cae_beit-base-p16_8xb256-amp-coslr-300e_in1k.py

_base_ = '../_base_/default_runtime.py'

# dataset settings
dataset_type = 'ImageNet'
data_root = 'data/imagenet/'
data_preprocessor = dict(
    type='TwoNormDataPreprocessor',
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    second_mean=[-31.875, -31.875, -31.875],
    second_std=[318.75, 318.75, 318.75],
    to_rgb=True)

train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='RandomFlip', prob=0.5),
    dict(
        type='RandomResizedCropAndInterpolationWithTwoPic',
        size=224,
        second_size=112,
        interpolation='bicubic',
        second_interpolation='lanczos',
        scale=(0.08, 1.0)),
    dict(
        type='BEiTMaskGenerator',
        input_size=(14, 14),
        num_masking_patches=75,
        max_num_patches=None,
        min_num_patches=16),
    dict(type='PackInputs')
]

train_dataloader = dict(
    batch_size=256,
    num_workers=8,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    collate_fn=dict(type='default_collate'),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        ann_file='meta/train.txt',
        data_prefix=dict(img_path='train/'),
        pipeline=train_pipeline))

# model settings
model = dict(
    type='CAE',
    backbone=dict(
        type='CAEPretrainViT',
        arch='b',
        patch_size=16,
        layer_scale_init_value=0.1,
        bias='qv_bias'),
    neck=dict(
        type='CAENeck',
        embed_dims=768,
        num_heads=12,
        regressor_depth=4,
        decoder_depth=4,
        mlp_ratio=4,
        layer_scale_init_value=0.1,
    ),
    head=dict(type='CAEHead', loss=dict(type='CAELoss', lambd=2)),
    target_generator=dict(
        type='DALL-E',
        init_cfg=dict(
            type='Pretrained',
            checkpoint=  # noqa: E251
            'https://download.openmmlab.com/mmselfsup/1.x/target_generator_ckpt/dalle_encoder.pth',  # noqa: E501
        )),
    base_momentum=0.0)

# optimizer wrapper
optim_wrapper = dict(
    type='AmpOptimWrapper',
    loss_scale='dynamic',
    optimizer=dict(
        type='AdamW', lr=1.5e-3, betas=(0.9, 0.999), weight_decay=0.05),
    clip_grad=dict(max_norm=3.0),
    paramwise_cfg=dict(
        bias_decay_mult=0.0, norm_decay_mult=0.0, flat_decay_mult=0.0))

# learning rate scheduler
param_scheduler = [
    dict(
        type='LinearLR',
        start_factor=1e-4,
        by_epoch=True,
        begin=0,
        end=10,
        convert_to_iter_based=True),
    dict(
        type='CosineAnnealingLR',
        T_max=290,
        eta_min=1e-5,
        by_epoch=True,
        begin=10,
        end=300,
        convert_to_iter_based=True)
]

# runtime settings
train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=300)
default_hooks = dict(
    # only keeps the latest 3 checkpoints
    checkpoint=dict(type='CheckpointHook', interval=10, max_keep_ckpts=3))

randomness = dict(seed=0, diff_rank_seed=True)

find_unused_parameters = True

# NOTE: `auto_scale_lr` is for automatically scaling LR
# based on the actual training batch size.
auto_scale_lr = dict(base_batch_size=2048)