, "eval_loss": eval_loss}
            results.update(result)

            output_eval_file = os.path.join(eval_output_dir, output_prefix+'_results.txt')

            with open(output_eval_file, "w") as writer:
                logger.info("***** Eval results {} *****".format(str(prefix) + " is test:" + str(test)))
                writer.write("model           =%s\n" % str(args.model_name_or_path))
                writer.write(
                    "total batch size=%d\n"
                    % (
                        args.per_gpu_train_batch_size
                        * args.gradient_accumulation_steps
                    )
                )
                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)
                for key in sorted(result.keys()):
                    logger.info("  %s = %s", key, str(result[key]))
                    writer.write("{:<16}={}\n".format(key, str(result[key])))

        output_preds_file = os.path.join(eval_output_dir, output_prefix+'_prediction_labels.txt')
        with open(output_preds_file, 'w') as writer:
            writer.write('\n'.join([str(x) for x in preds]))

        output_preds_file = os.path.join(eval_output_dir, output_prefix+'_prediction_probs.txt')
        with open(output_preds_file, 'w') as writer:
            writer.write('\n'.join([
                ','.join([str((1+np.tanh(x))/2) for x in scores if x != -float('Inf')])
                for scores in preds_scores]))
    return results


def load_and_cache_examples(args, task, tokenizer, evaluate=False, test=False):
    processor = processors[task](args.encode_type)
    # Load data features from cache or dataset file
    if evaluate:
        cached_mode = "dev"
    elif test:
        cached_mode = "test"
    else:
        cached_mode = "train"
    assert not (evaluate and test)
    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),
        ),
    )
    if os.path.exists(cached_features_file) and not args.overwrite_cache:
        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)
        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)

        logger.info("Training number: %s", str(len(examples)))
        features = convert_examples_to_features(
            examples,
            args.max_choice,
            args.max_seq_length,
            tokenizer,
            pad_token=tokenizer.pad_token_id,
            pad_token_segment_id=tokenizer.pad_token_type_id,
        )
        logger.info("Saving features into cached file %s", cached_features_file)
        torch.save(features, cached_features_file)


    # Convert to Tensors and build dataset
    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)
    all_predicate_positions = torch.tensor([f.predicate_position for f in features], dtype=torch.long)
    all_n_choice = torch.tensor([f.n_choice for f in features], dtype=torch.long)
    if not test:
        all_label_ids = torch.tensor([f.label for f in features], dtype=torch.long)
    else:
        all_label_ids = torch.tensor([-1 for f in features], dtype=torch.long)

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


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

    # Required parameters
    parser.add_argument(
        "--encode_type",
        default=None,
        type=str,
        required=True,
        help="",
    )
    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",
    )
    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.",
    )

    # Other parameters
    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_predict", action="store_true", help="Whether to predict 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=300000, help="Log every X updates steps.")
    parser.add_argument("--save_steps", type=int, default=300000, 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=300, 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("--server_ip", type=str, default="", help="For distant debugging.")
    parser.add_argument("--server_port", type=str, default="", help="For distant debugging.")
    parser.add_argument("--max_choice", type=int, default=10, help="Maximum number of choices (frames).")
    args = parser.parse_args()

    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
            )
        )

    # 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

        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
    device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
    args.n_gpu = 0 if args.no_cuda else torch.cuda.device_count()
    args.device = device

    # Setup logging
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        level=logging.INFO,
    )
    logger.warning(
        "Device: %s, n_gpu: %s, 16-bits training: %s",
        device,
        args.n_gpu,
        args.fp16,
    )

    # Set seed
    set_seed(args)

    # Prepare GLUE task
    if args.task_name not in processors:
        raise ValueError("Task not found: %s" % (args.task_name))

    args.model_type = args.model_type.lower()
    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
    config = config_class.from_pretrained(
        args.config_name if args.config_name else args.model_name_or_path,
        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,
    )

    model.to(args.device)

    logger.info("Training/evaluation parameters %s", args)
    best_steps = 0

    # Training
    if args.do_train:
        train_dataset = load_and_cache_examples(args, args.task_name, tokenizer, evaluate=False)
        global_step, tr_loss, best_steps = train(args, train_dataset, model, tokenizer)
        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()
        # Create output directory if needed
        if not os.path.exists(args.output_dir):
            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()`
        model_to_save = (
            model.module if hasattr(model, "module") else model
        )  # Take care of distributed/parallel training
        model_to_save.save_pretrained(args.output_dir)
        tokenizer.save_pretrained(args.output_dir)
        torch.save(args, os.path.join(args.output_dir, "training_args.bin"))

    # Evaluation
    results = {}
    if args.do_eval:
        checkpoints = [args.output_dir]
        logger.info("Evaluate the following checkpoints: %s", checkpoints)
        for checkpoint in checkpoints:
            global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
            prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""

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

    # Prediction
    if args.do_predict:
        checkpoints = [args.output_dir]
        logger.info("Evaluate the following checkpoints: %s", checkpoints)
        for checkpoint in checkpoints:
            global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
            prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""
            
            print('Loading model from pretrained {}'.format(checkpoint))
            model = model_class.from_pretrained(checkpoint)
            model.to(args.device)
            result = evaluate(args, model, tokenizer, prefix=prefix, test=True)
            result = dict((k + "_{}".format(global_step), v) for k, v in result.items())
            results.update(result)

    return results


if __name__ == "__main__":
    main()