_base_ = 'knet-s3_swin-t_upernet_8xb2-adamw-80k_ade20k-512x512.py'

checkpoint_file = 'https://download.openmmlab.com/mmsegmentation/v0.5/pretrain/swin/swin_large_patch4_window7_224_22k_20220308-d5bdebaf.pth'  # noqa
# model settings
crop_size = (640, 640)
data_preprocessor = dict(
    type='SegDataPreProcessor',
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    bgr_to_rgb=True,
    pad_val=0,
    size=crop_size,
    seg_pad_val=255)
model = dict(
    data_preprocessor=data_preprocessor,
    pretrained=checkpoint_file,
    backbone=dict(
        embed_dims=192,
        depths=[2, 2, 18, 2],
        num_heads=[6, 12, 24, 48],
        window_size=7,
        use_abs_pos_embed=False,
        drop_path_rate=0.4,
        patch_norm=True),
    decode_head=dict(
        kernel_generate_head=dict(in_channels=[192, 384, 768, 1536])),
    auxiliary_head=dict(in_channels=768))

crop_size = (640, 640)
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', reduce_zero_label=True),
    dict(
        type='RandomResize',
        scale=(2048, 640),
        ratio_range=(0.5, 2.0),
        keep_ratio=True),
    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
    dict(type='RandomFlip', prob=0.5),
    dict(type='PhotoMetricDistortion'),
    dict(type='PackSegInputs')
]
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='Resize', scale=(2048, 640), keep_ratio=True),
    # add loading annotation after ``Resize`` because ground truth
    # does not need to do resize data transform
    dict(type='LoadAnnotations', reduce_zero_label=True),
    dict(type='PackSegInputs')
]
train_dataloader = dict(dataset=dict(pipeline=train_pipeline))
val_dataloader = dict(dataset=dict(pipeline=test_pipeline))
test_dataloader = val_dataloader
# In K-Net implementation we use batch size 2 per GPU as default
train_dataloader = dict(batch_size=2, num_workers=2)
val_dataloader = dict(batch_size=1, num_workers=4)
test_dataloader = val_dataloader