neurips_medfm_2023/configs/dinov2-b_vpt/1-shot_chest.py at main · mvnowak/neurips_medfm_2023 · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
_base_ = [
    '../datasets/chest.py',
    '../schedules/chest.py',
    'mmpretrain::_base_/default_runtime.py',
    '../custom_imports.py',
]

lr = 5e-3
vpl = 1
dataset = 'chest'
exp_num = 2
nshot = 1

run_name = f'dinov2-b_{vpl}_bs4_lr{lr}_{nshot}-shot_{dataset}_exp{exp_num}'
work_dir = f'work_dirs/chest/{nshot}-shot/{run_name}'

model = dict(
    type='ImageClassifier',
    backbone=dict(
        type='PromptedViT',
        prompt_length=vpl,
        layer_scale_init_value=1e-5,
        out_type='avg_all',
        img_size=518,
        patch_size=14,
        arch='base',
        init_cfg=dict(
            type='Pretrained',
            checkpoint=
            'https://download.openmmlab.com/mmpretrain/v1.0/dinov2/vit-base-p14_dinov2-pre_3rdparty_20230426-ba246503.pth',
            prefix='backbone'),
        ),
    neck=None,
    head=dict(
        type='MultiLabelLinearClsHead',
        num_classes=19,
        in_channels=768,
    ))

train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='NumpyToPIL', to_rgb=True),
    dict(type='torchvision/RandomAffine', degrees=(-15, 15), translate=(0.05, 0.05), fill=128),
    dict(type='PILToNumpy', to_bgr=True),
    dict(
        type='RandomResizedCrop',
        scale=518,
        crop_ratio_range=(0.9, 1.0),
        backend='pillow',
        interpolation='bicubic'),
    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
    dict(type='PackInputs'),
]

test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='Resize', scale=518, backend='pillow', interpolation='bicubic'),
    dict(type='PackInputs'),
]

train_dataloader = dict(
    batch_size=1,
    dataset=dict(
        ann_file=f'data_anns/MedFMC/{dataset}/{dataset}_{nshot}-shot_train_exp{exp_num}.txt',
        pipeline=train_pipeline),
)

val_dataloader = dict(
    batch_size=2,
    dataset=dict(
        ann_file=f'data_anns/MedFMC/{dataset}/{dataset}_{nshot}-shot_val_exp{exp_num}.txt',
        pipeline=test_pipeline),
)

test_dataloader = dict(
    batch_size=2,
    dataset=dict(
        ann_file=f'data_anns/MedFMC/{dataset}/test_WithLabel.txt',
        pipeline=test_pipeline),
)

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

default_hooks = dict(
    checkpoint=dict(type='CheckpointHook', interval=1, max_keep_ckpts=1, save_best="auto"),
    logger=dict(interval=50),
)

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