ssl_thymus_quantification/train.yaml at public_release · AIM-Harvard/ssl_thymus_quantification · 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
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
trainer:
  _target_: pytorch_lightning.Trainer
  benchmark: True
  max_epochs: 500
  check_val_every_n_epoch: 1000
  accelerator: gpu
  strategy: ddp_find_unused_parameters_true
  enable_model_summary: False
  log_every_n_steps: 1
  fast_dev_run: False
  num_sanity_val_steps: 0
  precision: 16-mixed
  devices: 3
  detect_anomaly: False
  sync_batchnorm: True
  logger: False
  callbacks:
    - _target_: pytorch_lightning.callbacks.ModelCheckpoint
      dirpath:  /mnt/data16_r2d6/SSLThymus/training/checkpoints/ssl_thymus_idc_pretrain_swav
      verbose: True
      save_last: True
      every_n_epochs: 1
      save_on_train_epoch_end: True

    - _target_: lighter.callbacks.Freezer
      name_starts_with: ["backbone.conv1", "backbone.layer1", "backbone.layer2"]

network:
  _target_: project.models.swav.SwaV
  num_ftrs: 4096
  out_dim: 128
  n_prototypes: 300
  n_queues: 0
  n_steps_frozen_prototypes: 50
  queue_length: 768
  start_queue_at_epoch: 15
  backbone:
      _target_: monai.networks.nets.resnet.resnet50
      pretrained: False
      n_input_channels: 1
      widen_factor: 2
      conv1_t_stride: 2
      feed_forward: False

optimizer:
  _target_: torch.optim.SGD
  params: "$@network.parameters()"
  lr: 0.6
  weight_decay: 1.0e-6
  momentum: 0.9

scheduler:
  _target_: torch.optim.lr_scheduler.CosineAnnealingLR
  optimizer: "@optimizer"
  eta_min: 0.0006
  T_max: 1000

model:
  _target_: project.system.LighterSystem
  network: "@network"
  criterion:
    _target_: project.losses.swav_loss.SwaVLoss
    temperature: 0.1
    sinkhorn_gather_distributed: True
    sinkhorn_epsilon: 0.03
  optimizer: "@optimizer"
  scheduler: "@scheduler"
  train_metrics: null
  val_metrics: null
  test_metrics: null

train_dataloader:
  _target_: torch.utils.data.DataLoader
  batch_size: 128
  pin_memory: True
  drop_last: True
  num_workers: 6
  dataset:
    _target_: monai.data.CSVDataset
    src: /mnt/data16_r2d6/SSLThymus/processed_patches.csv
    transform:
        _target_: monai.transforms.Compose
        transforms:
          - _target_: monai.transforms.LoadImaged
            keys: ["positive"]
            image_only: True
            reader: "NumpyReader"

          - _target_: monai.transforms.EnsureChannelFirstd
            keys: ["positive"]

          - _target_: project.transforms.replicate.Replicated
            keys: ["positive"]
            transforms:
              - _target_: monai.transforms.Compose
                transforms:
                  - _target_: project.transforms.random_resized_crop.RandomResizedCrop3D
                    size: [75, 50, 50]
                    scale: [0.3, 1]
                  - _target_: monai.transforms.RandAffine
                    prob: 0.5
                    rotate_range: 0.1745
                    shear_range: 0.1
                    padding_mode: zeros
                  - _target_: monai.transforms.RandHistogramShift
                    prob: 0.5
                  - _target_: monai.transforms.RandGaussianSmooth
                    prob: 0.5
                  - _target_: monai.transforms.SpatialPad
                    spatial_size: [75, 50, 50]
              - _target_: monai.transforms.Compose
                transforms:
                  - _target_: project.transforms.random_resized_crop.RandomResizedCrop3D
                    size: [75, 50, 50]
                    scale: [0.3, 1]
                  - _target_: monai.transforms.RandAffine
                    prob: 0.5
                    rotate_range: 0.1745
                    shear_range: 0.1
                    padding_mode: zeros
                  - _target_: monai.transforms.RandHistogramShift
                    prob: 0.5
                  - _target_: monai.transforms.RandGaussianSmooth
                    prob: 0.5
                  - _target_: monai.transforms.SpatialPad
                    spatial_size: [75, 50, 50]
              - _target_: monai.transforms.Compose
                transforms:
                  - _target_: project.transforms.random_resized_crop.RandomResizedCrop3D
                    size: [75, 50, 50]
                    scale: [0.3, 1]
                  - _target_: monai.transforms.RandAffine
                    prob: 0.5
                    rotate_range: 0.1745
                    shear_range: 0.1
                    padding_mode: zeros
                  - _target_: monai.transforms.RandHistogramShift
                    prob: 0.5
                  - _target_: monai.transforms.RandGaussianSmooth
                    prob: 0.5
                  - _target_: monai.transforms.SpatialPad
                    spatial_size: [75, 50, 50]
              - _target_: monai.transforms.Compose
                transforms:
                  - _target_: project.transforms.random_resized_crop.RandomResizedCrop3D
                    size: [75, 50, 50]
                    scale: [0.3, 1]
                  - _target_: monai.transforms.RandAffine
                    prob: 0.5
                    rotate_range: 0.1745
                    shear_range: 0.1
                    padding_mode: zeros
                  - _target_: monai.transforms.RandHistogramShift
                    prob: 0.5
                  - _target_: monai.transforms.RandGaussianSmooth
                    prob: 0.5
                  - _target_: monai.transforms.SpatialPad
                    spatial_size: [75, 50, 50]
          - _target_: monai.transforms.SelectItemsd
            keys: ["positive"]
          - _target_: torchvision.transforms.Lambda
            lambd: "$lambda x: (x['positive'], False)"

data:
  _target_: lighter.LighterDataModule
  train_dataloader: "@train_dataloader"
  val_dataloader: "@train_dataloader"

method_args: {}