add mmpretrain/ part

mmpretrain/configs/mobilenet_v2/mobilenet_v2_8xb32_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs32_pil_resize import *
+    from .._base_.default_runtime import *
+    from .._base_.models.mobilenet_v2_1x import *
+    from .._base_.schedules.imagenet_bs256_epochstep import *

mmpretrain/configs/mobilenet_v3/mobilenet_v3_large_8xb128_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+
+# Refers to https://pytorch.org/blog/ml-models-torchvision-v0.9/#classification
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.models.mobilenet_v3_small import *
+    from .._base_.datasets.imagenet_bs128_mbv3 import *
+    from .._base_.default_runtime import *
+
+from mmengine.optim import StepLR
+from torch.optim import RMSprop
+
+# model settings
+model.merge(
+    dict(
+        backbone=dict(arch='large'),
+        head=dict(in_channels=960, mid_channels=[1280]),
+    ))
+# schedule settings
+optim_wrapper = dict(
+    optimizer=dict(
+        type=RMSprop,
+        lr=0.064,
+        alpha=0.9,
+        momentum=0.9,
+        eps=0.0316,
+        weight_decay=1e-5))
+
+param_scheduler = dict(type=StepLR, by_epoch=True, step_size=2, gamma=0.973)
+
+train_cfg = dict(by_epoch=True, max_epochs=600, val_interval=1)
+val_cfg = dict()
+test_cfg = dict()
+
+# NOTE: `auto_scale_lr` is for automatically scaling LR
+# based on the actual training batch size.
+# base_batch_size = (8 GPUs) x (128 samples per GPU)
+auto_scale_lr = dict(base_batch_size=1024)

mmpretrain/configs/mobilenet_v3/mobilenet_v3_small_050_8xb128_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+# Refers to https://pytorch.org/blog/ml-models-torchvision-v0.9/#classification
+
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.models.mobilenet_v3_small import *
+    from .._base_.datasets.imagenet_bs128_mbv3 import *
+    from .._base_.default_runtime import *
+
+from mmengine.optim import StepLR
+from torch.nn.modules.batchnorm import BatchNorm2d
+from torch.optim import RMSprop
+
+# model settings
+model.merge(
+    dict(
+        backbone=dict(
+            arch='small_050',
+            norm_cfg=dict(type=BatchNorm2d, eps=1e-5, momentum=0.1)),
+        head=dict(in_channels=288),
+    ))
+
+train_pipeline = [
+    dict(type=LoadImageFromFile),
+    dict(
+        type=RandomResizedCrop,
+        scale=224,
+        backend='pillow',
+        interpolation='bicubic'),
+    dict(type=RandomFlip, prob=0.5, direction='horizontal'),
+    dict(
+        type=AutoAugment,
+        policies='imagenet',
+        hparams=dict(pad_val=[round(x) for x in [103.53, 116.28, 123.675]])),
+    dict(
+        type=RandomErasing,
+        erase_prob=0.2,
+        mode='rand',
+        min_area_ratio=0.02,
+        max_area_ratio=1 / 3,
+        fill_color=[103.53, 116.28, 123.675],
+        fill_std=[57.375, 57.12, 58.395]),
+    dict(type=PackInputs),
+]
+
+test_pipeline = [
+    dict(type=LoadImageFromFile),
+    dict(
+        type=ResizeEdge,
+        scale=256,
+        edge='short',
+        backend='pillow',
+        interpolation='bicubic'),
+    dict(type=CenterCrop, crop_size=224),
+    dict(type=PackInputs),
+]
+
+train_dataloader.merge(dict(dataset=dict(pipeline=train_pipeline)))
+
+val_dataloader.merge(dict(dataset=dict(pipeline=test_pipeline)))
+# If you want standard test, please manually configure the test dataset
+test_dataloader = val_dataloader
+
+# schedule settings
+optim_wrapper = dict(
+    optimizer=dict(
+        type=RMSprop,
+        lr=0.064,
+        alpha=0.9,
+        momentum=0.9,
+        eps=0.0316,
+        weight_decay=1e-5))
+
+param_scheduler = dict(type=StepLR, by_epoch=True, step_size=2, gamma=0.973)
+
+train_cfg = dict(by_epoch=True, max_epochs=600, val_interval=10)
+val_cfg = dict()
+test_cfg = dict()
+
+# NOTE: `auto_scale_lr` is for automatically scaling LR
+# based on the actual training batch size.
+# base_batch_size = (8 GPUs) x (128 samples per GPU)
+auto_scale_lr = dict(base_batch_size=1024)

mmpretrain/configs/mobilenet_v3/mobilenet_v3_small_075_8xb128_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+# Refers to https://pytorch.org/blog/ml-models-torchvision-v0.9/#classification
+
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.models.mobilenet_v3_small import *
+    from .._base_.datasets.imagenet_bs128_mbv3 import *
+    from .._base_.default_runtime import *
+
+from mmengine.optim import StepLR
+from torch.nn.modules.batchnorm import BatchNorm2d
+from torch.optim import RMSprop
+
+# model settings
+model.merge(
+    dict(
+        backbone=dict(
+            arch='small_075',
+            norm_cfg=dict(type=BatchNorm2d, eps=1e-5, momentum=0.1)),
+        head=dict(in_channels=432),
+    ))
+
+train_pipeline = [
+    dict(type=LoadImageFromFile),
+    dict(
+        type=RandomResizedCrop,
+        scale=224,
+        backend='pillow',
+        interpolation='bicubic'),
+    dict(type=RandomFlip, prob=0.5, direction='horizontal'),
+    dict(
+        type=AutoAugment,
+        policies='imagenet',
+        hparams=dict(pad_val=[round(x) for x in [103.53, 116.28, 123.675]])),
+    dict(
+        type=RandomErasing,
+        erase_prob=0.2,
+        mode='rand',
+        min_area_ratio=0.02,
+        max_area_ratio=1 / 3,
+        fill_color=[103.53, 116.28, 123.675],
+        fill_std=[57.375, 57.12, 58.395]),
+    dict(type=PackInputs),
+]
+
+test_pipeline = [
+    dict(type=LoadImageFromFile),
+    dict(
+        type=ResizeEdge,
+        scale=256,
+        edge='short',
+        backend='pillow',
+        interpolation='bicubic'),
+    dict(type=CenterCrop, crop_size=224),
+    dict(type=PackInputs),
+]
+
+train_dataloader.merge(dict(dataset=dict(pipeline=train_pipeline)))
+val_dataloader.merge(dict(dataset=dict(pipeline=test_pipeline)))
+test_dataloader = val_dataloader
+
+# schedule settings
+optim_wrapper = dict(
+    optimizer=dict(
+        type=RMSprop,
+        lr=0.064,
+        alpha=0.9,
+        momentum=0.9,
+        eps=0.0316,
+        weight_decay=1e-5))
+
+param_scheduler = dict(type=StepLR, by_epoch=True, step_size=2, gamma=0.973)
+
+train_cfg = dict(by_epoch=True, max_epochs=600, val_interval=10)
+val_cfg = dict()
+test_cfg = dict()
+
+# NOTE: `auto_scale_lr` is for automatically scaling LR
+# based on the actual training batch size.
+# base_batch_size = (8 GPUs) x (128 samples per GPU)
+auto_scale_lr = dict(base_batch_size=1024)

mmpretrain/configs/mobilenet_v3/mobilenet_v3_small_8xb128_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+# Refers to https://pytorch.org/blog/ml-models-torchvision-v0.9/#classification
+
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.models.mobilenet_v3_small import *
+    from .._base_.datasets.imagenet_bs128_mbv3 import *
+    from .._base_.default_runtime import *
+
+from mmengine.optim import StepLR
+from torch.optim import RMSprop
+
+# schedule settings
+optim_wrapper = dict(
+    optimizer=dict(
+        type=RMSprop,
+        lr=0.064,
+        alpha=0.9,
+        momentum=0.9,
+        eps=0.0316,
+        weight_decay=1e-5))
+
+param_scheduler = dict(type=StepLR, by_epoch=True, step_size=2, gamma=0.973)
+
+train_cfg = dict(by_epoch=True, max_epochs=600, val_interval=1)
+val_cfg = dict()
+test_cfg = dict()
+
+# NOTE: `auto_scale_lr` is for automatically scaling LR
+# based on the actual training batch size.
+# base_batch_size = (8 GPUs) x (128 samples per GPU)
+auto_scale_lr = dict(base_batch_size=1024)

mmpretrain/configs/mobilenet_v3/mobilenet_v3_small_8xb16_cifar10.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.models.mobilenet_v3_small import *
+    from .._base_.datasets.cifar10_bs16 import *
+    from .._base_.schedules.cifar10_bs128 import *
+    from .._base_.default_runtime import *
+
+from mmengine.optim import MultiStepLR
+
+# model settings
+model.merge(
+    dict(
+        head=dict(
+            _delete_=True,
+            type=StackedLinearClsHead,
+            num_classes=10,
+            in_channels=576,
+            mid_channels=[1280],
+            act_cfg=dict(type=Hardswish),
+            loss=dict(type=CrossEntropyLoss, loss_weight=1.0),
+            topk=(1, 5))))
+# schedule settings
+param_scheduler.merge(
+    dict(
+        type=MultiStepLR,
+        by_epoch=True,
+        milestones=[120, 170],
+        gamma=0.1,
+    ))
+
+train_cfg.merge(dict(by_epoch=True, max_epochs=200))

mmpretrain/configs/resnet/resnet18_8xb32_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs32 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.resnet18 import *
+    from .._base_.schedules.imagenet_bs256 import *

mmpretrain/configs/simclr/simclr_resnet50_16xb256_coslr_200e_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs32_simclr import *
+    from .._base_.schedules.imagenet_lars_coslr_200e import *
+    from .._base_.default_runtime import *
+
+from mmengine.hooks.checkpoint_hook import CheckpointHook
+from mmengine.optim.optimizer.optimizer_wrapper import OptimWrapper
+
+from mmpretrain.engine.optimizers.lars import LARS
+from mmpretrain.models.backbones.resnet import ResNet
+from mmpretrain.models.heads.contrastive_head import ContrastiveHead
+from mmpretrain.models.losses.cross_entropy_loss import CrossEntropyLoss
+from mmpretrain.models.necks.nonlinear_neck import NonLinearNeck
+from mmpretrain.models.selfsup.simclr import SimCLR
+
+# dataset settings
+train_dataloader.merge(dict(batch_size=256))
+
+# model settings
+model = dict(
+    type=SimCLR,
+    backbone=dict(
+        type=ResNet,
+        depth=50,
+        norm_cfg=dict(type='SyncBN'),
+        zero_init_residual=True),
+    neck=dict(
+        type=NonLinearNeck,  # SimCLR non-linear neck
+        in_channels=2048,
+        hid_channels=2048,
+        out_channels=128,
+        num_layers=2,
+        with_avg_pool=True),
+    head=dict(
+        type=ContrastiveHead,
+        loss=dict(type=CrossEntropyLoss),
+        temperature=0.1),
+)
+
+# optimizer
+optim_wrapper = dict(
+    type=OptimWrapper,
+    optimizer=dict(type=LARS, lr=4.8, momentum=0.9, weight_decay=1e-6),
+    paramwise_cfg=dict(
+        custom_keys={
+            'bn': dict(decay_mult=0, lars_exclude=True),
+            'bias': dict(decay_mult=0, lars_exclude=True),
+            # bn layer in ResNet block downsample module
+            'downsample.1': dict(decay_mult=0, lars_exclude=True)
+        }))
+
+# runtime settings
+default_hooks.checkpoint = dict(
+    type=CheckpointHook, interval=10, max_keep_ckpts=3)

mmpretrain/configs/swin_transformer/swin_base_16xb64_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, LabelSmoothLoss, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_224 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(img_size=224, drop_path_rate=0.5, stage_cfgs=None),
+    head=dict(
+        init_cfg=None,  # suppress the default init_cfg of LinearClsHead.
+        loss=dict(
+            type=LabelSmoothLoss,
+            label_smooth_val=0.1,
+            mode='original',
+            loss_weight=0),
+        topk=None,
+        cal_acc=False),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))
+
+# schedule settings
+optim_wrapper = dict(clip_grad=dict(max_norm=5.0))

mmpretrain/configs/swin_transformer/swin_base_16xb64_in1k_384px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_384 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# schedule settings
+optim_wrapper = dict(clip_grad=dict(max_norm=5.0))

mmpretrain/configs/swin_transformer/swin_large_16xb64_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_224 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(arch='large', img_size=224, stage_cfgs=None),
+    head=dict(in_channels=1536),
+)
+
+# schedule settings
+optim_wrapper = dict(clip_grad=dict(max_norm=5.0))

mmpretrain/configs/swin_transformer/swin_large_16xb64_in1k_384px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_384 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(arch='large'),
+    head=dict(in_channels=1536),
+)
+
+# schedule settings
+optim_wrapper = dict(clip_grad=dict(max_norm=5.0))

mmpretrain/configs/swin_transformer/swin_large_8xb8_cub_384px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.hooks import CheckpointHook, LoggerHook
+from mmengine.model import PretrainedInit
+from torch.optim.adamw import AdamW
+
+from mmpretrain.models import ImageClassifier
+
+with read_base():
+    from .._base_.datasets.cub_bs8_384 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.cub_bs64 import *
+
+# model settings
+checkpoint = 'https://download.openmmlab.com/mmclassification/v0/swin-transformer/convert/swin-large_3rdparty_in21k-384px.pth'  # noqa
+
+model.update(
+    backbone=dict(
+        arch='large',
+        init_cfg=dict(
+            type=PretrainedInit, checkpoint=checkpoint, prefix='backbone')),
+    head=dict(num_classes=200, in_channels=1536))
+
+# schedule settings
+optim_wrapper = dict(
+    optimizer=dict(
+        _delete_=True,
+        type=AdamW,
+        lr=5e-6,
+        weight_decay=0.0005,
+        eps=1e-8,
+        betas=(0.9, 0.999)),
+    paramwise_cfg=dict(
+        norm_decay_mult=0.0,
+        bias_decay_mult=0.0,
+        custom_keys={
+            '.absolute_pos_embed': dict(decay_mult=0.0),
+            '.relative_position_bias_table': dict(decay_mult=0.0)
+        }),
+    clip_grad=dict(max_norm=5.0),
+)
+
+default_hooks = dict(
+    # log every 20 intervals
+    logger=dict(type=LoggerHook, interval=20),
+    # save last three checkpoints
+    checkpoint=dict(type=CheckpointHook, interval=1, max_keep_ckpts=3))

mmpretrain/configs/swin_transformer/swin_small_16xb64_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, LabelSmoothLoss, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_224 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(
+        arch='small', img_size=224, drop_path_rate=0.3, stage_cfgs=None),
+    head=dict(
+        in_channels=768,
+        init_cfg=None,  # suppress the default init_cfg of LinearClsHead.
+        loss=dict(
+            type=LabelSmoothLoss,
+            label_smooth_val=0.1,
+            mode='original',
+            loss_weight=0),
+        topk=None,
+        cal_acc=False),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))
+
+# schedule settings
+optim_wrapper = dict(clip_grad=dict(max_norm=5.0))

mmpretrain/configs/swin_transformer/swin_tiny_16xb64_in1k.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, LabelSmoothLoss, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_224 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(
+        arch='tiny', img_size=224, drop_path_rate=0.2, stage_cfgs=None),
+    head=dict(
+        in_channels=768,
+        init_cfg=None,  # suppress the default init_cfg of LinearClsHead.
+        loss=dict(
+            type=LabelSmoothLoss,
+            label_smooth_val=0.1,
+            mode='original',
+            loss_weight=0),
+        topk=None,
+        cal_acc=False),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))
+
+# schedule settings
+optim_wrapper = dict(clip_grad=dict(max_norm=5.0))

mmpretrain/configs/swin_transformer_v2/swinv2_base_w12_8xb128_in21k_192px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet21k_bs128 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_v2_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(
+        img_size=192, drop_path_rate=0.5, window_size=[12, 12, 12, 6]),
+    head=dict(num_classes=21841),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))
+
+# dataset settings
+data_preprocessor = dict(num_classes=21841)
+
+_base_['train_pipeline'][1]['scale'] = 192  # RandomResizedCrop
+_base_['test_pipeline'][1]['scale'] = 219  # ResizeEdge
+_base_['test_pipeline'][2]['crop_size'] = 192  # CenterCrop

mmpretrain/configs/swin_transformer_v2/swinv2_base_w16_16xb64_in1k_256px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_256 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_v2_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(
+        img_size=256, drop_path_rate=0.5, window_size=[16, 16, 16, 8]),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))

mmpretrain/configs/swin_transformer_v2/swinv2_base_w16_in21k_pre_16xb64_in1k_256px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_256 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_v2_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(
+        img_size=256,
+        window_size=[16, 16, 16, 8],
+        pretrained_window_sizes=[12, 12, 12, 6]),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))

mmpretrain/configs/swin_transformer_v2/swinv2_base_w24_in21k_pre_16xb64_in1k_384px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_384 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_v2_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(
+        window_size=[24, 24, 24, 12], pretrained_window_sizes=[12, 12, 12, 6]))

mmpretrain/configs/swin_transformer_v2/swinv2_base_w8_16xb64_in1k_256px.py

+# Copyright (c) OpenMMLab. All rights reserved.
+# This is a BETA new format config file, and the usage may change recently.
+from mmengine.config import read_base
+from mmengine.model import ConstantInit, TruncNormalInit
+
+from mmpretrain.models import CutMix, Mixup
+
+with read_base():
+    from .._base_.datasets.imagenet_bs64_swin_256 import *
+    from .._base_.default_runtime import *
+    from .._base_.models.swin_transformer_v2_base import *
+    from .._base_.schedules.imagenet_bs1024_adamw_swin import *
+
+# model settings
+model.update(
+    backbone=dict(img_size=256, drop_path_rate=0.5),
+    init_cfg=[
+        dict(type=TruncNormalInit, layer='Linear', std=0.02, bias=0.),
+        dict(type=ConstantInit, layer='LayerNorm', val=1., bias=0.)
+    ],
+    train_cfg=dict(
+        augments=[dict(type=Mixup, alpha=0.8),
+                  dict(type=CutMix, alpha=1.0)]))