main_mvlt_finetune.py

# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.

# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
# --------------------------------------------------------
# References:
# DeiT: https://github.com/facebookresearch/deit
# BEiT: https://github.com/microsoft/unilm/tree/master/beit
# --------------------------------------------------------

import argparse
import datetime
import json
import numpy as np
import os
import time
from pathlib import Path
import string

import torch
import torch.backends.cudnn as cudnn
from torch.utils.tensorboard import SummaryWriter

import timm

assert timm.__version__ == "0.3.2" # version check
from timm.models.layers import trunc_normal_
from timm.data.mixup import Mixup
from timm.loss import LabelSmoothingCrossEntropy, SoftTargetCrossEntropy

import util.lr_decay as lrd
import util.misc as misc
from util.datasets import build_dataset
from util.pos_embed import interpolate_pos_embed
from util.misc import NativeScalerWithGradNormCount as NativeScaler

import models_mvlt_finetune

from engine_mvlt_finetune import train, evaluate, test
from dataset import Batch_Balanced_Dataset, hierarchical_dataset, AlignCollate


def get_args_parser():
    parser = argparse.ArgumentParser(
        'MAE fine-tuning for image classification', add_help=False)
    parser.add_argument('--batch_size', default=64, type=int,
                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')
    parser.add_argument('--num_iter', default=20000, type=int,
                        help='number of iterations to train for')
    parser.add_argument('--accum_iter', default=1, type=int,
                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')

    # Model parameters
    parser.add_argument('--model', default='vit_large_patch16', type=str, metavar='MODEL',
                        help='Name of model to train')

    parser.add_argument('--imgH', default=112, type=int,
                        help='the height of the input image')
    parser.add_argument('--imgW', default=448, type=int,
                        help='the width of the input image')
    parser.add_argument('--iter_correct', default=3, type=int,
                        help='correction times')

    parser.add_argument('--drop_path', type=float, default=0.1, metavar='PCT',
                        help='Drop path rate (default: 0.1)')

    # Optimizer parameters
    parser.add_argument('--clip_grad', type=float, default=None, metavar='NORM',
                        help='Clip gradient norm (default: None, no clipping)')
    parser.add_argument('--weight_decay', type=float, default=0.05,
                        help='weight decay (default: 0.05)')

    parser.add_argument('--lr', type=float, default=None, metavar='LR',
                        help='learning rate (absolute lr)')
    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',
                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')
    parser.add_argument('--layer_decay', type=float, default=0.75,
                        help='layer-wise lr decay from ELECTRA/BEiT')

    parser.add_argument('--min_lr', type=float, default=1e-6, metavar='LR',
                        help='lower lr bound for cyclic schedulers that hit 0')

    parser.add_argument('--warmup_iters', type=int, default=8000, metavar='N',
                        help='iters to warmup LR')

    # * Finetuning params
    parser.add_argument('--finetune', default='',
                        help='finetune from checkpoint')

    # Dataset parameters
    parser.add_argument('--data_path', default='/datasets01/imagenet_full_size/061417/', type=str,
                        help='dataset path')
    parser.add_argument('--data_path_val', default='/datasets01/imagenet_full_size/061417/', type=str,
                        help='dataset path')
    parser.add_argument('--select_data', default='MJ-ST', type=str,
                        help='select training data (default is MJ-ST, which means MJ and ST used as training data)')
    parser.add_argument('--batch_ratio', default='0.5-0.5', type=str,
                        help='assign ratio for each selected data in the batch')
    parser.add_argument('--label_max_length', default=25, type=int,
                        help='maximum-label-length')
    parser.add_argument('--character', default='0123456789abcdefghijklmnopqrstuvwxyz', type=str,
                        help='character label')
    parser.add_argument('--random_crop', action='store_true',
                        help='Image random_crop')
    parser.add_argument('--abinet_augment', action='store_true',
                        help='abinet augment')
    parser.add_argument('--sensitive', action='store_true',
                        help='for sensitive character mode')
    parser.add_argument('--data_filtering_off', action='store_true',
                        help='for data_filtering_off mode')
    parser.add_argument('--val_interval', default=250, type=int, 
                        help='validation interval')

    parser.add_argument('--output_dir', default='./output_dir',
                        help='path where to save, empty for no saving')
    parser.add_argument('--log_dir', default='./output_dir',
                        help='path where to tensorboard log')
    parser.add_argument('--device', default='cuda',
                        help='device to use for training / testing')
    parser.add_argument('--seed', default=0, type=int)
    parser.add_argument('--resume', default='',
                        help='resume from checkpoint')

    parser.add_argument('--start_iter', default=0, type=int, metavar='N',
                        help='start iter')
    parser.add_argument('--test', action='store_true',
                        help='Perform evaluation only')
    parser.add_argument('--dist_eval', action='store_true', default=False,
                        help='Enabling distributed evaluation (recommended during training for faster monitor')
    parser.add_argument('--num_workers', default=10, type=int)
    parser.add_argument('--pin_mem', action='store_true',
                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')
    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')
    parser.set_defaults(pin_mem=True)

    # distributed training parameters
    parser.add_argument('--world_size', default=1, type=int,
                        help='number of distributed processes')
    parser.add_argument('--local_rank', default=-1, type=int)
    parser.add_argument('--dist_on_itp', action='store_true')
    parser.add_argument('--dist_url', default='env://',
                        help='url used to set up distributed training')

    return parser


def main(args):
    misc.init_distributed_mode(args)

    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))
    print("{}".format(args).replace(', ', ',\n'))

    device = torch.device(args.device)

    # fix the seed for reproducibility
    seed = args.seed + misc.get_rank()
    torch.manual_seed(seed)
    np.random.seed(seed)

    cudnn.benchmark = True

    if args.sensitive:
        args.character = string.printable[:-38]

    num_tasks = misc.get_world_size()
    global_rank = misc.get_rank()

    args.eval = False
    args.pretrain = False
    data_loader_train = Batch_Balanced_Dataset(args, num_tasks, global_rank)

    args.eval = True
    dataset_val, dataset_val_log = hierarchical_dataset(
        root=args.data_path_val, opt=args, data_filtering_off=True, global_rank=global_rank)

    if True:  # args.distributed:
        if args.dist_eval:
            if len(dataset_val) % num_tasks != 0:
                print('Warning: Enabling distributed evaluation with an eval dataset not divisible by process number. '
                      'This will slightly alter validation results as extra duplicate entries are added to achieve '
                      'equal num of samples per-process.')
            sampler_val = torch.utils.data.DistributedSampler(
                dataset_val, num_replicas=num_tasks, rank=global_rank, shuffle=True)  # shuffle=True to reduce monitor bias
        else:
            sampler_val = torch.utils.data.SequentialSampler(dataset_val)
    else:
        sampler_val = torch.utils.data.SequentialSampler(dataset_val)

    if global_rank == 0 and args.log_dir is not None and not args.test:
        os.makedirs(args.log_dir, exist_ok=True)
        log_writer = SummaryWriter(log_dir=args.log_dir)
        log = open(f'{args.output_dir}/log_dataset.txt', 'a')
        log.write(dataset_val_log)
        print('-' * 80)
        log.write('-' * 80 + '\n')
        log.close()
    else:
        log_writer = None

    collate_val = AlignCollate(
        imgH=args.imgH, imgW=args.imgW, opt=args, is_eval=True)
    data_loader_val = torch.utils.data.DataLoader(
        dataset_val, sampler=sampler_val,
        batch_size=args.batch_size,
        num_workers=args.num_workers,
        collate_fn=collate_val,
        pin_memory=args.pin_mem,
        drop_last=False
    )

    model = models_mvlt_finetune.__dict__[args.model](
        img_size=(args.imgH, args.imgW),
        num_char_class=len(args.character) + 2,
        txt_max_len=args.label_max_length + 2,
        iter_correct=args.iter_correct,
        drop_path_rate=args.drop_path,
    )

    if args.finetune and not args.test:
        checkpoint = torch.load(args.finetune, map_location='cpu')

        print("Load pre-trained checkpoint from: %s" % args.finetune)
        checkpoint_model = checkpoint['model']
        state_dict = model.state_dict()
        for k in ['mask_token', 'decoder_pred.bias', 'decoder_pred.weight']:
            if k in checkpoint_model:
                print(f"Removing key {k} from pretrained checkpoint")
                del checkpoint_model[k]

        # interpolate position embedding
        #interpolate_pos_embed(model, checkpoint_model)

        # load pre-trained model
        msg = model.load_state_dict(checkpoint_model, strict=False)
        print(msg)

    model.to(device)

    model_without_ddp = model
    n_parameters = sum(p.numel()
                       for p in model.parameters() if p.requires_grad)

    print("Model = %s" % str(model_without_ddp))
    print('number of params (M): %.2f' % (n_parameters / 1.e6))

    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()

    if args.lr is None:  # only base_lr is specified
        args.lr = args.blr * eff_batch_size / 256

    print("base lr: %.2e" % (args.lr * 256 / eff_batch_size))
    print("actual lr: %.2e" % args.lr)

    print("accumulate grad iterations: %d" % args.accum_iter)
    print("effective batch size: %d" % eff_batch_size)

    if args.distributed:
        model = torch.nn.parallel.DistributedDataParallel(
            model, device_ids=[args.gpu], find_unused_parameters=True)
        model_without_ddp = model.module

    # build optimizer with layer-wise lr decay (lrd)
    param_groups = lrd.param_groups_lrd(model_without_ddp, args.weight_decay,
                                        no_weight_decay_list=model_without_ddp.no_weight_decay(),
                                        layer_decay=args.layer_decay
                                        )
    optimizer = torch.optim.AdamW(param_groups, lr=args.lr)
    loss_scaler = NativeScaler()

    criterion = torch.nn.CrossEntropyLoss(ignore_index=0)

    print("criterion = %s" % str(criterion))

    misc.load_model(args=args, model_without_ddp=model_without_ddp,
                    optimizer=optimizer, loss_scaler=loss_scaler)

    if args.test:
        test(data_loader_val, model, device, args, global_rank)
        exit(0)

    print(f"Start training for {args.num_iter} iter")
    start_time = time.time()
    max_accuracy = 0.0

    train_stats = train(
        model, criterion, data_loader_train, data_loader_val,
        optimizer, device, args.num_iter, loss_scaler,
        args.clip_grad,
        log_writer=log_writer,
        args=args,
        is_main_process=misc.is_main_process()
    )

    log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},
                 'n_parameters': n_parameters}

    if args.output_dir and misc.is_main_process():
        if log_writer is not None:
            log_writer.flush()
        with open(os.path.join(args.output_dir, "log.txt"), mode="a", encoding="utf-8") as f:
            f.write(json.dumps(log_stats) + "\n")

    total_time = time.time() - start_time
    total_time_str = str(datetime.timedelta(seconds=int(total_time)))
    print('Training time {}'.format(total_time_str))


if __name__ == '__main__':
    args = get_args_parser()
    args = args.parse_args()
    if args.output_dir:
        Path(args.output_dir).mkdir(parents=True, exist_ok=True)
    main(args)