deepcv/input/input_ae_DA.json at master · rangsimanketkaew/deepcv · 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
{
  "_comment": "Configuration input file",
  "project": {
    "_comment": "Type of neural network to train",
    "name": "Demo",
    "neural_network": "daenn"
  },
  "dataset": {
    "_comment": "Dataset manipulation: dataset splitting and normalization",
    "primary": [
      "DA_dist.npz",
      "DA_angle.npz",
      "DA_torsion.npz",
      "DA_sprint.npz"
    ],
    "secondary": [
      "DA_dist_6carbons.npz"
    ],
    "split": true,
    "split_ratio": 0.8,
    "shuffle": true,
    "normalize_scale": 0.0,
    "max_scale": 1
  },
  "model": {
    "_comment": "Define the optimizer, loss function, number of epochs, and batch size",
    "optimizer": "adam",
    "main_loss": "RMSE",
    "penalty_loss": "RMSE",
    "loss_weights": [1, 0.1],
    "num_epoch": 1000,
    "batch_size": 55,
    "save_every_n_epoch": 100
  },
  "network": {
    "_comment": "Number of neurons and activation function for each hidden layer",
    "units": [32, 8, 2, 8, 32],
    "act_funcs": ["relu", "relu", "relu", "relu", "relu"]
  },
  "performance": {
    "_comment": "Setting for training performance",
    "enable_gpu": true
  },
  "settings": {
    "_comment": "User-defined settings",
    "verbosity": 1,
    "show_summary": true,
    "save_model": true,
    "save_weights": true,
    "save_weights_npz": true,
    "save_biases_npz": true,
    "save_graph": true,
    "save_loss": true,
    "show_loss": true,
    "save_metrics": true,
    "show_metrics": true
  },
  "output": {
    "_comment": "Set name of output files",
    "out_dir": "output",
    "out_tb": "tb",
    "out_model": "model",
    "out_weights": "model.weights.h5",
    "out_weights_npz": "model_weights.npz",
    "out_biases_npz": "model_biases.npz",
    "loss_plot": "loss.png",
    "metrics_plot": "metrics.png"
  }
}