pytorch_connectomics/tutorials/minimal.yaml at master · PytorchConnectomics/pytorch_connectomics · 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
experiment_name: minimal_demo
description: Minimal training config for quick startup and UX onboarding.

default:
  model:
    arch:
      type: monai_unet
    in_channels: 1
    out_channels: 1
    input_size: [32, 64, 64]
    output_size: [32, 64, 64]
    monai:
      filters: [16, 32, 64]
      num_res_units: 1
      kernel_size: 3
      dropout: 0.0
    loss:
      losses:
        - function: DiceLoss
          weight: 1.0
          pred_slice: "0:1"
          target_slice: "0:1"
  data:
    train:
      image: random://minimal/train_image
      label: random://minimal/train_label
    val:
      image: random://minimal/val_image
      label: random://minimal/val_label
    dataloader:
      batch_size: 1
      patch_size: [32, 64, 64]

train:
  optimization:
    max_epochs: 1
    n_steps_per_epoch: 1
    precision: "32"
    optimizer:
      name: AdamW
      lr: 1e-4
  monitor:
    checkpoint:
      dirpath: outputs/minimal/checkpoints
      save_top_k: 1
      save_last: true
  system:
    num_gpus: 1
    num_workers: 2
    seed: 42