main_finetune.py

# ------------------------------------------------------------------------
# Modified from MAE (https://github.com/facebookresearch/mae)
# Copyright (c) Meta Platforms, Inc. and affiliates. All Rights Reserved.
# Licensed under the CC-BY-NC 4.0 license.
# ------------------------------------------------------------------------
# Modifications Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
# SPDX-License-Identifier: Apache-2.0

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

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

from timm.models.layers import trunc_normal_
from timm.data.mixup import Mixup
from timm.loss import LabelSmoothingCrossEntropy, SoftTargetCrossEntropy
from timm.utils import ModelEma

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

from models import vit as models_vit

from engine_finetune import train_one_epoch, evaluate


def get_args_parser():
    parser = argparse.ArgumentParser('ViT supervised fine-tuning', 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('--epochs', default=50, type=int)
    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('--input_size', default=224, type=int,
                        help='images input size')

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

    parser.add_argument('--model_ema', action='store_true', default=False)
    parser.add_argument('--model_ema_decay', type=float, default=0.9999, help='')
    parser.add_argument('--model_ema_force_cpu', action='store_true', default=False, help='')

    # Optimizer parameters
    parser.add_argument('--opt_betas', default=None, type=float, nargs='+', metavar='BETA',
                        help='Optimizer Betas (default: None, use opt default)')
    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_epochs', type=int, default=5, metavar='N',
                        help='epochs to warmup LR')

    # Augmentation parameters
    parser.add_argument('--color_jitter', type=float, default=None, metavar='PCT',
                        help='Color jitter factor (enabled only when not using Auto/RandAug)')
    parser.add_argument('--aa', type=str, default='rand-m9-mstd0.5-inc1', metavar='NAME',
                        help='Use AutoAugment policy. "v0" or "original". " + "(default: rand-m9-mstd0.5-inc1)'),
    parser.add_argument('--smoothing', type=float, default=0.1,
                        help='Label smoothing (default: 0.1)')

    # * Random Erase params
    parser.add_argument('--reprob', type=float, default=0.25, metavar='PCT',
                        help='Random erase prob (default: 0.25)')
    parser.add_argument('--remode', type=str, default='pixel',
                        help='Random erase mode (default: "pixel")')
    parser.add_argument('--recount', type=int, default=1,
                        help='Random erase count (default: 1)')
    parser.add_argument('--resplit', action='store_true', default=False,
                        help='Do not random erase first (clean) augmentation split')

    # * Mixup params
    parser.add_argument('--mixup', type=float, default=0.8,
                        help='mixup alpha, mixup enabled if > 0.')
    parser.add_argument('--cutmix', type=float, default=1.0,
                        help='cutmix alpha, cutmix enabled if > 0.')
    parser.add_argument('--cutmix_minmax', type=float, nargs='+', default=None,
                        help='cutmix min/max ratio, overrides alpha and enables cutmix if set (default: None)')
    parser.add_argument('--mixup_prob', type=float, default=1.0,
                        help='Probability of performing mixup or cutmix when either/both is enabled')
    parser.add_argument('--mixup_switch_prob', type=float, default=0.5,
                        help='Probability of switching to cutmix when both mixup and cutmix enabled')
    parser.add_argument('--mixup_mode', type=str, default='batch',
                        help='How to apply mixup/cutmix params. Per "batch", "pair", or "elem"')

    # * Finetuning params
    parser.add_argument('--finetune', default='', help='finetune from MAE pretrained checkpoint')
    parser.add_argument('--super_finetune', default='', help='finetune from supervised pretrained checkpoint')
    parser.add_argument('--init_scale', default=0.001, type=float)
    parser.add_argument('--global_pool', action='store_true')
    parser.set_defaults(global_pool=True)
    parser.add_argument('--cls_token', action='store_false', dest='global_pool',
                        help='Use class token instead of global pool for classification')
    parser.add_argument('--use_fixed_pos_emb', action='store_true', help='Use sincos position embedding')
    parser.add_argument('--head_lr_scalar', default=1.0, type=float)

    # Dataset parameters
    parser.add_argument('--data_path', default='~/data/imagenet/', type=str,
                        help='dataset path')
    parser.add_argument('--nb_classes', default=1000, type=int,
                        help='number of the classification types')
    parser.add_argument('--trainindex', default=None, type=str, metavar='PATH',
                        help='path to train annotation index (default: None)')
    parser.add_argument('--anno_percent', type=float, default=1.0, help='number of labeled data')

    parser.add_argument('--output_dir', default='./save',
                        help='path where to save, empty for no saving')
    parser.add_argument('--save_ckpt_freq', default=50, type=int)
    parser.add_argument('--log_dir', default=None,
                        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_epoch', default=0, type=int, metavar='N',
                        help='start epoch')
    parser.add_argument('--eval', 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('--eval_freq', default=1, type=int)
    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')

    parser.add_argument('--print_freq', default=40, type=int, help='print frequency')

    return parser


def main(args):
    misc.init_distributed_mode(args)
    args.print_freq = round(args.print_freq / (args.batch_size * misc.get_world_size() / 1024.0))

    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

    dataset_train = build_dataset(is_train=True, args=args)
    dataset_val = build_dataset(is_train=False, args=args)

    if True:  # args.distributed:
        num_tasks = misc.get_world_size()
        global_rank = misc.get_rank()
        sampler_train = torch.utils.data.DistributedSampler(
            dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True
        )
        print("Sampler_train = %s" % str(sampler_train))
        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_train = torch.utils.data.RandomSampler(dataset_train)
        sampler_val = torch.utils.data.SequentialSampler(dataset_val)

    if global_rank == 0 and args.log_dir is not None and not args.eval:
        os.makedirs(args.log_dir, exist_ok=True)
        log_writer = SummaryWriter(log_dir=args.log_dir)
    else:
        log_writer = None

    data_loader_train = torch.utils.data.DataLoader(
        dataset_train, sampler=sampler_train,
        batch_size=args.batch_size,
        num_workers=args.num_workers,
        pin_memory=args.pin_mem,
        drop_last=True,
    )

    data_loader_val = torch.utils.data.DataLoader(
        dataset_val, sampler=sampler_val,
        batch_size=int(1.5 * args.batch_size),
        num_workers=args.num_workers,
        pin_memory=args.pin_mem,
        drop_last=False
    )

    mixup_fn = None
    mixup_active = args.mixup > 0 or args.cutmix > 0. or args.cutmix_minmax is not None
    if mixup_active:
        print("Mixup is activated!")
        mixup_fn = Mixup(
            mixup_alpha=args.mixup, cutmix_alpha=args.cutmix, cutmix_minmax=args.cutmix_minmax,
            prob=args.mixup_prob, switch_prob=args.mixup_switch_prob, mode=args.mixup_mode,
            label_smoothing=args.smoothing, num_classes=args.nb_classes)
    
    model = models_vit.__dict__[args.model](
        num_classes=args.nb_classes,
        drop_path_rate=args.drop_path,
        global_pool=args.global_pool,
        use_fixed_pos_emb=args.use_fixed_pos_emb,
        init_scale=args.init_scale,
    )

    if args.finetune and not args.eval:
        checkpoint = torch.load(args.finetune, map_location='cpu')
        print("Load pre-trained checkpoint from: %s" % args.finetune)
        checkpoint_model = checkpoint['model'] if 'model' in checkpoint else checkpoint
        state_dict = model.state_dict()
        for k in ['head.weight', 'head.bias']:
            if k in checkpoint_model and checkpoint_model[k].shape != state_dict[k].shape:
                print(f"Removing key {k} from pretrained checkpoint")
                del checkpoint_model[k]

        # interpolate position embedding
        interpolate_pos_embed(model, checkpoint_model)

        # interpolate patch_embed
        new_size = model.patch_embed.proj.weight.shape[-1]
        orig_size = checkpoint_model['patch_embed.proj.weight'].shape[-1]
        if new_size != orig_size:
            print("patch_embed.proj interpolate from %dx%d to %dx%d" % (orig_size, orig_size, new_size, new_size))
            patch_embed_weight = torch.nn.functional.interpolate(
                checkpoint_model['patch_embed.proj.weight'], size=(new_size, new_size), mode='bicubic', align_corners=False)
            checkpoint_model['patch_embed.proj.weight'] = patch_embed_weight

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

        if args.global_pool:
            assert set(msg.missing_keys) == {'head.weight', 'head.bias', 'fc_norm.weight', 'fc_norm.bias'}
        else:
            assert set(msg.missing_keys) == {'head.weight', 'head.bias'}

        # manually initialize fc layer
        trunc_normal_(model.head.weight, std=2e-5)
    elif args.super_finetune:
        checkpoint = torch.load(args.super_finetune, map_location='cpu')
        print("Load supervised pre-trained checkpoint from: %s" % args.super_finetune)
        checkpoint_model = checkpoint['model']
        # load pre-trained model
        checkpoint_model_num_cls = checkpoint_model['head.weight'].shape[0]
        if checkpoint_model_num_cls != args.nb_classes:
            # if num_cls don't match, remove the last layer
            del checkpoint_model['head.weight']
            del checkpoint_model['head.bias']
            msg = model.load_state_dict(checkpoint_model, strict=False)
            assert set(msg.missing_keys) == {"head.weight", "head.bias"}, \
                "missing keys:\n{}".format('\n'.join(msg.missing_keys))
        else:
            msg = model.load_state_dict(checkpoint_model, strict=True)
        print(msg)

    model.to(device)

    model_ema = None
    if args.model_ema:
        # Important to create EMA model after cuda(), DP wrapper, and AMP but before SyncBN and DDP wrapper
        model_ema = ModelEma(
            model,
            decay=args.model_ema_decay,
            device='cpu' if args.model_ema_force_cpu else '',
            resume='')
        print("Using EMA with decay = %.8f" % args.model_ema_decay)

    model_without_ddp = model
    n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad)
    n_parameters_tensors = sum(1 for p in model.parameters() if p.requires_grad)

    print("Model = %s" % str(model_without_ddp))
    print('number of params (M): %f' % (n_parameters / 1.e6))
    print('number of learnable param tensors: %d/%d' % (n_parameters_tensors, len(list(model.parameters()))))

    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])
        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, head_scalar=args.head_lr_scalar,
    )
    opt_args = dict(lr=args.lr)
    if hasattr(args, 'opt_eps') and args.opt_eps is not None:
        opt_args['eps'] = args.opt_eps
    if hasattr(args, 'opt_betas') and args.opt_betas is not None:
        opt_args['betas'] = args.opt_betas
    optimizer = torch.optim.AdamW(param_groups, **opt_args)
    print(f"optimizer: {optimizer}")
    loss_scaler = NativeScaler()

    if mixup_fn is not None:
        # smoothing is handled with mixup label transform
        criterion = SoftTargetCrossEntropy()
    elif args.smoothing > 0.:
        criterion = LabelSmoothingCrossEntropy(smoothing=args.smoothing)
    else:
        criterion = torch.nn.CrossEntropyLoss()

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

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

    if args.eval:
        test_stats = evaluate(data_loader_val, model, device)
        print(f"Accuracy of the network on the {len(dataset_val)} test images: {test_stats['acc1']:.1f}%")
        if model_ema is not None:
            test_stats_ema = evaluate(data_loader_val, model_ema.ema, device)
            print(f"Accuracy of the EMA network on the {len(dataset_val)} test images: {test_stats_ema['acc1']:.1f}%")
        exit(0)

    print(f"Start training for {args.epochs} epochs")
    start_time = time.time()
    max_accuracy = 0.0
    for epoch in range(args.start_epoch, args.epochs):
        if args.distributed:
            data_loader_train.sampler.set_epoch(epoch)
        train_stats = train_one_epoch(
            model, criterion, data_loader_train,
            optimizer, device, epoch, loss_scaler,
            args.clip_grad, model_ema, mixup_fn,
            log_writer=log_writer, args=args
        )
        if args.output_dir:
            save_names = ['checkpoint.pth']
            if (epoch + 1) % args.save_ckpt_freq == 0 or epoch + 1 == args.epochs:
                save_names.append('checkpoint-%s.pth' % str(epoch))
            misc.save_model(
                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,
                loss_scaler=loss_scaler, epoch=epoch, save_names=save_names, model_ema=model_ema)

        if (epoch + 1) % args.eval_freq == 0:
            test_stats = evaluate(data_loader_val, model, device)
            print(f"Accuracy of the network on the {len(dataset_val)} test images: {test_stats['acc1']:.1f}%")
            if max_accuracy < test_stats["acc1"]:
                max_accuracy = test_stats["acc1"]
                if args.output_dir:
                    save_names = ['checkpoint-best.pth']
                    misc.save_model(
                        args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,
                        loss_scaler=loss_scaler, epoch=epoch, save_names=save_names, model_ema=model_ema)
            print(f'Max accuracy: {max_accuracy:.2f}%')

            if model_ema is not None:
                test_stats_ema = evaluate(data_loader_val, model_ema.ema, device)
                print(f"Accuracy of the EMA network on the {len(dataset_val)} test images: {test_stats_ema['acc1']:.1f}%")
                """if epoch >= 205:
                    print("copying from EMA model")
                    copy_model_from_ema(model, model_ema)
                    test_stats_after = evaluate(data_loader_val, model, device)
                    print(f"Accuracy of the network after copying from EMA model: {test_stats_after['acc1']:.1f}%")"""

            if log_writer is not None:
                log_writer.add_scalar('perf/test_acc1', test_stats['acc1'], epoch)
                log_writer.add_scalar('perf/test_acc5', test_stats['acc5'], epoch)
                log_writer.add_scalar('perf/test_loss', test_stats['loss'], epoch)

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