-
Notifications
You must be signed in to change notification settings - Fork 0
/
template_cnn.yml
37 lines (34 loc) · 1.53 KB
/
template_cnn.yml
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
# PREPROCESS CONFIGURATION
preprocess:
directory: # path to data directory
infotable: # name of infotable in data directory
sample: # number of total samples
time_start: # time start of observation
time_stop: # time stop of observation
binning: # map binning
smoothing: # Gaussian smoothing kernel
mode: # pipeline mode: clean, localise
norm_value: # 1 = single map normalisation; 0 = no normalisation; float = dataset overall normalisation value
stretch: # add stretch to normalisation;
saveas: # name of dataset to save
# CNN CONFIGURATION
cnn:
directory: # path to dataset directory
dataset: # dataset name
saveas: # name of model to save
mode: # CNN modality: clean, localise
reshape: # reshape input
split: # slipt ratio (train percentage)
batch_sz: # training batch size
epochs: # training epochs
learning: # model learning rate
shuffle: # training with shuffle
layers:
# CLEANER AND REGRESSOR LAYERS
conv_filter: # Conv2D filter
conv_kernel: # Conv2D kernel
sampling_kernel: # AvgPool kernel
# REGRESSOR LAYERS ONLY
number_convs: # Number of Conv2D layers after the 1st one
dropout: # Dropout rate
dense: # Dense layer size