_base_ = './rtmdet_l_8xb32-300e_coco.py'

norm_cfg = dict(type='GN', num_groups=32)
checkpoint = 'https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window12_384_22k.pth'  # noqa
model = dict(
    type='RTMDet',
    data_preprocessor=dict(
        _delete_=True,
        type='DetDataPreprocessor',
        mean=[123.675, 116.28, 103.53],
        std=[58.395, 57.12, 57.375],
        bgr_to_rgb=True,
        batch_augments=None),
    backbone=dict(
        _delete_=True,
        type='SwinTransformer',
        pretrain_img_size=384,
        embed_dims=128,
        depths=[2, 2, 18, 2],
        num_heads=[4, 8, 16, 32],
        window_size=12,
        mlp_ratio=4,
        qkv_bias=True,
        qk_scale=None,
        drop_rate=0.,
        attn_drop_rate=0.,
        drop_path_rate=0.3,
        patch_norm=True,
        out_indices=(1, 2, 3),
        with_cp=True,
        convert_weights=True,
        init_cfg=dict(type='Pretrained', checkpoint=checkpoint)),
    neck=dict(in_channels=[256, 512, 1024], norm_cfg=norm_cfg),
    bbox_head=dict(norm_cfg=norm_cfg))

max_epochs = 100
stage2_num_epochs = 10
interval = 10
base_lr = 0.001

train_cfg = dict(
    max_epochs=max_epochs,
    val_interval=interval,
    dynamic_intervals=[(max_epochs - stage2_num_epochs, 1)])

optim_wrapper = dict(optimizer=dict(lr=base_lr))

# learning rate
param_scheduler = [
    dict(
        type='LinearLR',
        start_factor=1.0e-5,
        by_epoch=False,
        begin=0,
        end=1000),
    dict(
        # use cosine lr from 50 to 100 epoch
        type='CosineAnnealingLR',
        eta_min=base_lr * 0.05,
        begin=max_epochs // 2,
        end=max_epochs,
        T_max=max_epochs // 2,
        by_epoch=True,
        convert_to_iter_based=True),
]

custom_hooks = [
    dict(
        type='EMAHook',
        ema_type='ExpMomentumEMA',
        momentum=0.0002,
        update_buffers=True,
        priority=49),
    dict(
        type='PipelineSwitchHook',
        switch_epoch=max_epochs - stage2_num_epochs,
        switch_pipeline={{_base_.train_pipeline_stage2}})
]