main_linprobe.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
# MoCo v3: https://github.com/facebookresearch/moco-v3
# --------------------------------------------------------

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

import paddle
import paddle.nn as nn
import paddle.vision.transforms as transforms
import paddle.vision.datasets as datasets
from paddle.fluid.optimizer import LarsMomentumOptimizer
from paddle.io import BatchSampler, DataLoader, DistributedBatchSampler
from paddle.nn.initializer import TruncatedNormal

import util.misc as misc
from util.pos_embed import interpolate_pos_embed
from util.misc import NativeScalerWithGradNormCount as NativeScaler
from util.misc import WandbLogger
from util.crop import RandomResizedCrop

import models_vit

from engine_finetune import train_one_epoch, evaluate


def get_args_parser():
    parser = argparse.ArgumentParser('MAE linear probing for image classification', add_help=False)
    parser.add_argument('--batch_size', default=512, type=int,
                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')
    parser.add_argument('--epochs', default=90, 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')

    # Optimizer parameters
    parser.add_argument('--weight_decay', type=float, default=0,
                        help='weight decay (default: 0 for linear probe following MoCo v1)')

    parser.add_argument('--lr', type=float, default=None, metavar='LR',
                        help='learning rate (absolute lr)')
    parser.add_argument('--blr', type=float, default=0.1, metavar='LR',
                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')

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

    parser.add_argument('--warmup_epochs', type=int, default=10, metavar='N',
                        help='epochs to warmup LR')

    # * Finetuning params
    parser.add_argument('--finetune', default='',
                        help='finetune from checkpoint')
    parser.add_argument('--global_pool', action='store_true')
    parser.set_defaults(global_pool=False)
    parser.add_argument('--cls_token', action='store_false', dest='global_pool',
                        help='Use class token instead of global pool for classification')

    # Dataset parameters
    parser.add_argument('--data_path', default='/datasets01/imagenet_full_size/061417/', type=str,
                        help='dataset path')
    parser.add_argument('--nb_classes', default=1000, type=int,
                        help='number of the classification types')

    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_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('--num_workers', default=10, type=int)

    # logger training parameters
    parser.add_argument('--log_wandb', action='store_true',
                        help='log training and validation metrics to wandb')
    parser.add_argument('--wandb_entity', default=None, type=str,
                        help='user or team name of wandb')
    parser.add_argument('--wandb_project', default=None, type=str,
                        help='log training and validation metrics to wandb')
    parser.add_argument('--debug', action='store_true')

    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'))

    # fix the seed for reproducibility
    seed = args.seed + misc.get_rank()
    paddle.seed(args.seed)
    np.random.seed(seed)
    random.seed(seed)
    if args.debug:
        paddle.version.cudnn.FLAGS_cudnn_deterministic = True

    # linear probe: weak augmentation
    transform_train = transforms.Compose([
            RandomResizedCrop(224, interpolation='bicubic'),
            transforms.RandomHorizontalFlip(),
            transforms.ToTensor(),
            transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
    transform_val = transforms.Compose([
            transforms.Resize(256, interpolation='bicubic'),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])
    dataset_train = datasets.DatasetFolder(os.path.join(args.data_path, 'train' if not args.debug else 'val'), transform=transform_train)
    dataset_val = datasets.DatasetFolder(os.path.join(args.data_path, 'val'), transform=transform_val)
    print(dataset_train)
    print(dataset_val)

    num_tasks = misc.get_world_size()
    global_rank = misc.get_rank()
    sampler_train = DistributedBatchSampler(
        dataset_train, args.batch_size, shuffle=True, drop_last=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 = DistributedBatchSampler(
            dataset_val, args.batch_size, shuffle=True, drop_last=False)  # shuffle=True to reduce monitor bias
    else:
        sampler_val = BatchSampler(dataset=dataset_val, batch_size=args.batch_size)

    if global_rank == 0 and args.log_wandb and not args.eval:
        log_writer = WandbLogger(args, entity=args.wandb_entity, project=args.wandb_project)
    else:
        log_writer = None

    data_loader_train = DataLoader(dataset_train, batch_sampler=sampler_train, num_workers=args.num_workers)
    data_loader_val = DataLoader(dataset_val, batch_sampler=sampler_val, num_workers=args.num_workers)

    model = models_vit.__dict__[args.model](
        num_classes=args.nb_classes,
        global_pool=args.global_pool,
    )

    if args.finetune and not args.eval:
        checkpoint = paddle.load(args.finetune)

        print("Load pre-trained checkpoint from: %s" % args.finetune)
        checkpoint_model = checkpoint['model']
        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)

        # load pre-trained model
        model.set_state_dict(checkpoint_model)

        # manually initialize fc layer: following MoCo v3
        TruncatedNormal(std=0.01)(model.head.weight)

    # for linear prob only
    # hack: revise model's head with BN
    model.head = nn.Sequential(
        nn.BatchNorm1D(model.head.weight.shape[0], epsilon=1e-6, weight_attr=False, bias_attr=False),
        model.head)
    # freeze all but the head
    for _, p in model.named_parameters():
        p.stop_gradient = True
    for _, p in model.head[1].named_parameters():
        p.stop_gradient = False

    model_without_ddp = model
    n_parameters = sum(p.numel().item() for p in model.parameters() if not p.stop_gradient)

    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)

    model = paddle.DataParallel(model)
    model_without_ddp = model._layers

    optimizer = LarsMomentumOptimizer(
        learning_rate=args.lr,
        momentum=0.9,
        lars_weight_decay=args.weight_decay,
        parameter_list=model_without_ddp.head[1].parameters()
    )
    print(optimizer)
    loss_scaler = NativeScaler()

    criterion = nn.CrossEntropyLoss()

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

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

    if args.eval:
        test_stats = evaluate(data_loader_val, model)
        print(f"Accuracy of the network on the {len(dataset_val)} test images: {test_stats['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):
        data_loader_train.batch_sampler.set_epoch(epoch)
        if log_writer is not None:
            num_training_steps_per_epoch = len(dataset_train) // eff_batch_size
            log_writer.set_step(epoch * num_training_steps_per_epoch)
        train_stats = train_one_epoch(
            model, criterion, data_loader_train,
            optimizer, epoch, loss_scaler,
            log_writer=log_writer,
            args=args
        )
        test_stats = evaluate(data_loader_val, model)

        if args.output_dir:
            misc.save_model(
                args=args, model_without_ddp=model_without_ddp, optimizer=optimizer,
                loss_scaler=loss_scaler, epoch=epoch, tag='latest')
            if test_stats["acc1"] > max_accuracy:
                misc.save_model(
                    args=args, model_without_ddp=model_without_ddp, optimizer=optimizer,
                    loss_scaler=loss_scaler, epoch=epoch, tag='best')
            if (epoch + 1) % 20 == 0 or epoch + 1 == args.epochs:
                misc.save_model(
                    args=args, model_without_ddp=model_without_ddp, optimizer=optimizer,
                    loss_scaler=loss_scaler, epoch=epoch)

        print(f"Accuracy of the network on the {len(dataset_val)} test images: {test_stats['acc1']:.1f}%")
        max_accuracy = max(max_accuracy, test_stats["acc1"])
        print(f'Max accuracy: {max_accuracy:.2f}%')

        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.update(log_stats)
                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)