1-shot_colon.py

_base_ = [
    '../datasets/colon.py',
    '../schedules/adamw_inverted_cosine_lr.py',
    'mmpretrain::_base_/default_runtime.py',
    '../custom_imports.py',
]

lr = 5e-4
train_bs = 8
dataset = 'colon'
model_name = 'swin'
exp_num = 1
nshot = 1

run_name = f'{model_name}_bs{train_bs}_lr{lr}_exp{exp_num}_'
work_dir = f'work_dirs/colon/{nshot}-shot/{run_name}'

model = dict(
    type='ImageClassifier',
    backbone=dict(
        type='PromptedSwinTransformer',
        prompt_length=5,
        arch='base',
        img_size=384,
        init_cfg=dict(
            type='Pretrained',
            checkpoint=
            'https://download.openmmlab.com/mmclassification/v0/swin-transformer/convert/swin_base_patch4_window12_384_22kto1k-d59b0d1d.pth',
            prefix='backbone',
        ),
        stage_cfgs=dict(block_cfgs=dict(window_size=12))),
    neck=None,
    head=dict(
        type='LinearClsHead',
        num_classes=2,
        in_channels=1024,
        loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
    )
)

train_dataloader = dict(
    batch_size=train_bs,
    dataset=dict(
        ann_file=f'data_anns/MedFMC/{dataset}/{dataset}_{nshot}-shot_train_exp{exp_num}.txt',
    )
)

val_dataloader = dict(
    batch_size=32,
    dataset=dict(
        ann_file=f'data_anns/MedFMC/{dataset}/{dataset}_{nshot}-shot_val_exp{exp_num}.txt',
    )
)

test_dataloader = dict(
    batch_size=8,
    dataset=dict(ann_file=f'data_anns/MedFMC/{dataset}/test_WithLabel.txt'),
)

visualizer = dict(type='Visualizer', vis_backends=[dict(type='TensorboardVisBackend')])

train_cfg = dict(by_epoch=True, val_interval=25, max_epochs=500)

auto_scale_lr = dict(base_batch_size=1024)

randomness = dict(seed=0)

default_hooks = dict(
    checkpoint=dict(interval=250, max_keep_ckpts=1, save_best="accuracy/top1", rule="greater"),
    logger=dict(interval=10),
)