GLM/params.yaml at master · Emvlt/GLM · 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
demo:
  data:
    raw_path: src/glm/datasets/raw/2detect
    processed_path: src/glm/datasets/processed/2detect

data:
  raw_path: /home/emilien/datasets/raw/2detect
  processed_path: /home/emilien/datasets/processed/2detect

experiment:
  name: end-to-end reconstruction
  run_id: run_001
  description: "GLM with 16 channels"

pretrain_parameters:
  hyperparameters:
    learning_rate: 5e-4
    epochs: 1
    batch_size: 8
    num_workers: 2
    downsampling: 1
  active_model: GLM
  models:
    GLM:
      # CNN parameters
      kernel_size: 7
      kernel_dimension: 1
      padding: 3
      n_channels: 16
      # GNN parameters
      normalize: true
      improved: true
      cached: false
      add_self_loops: true
      aggr: add
    sinogram_CNN:
      # CNN parameters
      kernel_size: 7
      kernel_dimension: 2
      padding: 3
      n_channels: 24

train_parameters:
  hyperparameters:
    learning_rate: 5e-5
    epochs: 40
    batch_size: 8
    num_workers: 2
    downsampling: 1
  active_pseudo_inverse: filtered_backprojection
  pseudo_inverse:
    backprojection:
      n_voxels: 1024
      impl: pytorch
    filtered_backprojection:
      n_voxels: 1024
      impl: pytorch
      filter_name: Ram-Lak
  active_image_model: image_CNN
  image_models:
    image_CNN:
      n_channels: 32
      in_channels: 1
      out_channels: 1
      kernel_dimension: 2
      kernel_size: 7
      padding: 3