SETR_MLA_DeiT_768x768_40k_cityscapes_bs_8.py

_base_ = [
    '../_base_/models/setr_mla.py',
    '../_base_/datasets/cityscapes_768x768.py', '../_base_/default_runtime.py',
    '../_base_/schedules/schedule_40k.py'
]

model = dict(
    backbone=dict(img_size=768, pos_embed_interp=True, drop_rate=0., mla_channels=256,
                  model_name='deit_base_distilled_path16_384', mla_index=(2, 5, 8, 11), embed_dim=768, depth=12, num_heads=12),
    decode_head=dict(img_size=768, mla_channels=256,
                     mlahead_channels=128, num_classes=19),
    auxiliary_head=[
        dict(
            type='VIT_MLA_AUXIHead',
            in_channels=256,
            channels=512,
            in_index=0,
            img_size=768,
            num_classes=19,
            align_corners=False,
            loss_decode=dict(
                type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
        dict(
            type='VIT_MLA_AUXIHead',
            in_channels=256,
            channels=512,
            in_index=1,
            img_size=768,
            num_classes=19,
            align_corners=False,
            loss_decode=dict(
                type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
        dict(
            type='VIT_MLA_AUXIHead',
            in_channels=256,
            channels=512,
            in_index=2,
            img_size=768,
            num_classes=19,
            align_corners=False,
            loss_decode=dict(
                type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
        dict(
            type='VIT_MLA_AUXIHead',
            in_channels=256,
            channels=512,
            in_index=3,
            img_size=768,
            num_classes=19,
            align_corners=False,
            loss_decode=dict(
                type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
    ])

optimizer = dict(lr=0.002, weight_decay=0.0,
                 paramwise_cfg=dict(custom_keys={'head': dict(lr_mult=10.)})
                 )

crop_size = (768, 768)
test_cfg = dict(mode='slide', crop_size=crop_size, stride=(512, 512))
find_unused_parameters = True
data = dict(samples_per_gpu=1)