HistAug/src/histaug/config/Histaug_conch.yaml at master · MICS-Lab/HistAug · 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
General:
    comment:
    seed: 2025
    precision: 16-mixed
    epochs: &epoch 2000
    grad_acc: 1
    server: train # 'train' or 'test'
    log_path: logs/
    strategy: "auto"
    devices: 1
    accelerator: "gpu"
    ckpt_path_resume_training: null # Path to a checkpoint to resume training from. Set to null to start training from scratch.
Data:
    dataset_name: patch_dataset
    shuffle_data: True
    data_path: /path/to/WSI_data # Root folder containing your WSIs.
    data_path_extension: svs # Extension of the WSI files (e.g., svs, ndpi, tiff).
    patching:
        blca: /path/to/BLCA/patches # Default path to BLCA patches (extracted using CLAM toolbox).
        brca: /path/to/BRCA/patches # Default path to BRCA patches (extracted using CLAM toolbox).
        lusc: /path/to/LUSC/patches # Default path to LUSC patches (extracted using CLAM toolbox).

    Transforms:
        transform_class: PatchAugmentation
        parameters:
            rotation: 0.75 # Probability to apply a rotation.
            h_flip: 0.75 # Probability to apply a horizontal flip.
            v_flip: 0.75 # Probability to apply a vertical flip.
            erosion: 0.75 # Probability to apply an erosion.
            dilation: 0.75 # Probability to apply a dilation.
            crop: 0.75 # Probability to apply a crop.
            gaussian_blur: 0.75 # Probability to apply gaussian blurring.
            hed: [-0.5, 0.5] # Range of HED transform parameters.
            hue: [-0.5, 0.5] # Range of hue transform parameters.
            brightness: [-0.5, 0.5] # Range of brightness transform parameters.
            contrast: [-0.5, 0.5]  # Range of contrast transform parameters.
            saturation: [-0.5, 0.5] # Range of saturation transform parameters.
            gamma: [-0.5, 0.5] # Range of gamma transform parameters.

    train_dataloader:
        batch_size: 32
        num_workers: 25

    test_dataloader:
        batch_size: 32
        num_workers: 10

Foundation_model:
    name: CONCH
    ckpt_path: null # Path to pytorch_model.bin. If null, the default checkpoint will be used from Hugging Face cache (~/.cache).

Model:
    name: histaug_model
    input_dim: 512
    chunk_size: 4
    depth: 16
    num_heads: 8
    mlp_ratio: 4
    use_transform_pos_embeddings: true
    positional_encoding_type: learnable

Optimizer:
    name: AdamW
    parameters:
        lr: 5e-4
        weight_decay: 1e-5

Scheduler:
    name: null # No learning rate scheduler

    # Example of learning rate scheduler :
    # name: CosineAnnealingLR
    # parameters:
    #     T_max: 500
    #     eta_min: 5e-4
    #     last_epoch: -1
Loss:
    base_loss: MSELoss

    # Example using combination of losses
    # base_loss: [CosineSimilarityLoss, SmoothL1Loss]
    # loss_weights: [1, 1]